Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Review - Intervention

Body weight management in overweight and obese breast cancer survivors

Authors' declarations of interest

Version published: 11 December 2020 Version history

https://doi.org/10.1002/14651858.CD012110.pub2
Collapse all Expand all

Background

Studies suggest that overweight and obese breast cancer survivors are at increased risk of cancer recurrence and have higher all‐cause mortality. Obesity has an impact on breast cancer survivor's quality of life (QOL) and increases the risk of longer‐term morbidities such as type 2 diabetes mellitus and cardiovascular disease. Many cancer guidelines recommend survivors maintain a healthy weight but there is a lack of evidence regarding which weight loss method to recommend.

Objectives

To assess the effects of different body weight loss approaches in breast cancer survivors who are overweight or obese (body mass index (BMI) ≥ 25 kg/m2).

Search methods

We carried out a search in the Cochrane Breast Cancer Group's (CBCG's) Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL, Issue 6), MEDLINE (2012 to June 2019), Embase (2015 to June 2019), the World Health Organisation International Clinical Trials Registry Platform (WHO ICTRP) and Clinicaltrials.gov on 17 June 2019. We also searched Mainland Chinese academic literature databases (CNKI), VIP, Wan Fang Data and SinoMed on 25 June 2019. We screened references in relevant manuscripts.

Selection criteria

We included randomised controlled trials (RCTs), quasi‐RCTs and randomised cross‐over trials evaluating body weight management for overweight and obese breast cancer survivors (BMI ≥ 25 kg/m2). The aim of the intervention had to be weight loss.

Data collection and analysis

Two review authors independently performed data extraction and assessed risk of bias for the included studies, and applied the quality of the evidence using the GRADE approach. Dichotomous outcomes were analysed as proportions using the risk ratio (RR) as the measure of effect. Continuous data were analysed as means with the measure of effect being expressed as the mean differences (MDs) between treatment groups in change from baseline values with 95% confidence intervals (CIs), when all studies reported exactly the same outcomes on the same scale. If similar outcomes were reported on different scales the standardised mean difference (SMD) was used as the measure of effect. Quality of life data and relevant biomarkers were extracted where available.

Main results

We included a total of 20 studies (containing 23 intervention‐comparisons) and analysed 2028 randomised women. Participants in the experimental groups received weight loss interventions using the core element of dietary changes, either in isolation or in combination with other core elements such as 'diet and exercise', 'diet and psychosocial support' or 'diet, exercise and psychosocial support'. Participants in the controls groups either received usual care, written materials or placebo, or wait‐list controls. The duration of interventions ranged from 0.5 months to 24 months. The duration of follow‐up ranged from three months to 36 months.

There were no time‐to‐event data available for overall survival, breast cancer recurrence and disease‐free survival. There was a relatively small amount of data available for breast cancer recurrence (281 participants from 4 intervention‐comparisons with 14 recurrence events; RR 1.95, 95% CI 0.68 to 5.60; low‐quality evidence) and the analysis was likely underpowered.

Overall, we found low‐quality evidence that weight loss interventions for overweight and obese breast cancer survivors resulted in a reduction in body weight (MD: ‐2.25 kg, 95% CI: ‐3.19 to ‐1.3 kg; 21 intervention‐comparisons; 1751 women), body mass index (BMI) (MD: ‐1.08 kg/m2, 95% CI: ‐1.61 to ‐0.56 kg/m2; 17 intervention‐comparisons; 1353 women), and waist circumference (MD:‐1.73 cm, 95% CI: ‐3.17 to ‐0.29 cm; 13 intervention‐comparisons; 1193 women), and improved overall quality of life (SMD: 0.74; 95% CI: 0.20 to 1.29; 10 intervention‐comparisons; 867 women). No increase was seen in adverse events for women in the intervention groups compared to controls (RR 0.94, 95% CI: 0.76 to 1.17; 4 intervention‐comparisons; 394 women; high‐quality evidence). Subgroup analyses revealed that decreases in body weight, BMI and waist circumference were present in women regardless of their ethnicity and menopausal status.

Multimodal weight loss interventions (which referred to 'diet, exercise and psychosocial support') appeared to result in greater reductions in body weight (MD: ‐2.88 kg, 95% CI: ‐3.98 to ‐1.77 kg; 13 intervention‐comparisons; 1526 participants), BMI (MD: ‐1.44 kg/m2, 95% CI: ‐2.16 to ‐0.72 kg/m2; 11 studies; 1187 participants) and waist circumference (MD:‐1.66 cm, 95% CI: ‐3.49 to ‐0.16 cm; 8 intervention‐comparisons; 1021 participants) compared to dietary change alone, however the evidence was low quality.

Authors' conclusions

Weight loss interventions, particularly multimodal interventions (incorporating diet, exercise and psychosocial support), in overweight or obese breast cancer survivors appear to result in decreases in body weight, BMI and waist circumference and improvement in overall quality of life. There was no increase in adverse events. There is a lack of data to determine the impact of weight loss interventions on survival or breast cancer recurrence. This review is based on studies with marked heterogeneity regarding weight loss interventions. Due to the methods used in included studies, there was a high risk of bias regarding blinding of participants and assessors.
Further research is required to determine the optimal weight loss intervention and assess the impact of weight loss on survival outcomes. Long‐term follow‐up in weight loss intervention studies is required to determine if weight changes are sustained beyond the intervention periods.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Weight loss programmes for overweight and obese breast cancer survivors: what are their benefits and harms, and do they help survivors to live longer?

What is a healthy weight?

Body mass index (BMI) assesses whether people are a healthy weight for their height. A BMI of 18 to 25 shows a healthy weight, a BMI over 25 indicates being overweight, and a BMI over 30 indicates obesity.

Breast cancer and weight

People with a BMI over 25 are more likely to develop a recurrence of their breast cancer. Obesity can also affect people's quality of life (well‐being) and can lead to serious and life‐threatening conditions, including type 2 diabetes, coronary heart disease and stroke. After successful treatment for breast cancer, people with a BMI over 25 are advised to lose weight.

Losing weight

The most common method for losing weight is to reduce the number of calories eaten and to increase physical activity. A healthy, reduced‐calorie diet and regular exercise may be combined with psychosocial support. Some weight loss programmes include all three elements.

Why we did this Cochrane Review

We wanted to identify which weight‐loss programmes work best to help overweight and obese breast cancer survivors to lose weight; and whether the programmes had advantages or unwanted effects.

What did we do?

We searched for studies that assessed weight loss programmes in survivors of early‐stage breast cancer who had a BMI over 25 and no evidence that their cancer had returned. We looked for randomised controlled studies, in which the programmes people received were decided at random. This type of study usually gives the most reliable evidence about the effects of a programme.

We wanted to know how weight loss programmes affected:
‐how long people lived;
‐whether their breast cancer returned;
‐the length of time before the cancer returned;
‐how many people died;
‐body weight;
‐measurements of waist size;
‐people's quality of life (well‐being); or
‐had any unwanted effects.

Search date: we included evidence published up to June 2019.

What we found

We found 20 relevant studies in 2028 women. The studies compared participation in a weight‐loss programme to not participating in one but receiving usual care, a placebo (dummy) treatment, a different type of weight‐loss programme, written information, or being on a waiting list instead. All the programmes included dietary changes; some combined these with exercise or psychosocial support, or both.

Most studies were conducted in the USA. The weight loss programmes lasted from two weeks to two years; the people participating were followed for three months to 36 months after starting their programme.

None of the studies reported results for: how long people lived; or the length of time before their cancer returned, or how many people died. Few studies reported about the effect of weight loss programmes on the return of breast cancer.

What are the results of our review?

Compared with those not participating in a weight loss programme, breast cancer survivors with a BMI over 25 who take part in one may:
‐lose more body weight; 
‐have greater reductions in their waist size and BMI; and
‐improve their well‐being.

Taking part in a weight loss programme did not cause more unwanted effects.

Programmes combining diet with exercise or psychosocial support, or both, seemed to reduce body weight and waist size more than programmes based on dietary changes alone.

Our confidence in these results

Our confidence in these results is generally low. We identified limitations in the ways that some of the studies were designed and conducted, and the people taking part and those assessing them knew who received which treatments, which could have affected the study results.

Conclusions

Weight loss programmes may help overweight and obese breast cancer survivors to lose weight, reduce their BMI and waist size, and may improve their quality of life, without increasing unwanted effects. We did not find evidence about whether they could help people live longer, or delay the return of breast cancer.

We need more studies to find out which weight loss programmes work best to help breast cancer survivors lose weight, and whether this helps them to live longer.

Authors' conclusions

Implications for practice

This review identified insufficient data to reach firm conclusions regarding survival outcomes and breast cancer recurrence after body weight interventions in overweight and obese breast cancer survivors. Whilst there was sufficient data to perform meta‐analyses on breast cancer recurrence (which suggested no evidence of effect), the total number of included intervention‐comparisons, participants and recurrence events were very small (and thus likely underpowered), and the quality of the evidence was low. The effect of these interventions on survival outcomes and cancer recurrence remain unclear and clinical recommendations cannot be made when considering these primary outcomes in isolation. Our secondary outcomes suggest that these body weight interventions compared favourably to controls (which included usual care) and resulted in significant improvements in anthropometric outcomes (change in body weight, waist circumference, body mass index (BMI)) and various aspects of quality of life (QOL) (including overall QOL, physical subscales and mental health subscales), and were safe. Whilst the quality of the evidence was high for safety (adverse events), it was low for the remaining outcomes (change in body weight, waist circumference, overall QOL) due to a high risk of bias and substantial heterogeneity. In contrast, there appeared to be no evidence of effect of these weight loss interventions on other QOL subscales (social subscales, emotional subscales and anxiety and depression subscales). There also appeared to be no evidence of effect for analysed biomarkers, with the exception of triglycerides and leptin. However biomarker analysis may have been underpowered as the interventions may not have been of sufficient duration and/or intensity to cause a detectable decrease in biomarkers. There was some evidence to suggest using a combination of different intervention modalities (e.g. 'diet and exercise' or 'diet and exercise and psychosocial support') may lead to a greater effect than an exclusively 'diet' intervention. However, there was a wide range of intervention programmes using multimodal interventions, and it is not known which particular intervention programme confers the most benefit. Thus, whilst the overall quality of evidence was low and the effects on survival and recurrence are unclear, the interventions were safe and were associated with favourable improvements in anthropometric outcomes and various aspects of QOL. Therefore clinicians may consider utilising such interventions in their clinical practice, although no definite recommendations can be made from this current review.

Implications for research

A major gap in the literature is the lack of studies assessing survival outcomes (overall survival, breast cancer‐specific survival and disease‐free survival) and very limited data regarding breast cancer recurrence in overweight and obese breast cancer survivors undergoing a body weight management intervention with the aim of weight loss. There is a need for adequately powered studies with long‐term follow‐up to assess these survival outcomes, such as the BWEL study. The BWEL study is an ongoing weight loss study with a sample size of 3136 breast cancer participants which will assess survival outcomes (overall survival, invasive disease‐free survival, distant disease‐free survival) at 10‐year follow‐up BWEL. Many biomarker outcomes had insufficient data for meta‐analyses and therefore require further investigation. This includes sex hormones (oestradiol and testosterone), inflammatory markers (C‐reactive protein (CRP), tumour necrosis factor alpha (TNF‐α) and interleukin‐6 (IL‐6), insulin growth factor (IGF‐1) and adiponectin. Further research is required to determine which particular intervention programme leads to the most favourable outcomes, and to determine whether weight loss is sustained after the intervention period.

Summary of findings

Open in table viewer
Summary of findings 1. All weight loss interventions for overweight and obese breast cancer survivors

All weight loss interventions compared to comparator groups for overweight and obese breast cancer survivors

Patient or population: overweight and obese breast cancer survivors
Setting: various settings (individual or group‐based, in‐person or remote, exercise centre or external location)
Intervention: all weight loss interventions (diet, diet & exercise, diet & psychosocial, diet & exercise & psychosocial)
Comparison: control programs (usual care, wait‐list, written materials, placebo or active control)

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(treatment‐ comparisons)

Quality of the evidence
(GRADE)

Comments

Risk with control programs

Risk with all weight loss interventions

Overall survival

‐

‐

‐

‐

‐

No trials reported this outcome as time‐to‐event data.

Breast cancer recurrence

Study population

RR 1.95
(0.68 to 5.60)

281 (4)

⊕⊕⊝⊝
LOW 1

32 per 1,000

62 per 1,000
(22 to 178)

Change in body weight

The mean body weight loss4 was 1.01 kg

MD 2.25 kg lower
(3.19 lower to 1.30 lower)

‐

1751 (21)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.00001, I2 = 69%

Change in BMI

The mean BMI reduction4 was 0.42 kg/m2

MD 1.08 kg/m2 lower (1.61 lower to 0.56 lower)

1353 (17)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.0001, I2 = 84%

Change in waist circumference

The mean waist circumference4 reduction was ‐0.28 cm (i.e. 0.28 cm increase)

MD 1.73 cm lower
(3.17 lower to 0.29 lower)

‐

1193 (13)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.0001, I2 = 73%

Disease‐free survival

‐

‐

‐

‐

‐

No trials reported this outcome.

Adverse events

Study population

RR 0.94
(0.76 to 1.17)

394 (4)

⊕⊕⊕⊕
HIGH

471 per 1,000

443 per 1,000
(358 to 551)

Change in quality of life ‐ overall scales

‐

SMD 0.74 higher
(0.20 higher to 1.29 higher)

‐

867 (10)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.00001, I2 = 89%; An SMD of 0.74 is a moderate sized effect according to Cohen’s interpretation of effect sizes.

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; MD: mean difference; RR: Risk ratio.

GRADE Working Group grades of evidence

High quality: further research is very unlikely to change our confidence in the estimate of effect.

Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality: we are very uncertain about the estimate.

1Quality of evidence downgraded two levels for 'imprecision' because 95% CI for risk ratio estimate suggests that the intervention might reduce the risk of progression by up to 32% or increase the risk of progression by up to 460% and risk ratio estimate is based on only 14 events in 281 patients.
2Quality of evidence downgraded one level for 'risk of bias' because >20% attrition of randomised participants for measurement of this outcome at follow‐up and because most studies were unblinded for participants, personnel and assessment of patient‐reported outcomes.
3Quality of evidence downgraded one level for 'inconsistency' because I‐squared statistic suggests substantial heterogeneity.
4Calculated as inverse variance weighted average of control group measurements.

Open in table viewer
Summary of findings 2. Weight loss interventions involving diet, exercise and psychosocial support for overweight and obese breast cancer survivors

Weight loss interventions involving all three diet, exercise and psychosocial support for overweight and obese breast cancer survivors

Patient or population: overweight and obese breast cancer survivors
Setting: various settings (individual or group‐based, in‐person or remote, exercise centre or external location)
Intervention: weight loss interventions involving the following components: diet, exercise and psychosocial support
Comparison: control programs (usual care, wait‐list, written materials or placebo)

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(treatment‐ comparisons)

Quality of the evidence
(GRADE)

Comments

Risk with control programs

Risk with All weight loss interventions

Overall survival

‐

‐

‐

‐

‐

No trials reported this outcome as time‐to‐event data.

Breast cancer recurrence

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

Change in body weight

The mean body weight loss3 was 0.97 kg

MD 2.88 kg lower
(3.98 lower to 1.77 lower)

‐

1526 (13)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 69%.

Change in BMI

The mean BMI reduction3 was 0.41 kg/m2

MD 1.44 kg/m2 lower (2.16 lower to 0.72 lower)

1187 (11)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 89%.

Change in waist circumference

The mean waist circumference3 reduction was ‐0.33 cm (.i.e. 0.33 cm increase)

MD 1.66 cm lower
(3.49 lower to 0.16 lower)

‐

1021 (8)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 79%.

Disease‐free survival ‐ not reported

‐

‐

‐

‐

‐

No trials reported this outcome.

Adverse events

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

Change in quality of life ‐ overall scales

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence

High quality: further research is very unlikely to change our confidence in the estimate of effect.

Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality: we are very uncertain about the estimate.

1Quality of evidence downgraded one level for 'risk of bias' because > 20% attrition of randomised participants for measurement of this outcome at follow‐up and because most studies were unblinded for participants, personnel and assessment of patient‐reported outcomes.
2Quality of evidence downgraded one level for 'inconsistency' because I2 statistic suggests substantial heterogeneity.
3Calculated as inverse variance weighted average of control group measurements.

Background

Description of the condition

Breast cancer is the second most common cancer worldwide, with >2 million new cases diagnosed in 2018 (Bray 2018). With early diagnosis, and an increase in the use of neoadjuvant and adjuvant chemotherapy, hormonal therapy and immunotherapy, survival rates have continued to improve (IARC of WHO 2012). The five‐year survival rate for breast cancer is now 90% in North America (Siegal 2019), and 87% in England (Office for National Statistics 2013). In light of the rapid increase in the number of breast cancer survivors, there is a new emphasis on the need for appropriate survivorship care (Yu 2014).

In its broadest definition, a person becomes a cancer survivor when they are diagnosed with cancer, and remain a survivor for the rest of their life (Centers for Disease Control and Prevention 2011). Approximately 50% of breast cancer survivors worldwide are classified as overweight or obese (57% to 66% in the USA (Coups 2005; Morimoto 2002; Imayama 2013), 55% in Switzerland (Eichholzer 2012) and 42% in Mexico (Ortiz‐Mendoza 2014)). Many breast cancer survivors gain more body weight after primary neoadjuvant or adjuvant treatment (Arce‐Salinas 2014; Kann 2014; Kim 2013; Vance 2011; Vagenas 2015) and hormonal therapy (Lorizio 2012).

Being overweight or obese is defined as having an abnormal or excessive amount of total body fat, that may affect health status (WHO 2015). It is usually measured by Body Mass Index (BMI (kg/m2)), with overweight BMI classified as ranging from 25 to 29.9 kg/m2, obesity as a BMI of 30 kg/m2 and above, and morbid obesity as a BMI of 40 kg/m2 and above (WHO 1997).

Body fat is composed mostly of adipocytes and other cells including preadipocytes, fibroblasts, vascular endothelial cells and a variety of immune cells. It can store energy, and cushion and insulate the body. It has been recognised as a major endocrine organ (Ferlay 2015; Kershaw 2004) as it produces hormones such as leptin, oestrogen, and the cytokine tumour necrosis factor alpha (TNF‐α), which stimulates insulin secretion leading to insulin resistance (Rock 2013; Su H 2013). Research has demonstrated that breast cancer survivors that achieve a weight loss of < 5% of their initial weight have higher levels of oestrogen and leptin, and lower levels of adiponectin than those who achieve a weight loss of ≥ 5% of their initial weight (Rock 2013).

Studies suggest that women who are overweight or obese are at an increased risk of cancer recurrence, and higher all‐cause mortality (Dignam 2003; Demark‐Wahnefried 2018a; Ewertz 2011). Obesity has been found to increase the risk of total mortality by 17%, and 18% for breast cancer‐specific mortality, for every 5 kg/m2 increment before their cancer diagnosis (Chan 2014). In addition, morbid obesity may be a prognostic factor for diabetes and cardiovascular disease (Vance 2011). Obesity also has a significant impact on a woman's quality of life (QOL) and ability to function in relation to everyday activities (Imayama 2013). Hence, many cancer guidelines recommend survivors maintain a healthy weight (Ligibel 2014; World Cancer Research Fund 2019).

Description of the intervention

A number of interventions have been adopted into clinical practice for breast cancer survivors who are morbidly obese, to reduce body weight and maintain it within a healthy weight range (BMI of 18.5 to 24.9 kg/m2). These include: physical activity programmes (Thomas 2013), dietary changes (Pierce 2009), medication (Goodwin 2011) and bariatric surgery (Philip 2015). The weight loss approach selected needs to be matched to an individual patient's needs, comorbidities, and risk profile. The most common first‐line strategy used for weight loss is comprehensive lifestyle modification. The basic components are to facilitate energy deficiency through increased physical activity and reduced calorie intake, generally with the goal of losing 3% to 5% of initial body weight for at least six months.

Individually‐tailored physical activity programmes generally consist of a combination of resistance or weight load (strength training) with aerobic exercises (such as walking, jogging, running, cycling, swimming, dancing etc.). Recommendations are to undertake at least 150 minutes per week of activity of moderate intensity (Cormie 2018; Rock 2012; Subirats Bayego 2012; U.S. Department of Health and Human Services 2008).

Weight loss diets have been designed to provide a balanced diet with low energy intake of 1200 to 1500 kilocalories per day (kcal/d) for women and 1500 to 1800 kcal/d for men (Jensen 2014) (or 1000 to 1600 kcal/d for a low‐calorie diet (Commonwealth of Australia 2013)) for the management of overweight and obesity in adults. This can be achieved through a low‐fat, high‐fibre diet. A very‐low‐calorie diet (< 800 kcal/d) may be appropriate if supervised in a medical setting. Behavioural, psychological and/or social interventions are often used in conjunction with exercise and dietary interventions.

Physical activity and weight loss dietary programmes are usually administered in medical clinics, at home or in gyms. They can be conducted face‐to‐face, by telephone or through web sites (Rogers 2008).

Pharmacotherapy can be used either as an adjunct to comprehensive lifestyle modification, or in isolation for cancer survivors who cannot participate in lifestyle programmes due, for example, to comorbidities limiting their physical activity. Medications such as sibutramine and orlistat are most commonly used to achieve or maintain weight reduction in those with a BMI of 30 or above, or BMI of 27 or above in the presence of obesity‐related comorbidities (e.g. diabetes, cardiovascular diseases) (Jensen 2014). Sibutramine has been associated with an increased risk of elevated blood pressure and tachycardia (fast heart rate), while orlistat can lead to gastrointestinal side effects, deficiency of fat‐soluble vitamins (e.g. vitamins A, D, E and K), and interactions with other medications, e.g. warfarin (National Health and Medical Research Council 2013).

Bariatric surgery may be considered for individuals with a BMI of at least 40, or a BMI of 35 or above with high‐risk comorbidities (Jensen 2014) who have not responded to behavioural treatments, with or without pharmacotherapy. Bariatric surgery includes a variety of procedures where the aim is to achieve weight loss by reducing the size of the stomach using a gastric band, or removing a portion of the stomach (sleeve gastrectomy or biliopancreatic diversion with duodenal switch), or by resecting and re‐routing the small stomach pouch (gastric bypass surgery) to the small intestine. Bariatric surgical interventions should be performed only in highly‐selected patients and in specialist centres by experienced surgeons. Patients need to be fully informed about the potential risks and side effects. Adverse effects include misplacement of the band, erosion of the gastric wall, port complications, anastomotic leak, wound complications, haemorrhage, pulmonary embolism, deep vein thrombosis, deficiency of nutrients or malnutrition, elevated parathyroid hormone, ventral hernia and, occasionally, death (Jensen 2014).

In this Cochrane Review, we will include randomised controlled clinical trials (RCTs), in which the participants are randomly allocated to one or other of the different treatments under study. We will evaluate several types of comparator interventions such as placebo medications and supportive treatments (e.g. vitamins), as well as evidence‐based positive controlled interventions (e.g. increased physical activity and diet modification regimens).

How the intervention might work

Obesity after a breast cancer diagnosis is known to be a poor prognostic risk factor, particularly for postmenopausal women (Chan 2014). This may be due to changes in energy metabolism secondary to the side effects of chemotherapy treatment (Gadea 2013). There is likely to be higher than normal oestrogen conversion and secretion in excess body fat tissue and less sex hormone binding globulin in the circulation (Siiteri 1987), which will increase stimulation to breast tissue. Elevation of inflammatory cytokines such as TNF‐α, interleukin‐6 (IL‐6) and adipokines such as leptin (Khandekar 2011) in adipose tissue can activate cancer cells by activating oncogenic transcription in breast tissue, and a fall in adiponectin decreases the inhibition of proliferation and metastasis of breast tumour cells. Finally, high insulin levels and insulin resistance (Oh 2011) may exacerbate the loco‐regional metabolic microenvironment leading to an imbalance in homeostasis, with a depletion of oxygen, and energy dysfunction in localised breast lesions, which are considered to be an ideal microenvironment for tumour recurrence.

Under normal circumstances, obesity occurs when energy expenditure is less than the energy intake over a period of time (Davoodi 2013). However morbidly obese breast cancer survivors are often characterised by fat gain and loss of lean tissue (Vance 2011). The loss of muscle mass in the setting of increased body fat is known as sarcopenic obesity. This is a multifactorial condition. In addition to the metabolic and neuroendocrine alterations that may be induced by chemotherapy and genetics, other lifestyle‐dependent factors such as inactivity are likely to be important risk determinants (Davoodi 2013). Other causes such as psycho‐social and environmental factors may also play a role in the excess accumulation of adipose tissue (Deusinger 2012; Mastorakos 2010; Nahas 2012; Waxler‐Morrison 1991).

Physical activity promotes blood circulation and oxygen concentration, and reduces the concentration of plasma cytokines and inflammatory factors (nuclear factor kappa B (NF‐κB), IL‐6, C‐reactive protein (CRP)), insulin‐like growth factor (IGF); (Imayama 2013; Jones 2013), leptin (Iantorno 2014), and hormones (insulin and oestrogens) (Borer 2014; Rock 2004). Exercise can also improve body composition (Guinan 2013). Studies of dietary modification suggest that exercise has positive effects on loss of excess body weight and weight loss maintenance for breast cancer survivors (Carpenter 2012; Reeves 2014). These effects may be helpful in improving the local breast microenvironment. In addition, management of body weight may have preventative effects on secondary events associated with breast cancer such as type 2 diabetes, cardiovascular disease, dyslipidaemia and other comorbidities (Jensen 2014; Patnaik 2011). Loss of body weight can also have a positive impact on psychosocial well being (Demark‐Wahnefried 2012b; Dignam 2003; Imayama 2013).

Over‐the‐counter medications for obesity in the non‐cancer population may reduce excess body weight by decreasing macronutrient absorption (orlistat) and suppressing appetite (sibutramine) (National Health and Medical Research Council 2013). Bariatric surgery results in weight loss through limiting food intake or diminishing the area available for digestion and absorption in the gastrointestinal tract (Bordalo 2011). However, there is a lack of evidence for use of these weight loss methods in breast cancer survivors.

Why it is important to do this review

Changes in body weight associated with cancer treatment have been noted for decades (Dixon 1978). Studies indicate that body weight changes can occur in either direction, with weight gain or loss (Vagenas 2015). This heterogeneity may be associated with complex factors in breast cancer survivors, including: genetic predisposition (Slattery 2015), socio‐demographic factors (Sedjo 2014; Thompson 2014), menopausal status (Irwin 2007), hormone receptor status (Ewertz 2012), clinical presentation (tumour size, histological grade and degree of differentiation, lymph node metastasis) and treatment modality.

Cancer survivors gain weight mainly through sedentary and inactive lifestyles, with a peak in the third year, when followed for six years, after diagnosis (Makari‐Judson 2014; Vagenas 2015). Studies suggest that body weight loss of more than 5% is feasible for obese or overweight breast cancer survivors (Davoodi 2013; Vitolins 2014). However, many studies have been observational in design (Makari‐Judson 2014), with a paucity of high‐quality evidence from clinical interventions to guide effective body weight management for cancer survivors. Consequently, body weight reduction is frequently not addressed in overweight breast cancer survivors (Chan 2014). There is a need to synthesise the evidence in terms of which modality to recommend, at what intensity, and for which group of participants. In this review, we aim to assess the benefits, risks and efficacy of different body weight loss approaches in overweight breast cancer survivors in order to guide survivors, clinicians, and policy makers associated with survivorship care.

Objectives

To assess the effects of different body weight loss approaches in breast cancer survivors who are overweight or obese (body mass index (BMI) ≥ 25 kg/m2).

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs), quasi‐RCTs and randomised cross‐over trials that evaluated body weight loss for overweight and obese breast cancer survivors. No restrictions were based on the language of the publication. We translated studies published in other languages into English.

Types of participants

This review included overweight or obese breast cancer survivors diagnosed with early stage breast cancer with no evidence of a recurrence of their cancer. Participants could have had surgery and/or be receiving or have received adjuvant chemotherapy and/or radiotherapy, adjuvant hormonal treatment and/or targeted therapies such as trastuzumab for women with HER2+ status. Whilst the protocol stated that ductal carcinoma in situ (DCIS) was excluded, many studies included some patients with DCIS. We included cohorts that had < 10% of participants with DCIS. No restrictions were placed on participants' residence, race or ethnicity, occupation, gender, religion, education, socio‐economic status (SES), or time from diagnosis.

Types of interventions

The experimental interventions could involve the following.

  • Physical activity that included exercise as well as other activities involving body movements done as part of playing, working, active transportation, household chores and recreational activities (WHO 1997). This could include physical activity programmes alone, or in combination with various other treatments. There were no limitations on the setting, duration and delivery of the physical activity programme, but the aim of the intervention had to be weight loss.

  • Dietary interventions that adhered to a foundation diet, with the goal of losing 3% to 5% of body weight (Jensen 2014). Categories included a low‐calorie diet (< 1200 kcal/d) and very‐low‐calorie diet (< 800 kcal/d), based on participants' needs and risk factors. There were no restrictions on dietary type.

  • Drugs aimed primarily at reducing obesity such as: orlistat; sibutramine; L‐carnitine, or other drugs such as metformin used for secondary event prevention.

  • Social, psychological, and behavioural interventions aimed at improving the social environment, and cognitive and behavioural factors in relation to weight loss.

  • Bariatric surgery, which may include a variety of surgical procedures including gastric banding, removal of a portion of the stomach, or gastric bypass surgery.

  • Multifactorial interventions with a combination of the following regimens: physical activity ± diet intervention ± obesity drug ± bariatric surgery ± social, psychological, and behavioural interventions.

The control interventions of comparisons included: placebo, no treatment or waiting list, supportive treatments such as vitamins or minerals, or both, conventional treatments or active control interventions to prevent the recurrence of cancer.

We grouped the comparisons by interventions and controls in the pooling.

We included studies with co‐interventions if they were applied in exactly the same way to both the control and intervention group.

Types of outcome measures

Primary outcomes

  • Overall survival: generally defined in RCTs as time elapsed between randomisation (or study enrolment or intervention initiation) to date of death from any cause.

  • Breast cancer recurrence: generally defined in RCTs as time elapsed between randomisation (or study enrolment or intervention initiation) and event, with event defined as disease recurrence.

Secondary outcomes

  • Change in body weight (from baseline weight).

  • Change in body mass index (BMI).

  • Change in skinfold thickness.

  • Change in waist circumference.

  • Disease‐free survival:generally defined in RCTs as time elapsed between randomisation (or study enrolment or intervention initiation) and event, with event defined as disease recurrence or death from any cause.

  • Adverse events such as exacerbation of symptoms (pain, fatigue, nausea, dyspnoea), falls and bone fractures.

  • Change in quality of life (QOL) and patient‐reported outcomes, as defined by the perceived quality of an individual's daily life. This included emotional, social and physical, domains. We included data from self‐administered structured questionnaires including participants' self‐reported socio‐demographic (race/ethnicity) and medical characteristics (menopausal status) and QOL.

  • Changes in concentration of oestradiol, androgen, insulin, insulin‐like growth factor (IGF), fasting glucose and lipids profile.

  • Changes in adipokine concentrations: plasma leptin, adiponectin.

  • Changes in inflammatory marker concentrations: IL‐6, TNF‐α, CRP.

'Change in BMI' was not listed as an outcome in the protocol, but was added after it became apparent that many studies were reporting this outcome. 'Crude death rates' were incorrectly listed as a secondary outcome in the protocol. However, this has been removed from the list above as 'crude death rates' are group summary statistics corresponding to the outcome of death (or time to death) which is already listed as a primary outcome.

Main outcomes for 'Summary of findings' tables

According to the protocol the following outcomes were to be included in the 'Summary of findings' table(s).

  • Overall survival.

  • Breast cancer recurrence.

  • Adverse events.

  • Change in body weight.

  • Disease‐free survival.

  • Mortality.

  • Change in skinfold thickness.

  • Change in waist circumference.

However due to no trials reporting time‐to‐event data for overall survival and no trials reporting disease‐free survival or skinfold thickness, the following outcomes were included comparing all weight loss interventions (diet, exercise and/or psychosocial) versus comparators (usual care, written materials, wait‐list, placebo or active control).

  • Overall survival

  • Breast cancer recurrence

  • Change in body weight

  • Change in BMI

  • Change in waist circumference

  • Disease‐free survival

  • Adverse events

  • Change in quality of life (overall scales).

Search methods for identification of studies

Electronic searches

We searched the following databases to obtain relevant studies. No language restrictions were applied to the search.

  1. The Cochrane Breast Cancer Group's (CBCG's) Specialised Register on 25 June 2019. Details of the search strategies used by the Group for the identification of studies and the procedure used to code references are outlined in the Group's module (http://www.mrw.interscience.wiley.com/cochrane/clabout/articles/BREASTCA/frame.html). Trials with the key words 'breast cancer survivor', 'survivorship', 'breast cancer', 'obesity', 'overweight', 'body weight management', 'weight loss', 'weight reduction', 'body mass index', 'lifestyle intervention', 'lifestyle activity', 'exercise', 'diet', 'bariatric surgery', 'obesity medication', 'behavior', 'drug therapy' and 'cognitive therapy' were extracted and considered for inclusion in the review.

  2. Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, 2019, Issue 6. See Appendix 1.

  3. MEDLINE (2012 to 17 June 2019) (via OvidSP). See Appendix 2.

  4. Embase (2015 to 17 June 2019) (via OvidSP). See Appendix 3.

  5. The WHO International Clinical Trials Registry Platform (ICTRP) search portal (http://apps.who.int/trialsearch/Default.aspx) for all prospectively registered and ongoing trials on 17 June 2019. See Appendix 4.

  6. Clinicaltrials.gov (http://clinicaltrials.gov/) on 17 June 2019. See Appendix 5.

  7. Mainland Chinese academic literature databases using keywords in Chinese: CNKI (via http://www.cnki.net/) (1979 to 25 June 2019); VIP (http://edu.cqvip.com/) (1989 to 25 June 2019); Wan Fang Data (http://www.wanfangdata.com.cn/) (1980 to 25 June 2019); SinoMed (http://www.sinomed.ac.cn/zh/) (1978 to 25 June 2019).

Searching other resources

Bibliographic searching

We attempted to identify further studies from reference lists of identified relevant trials or reviews. A copy of the full article for each reference reporting a potentially‐eligible trial was obtained. Where this was not possible, we attempted to contact authors to obtain additional information.

We handsearched the retrieved articles and bibliographies in order to identify other potentially‐eligible studies and unpublished data.

Data collection and analysis

Selection of studies

Two review authors (HS and PB) independently screened the titles and abstracts. HS and PB obtained full‐text copies of the relevant articles and included the eligible studies in accordance with the inclusion criteria. During this process any disagreements were resolved by consensus and by the involvement of other review authors (SYT and JV). We recorded excluded trials in the 'Characteristics of excluded studies' table.

Data extraction and management

Data were collected in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) based on the inclusion criteria.

Data were extracted from eligible studies using a data extraction form designed and pilot‐tested by the review authors. We collected the following information: study design, participants, setting, interventions, outcomes, follow‐up, and any other items relevant to this review. Where studies had multiple publications, the main trial report was used as the primary reference and additional details supplemented from all papers. Two review authors (HS and PB) independently extracted the data, and any uncertainties were discussed with other authors (SYT and JV). We contacted study authors if required.

Assessment of risk of bias in included studies

Two review authors (HS and PB) assessed the risk of bias in the included studies using Cochrane's 'Risk of bias' assessment tool. Relevant items included: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessors; completeness of outcome data; selective outcome reporting; and other potential sources of bias. Two review authors (HS and PB) assessed these 'Risk of bias' domains independently, with any disagreements resolved by consensus or by discussion with other review authors (SYT and JV). All judgments were described in the 'Risk of bias' tables and classified as 'low risk', 'high risk', and 'unclear risk'. The results were incorporated into the interpretation of review findings by means of sensitivity analyses.

Measures of treatment effect

For time‐to‐event outcomes, the hazard ratio (HR) is the most appropriate measure of effect. However, no time‐to‐event data were available for extraction.

Dichotomous outcomes were analysed as proportions using the risk ratio (RR) as the measure of effect. Pooled RRs and 95% confidence intervals (CIs) were obtained through Mantel‐Haenszel random‐effects analysis.

Continuous data (including change in body weight, waist circumference, BMI, levels of hormones, cytokines and adipokines, and metabolic effects) were analysed as means with the measure of effect being the mean difference (MD) between treatment groups in change from baseline values, when all studies reported outcomes on the same scale. If similar outcomes were reported on different scales (for example, change in QOL) the standardised mean difference (SMD) was used as the measure of effect. Pooled MDs and SMDs and 95% CIs were obtained through inverse variance random‐effects analysis.

Unit of analysis issues

Intervention‐comparisons were the unit of analysis in this review and corresponded to pairwise comparisons of intervention and control groups. Individual studies assessing more than one intervention group or more than one control group (or both) contributed more than one intervention‐comparison to the review. Consequently, there were more intervention‐comparisons in this review than there were studies.

One study contained three intervention groups (Djuric 2002a;Djuric 2002b;Djuric 2002c) for comparison against a single control group, and another study contained two intervention groups (Demark‐Wahnefried 2014a;Demark‐Wahnefried 2014b) for comparison against a single control group. This was taken into account when treatment effect statistics were calculated by reducing the number of participants in the control groups proportional to the number of intervention‐comparisons. These methods for correcting for multiple intervention and/or control groups were suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Cross‐over trial designs were eligible for inclusion, but only data collected from the first stage were used in meta‐analysis.

Dealing with missing data

If the results of an RCT were published but information on an outcome of interest had not been reported, an attempt was made to contact the trial authors for the missing information (a number of studies provided additional results sufficient for extraction).

With regards to missing outcome data from individual participants, we chose not to impute best/worst case values because this method may not be informative for the most plausible scenarios, particularly for continuous outcomes with no limits to their potential values (Deeks 2019). Moreover, missing outcome data from individual participants probably had a limited impact on effect estimates in this review because attrition was similar in the intervention and control groups (Table 1).

Open in table viewer
Table 1. Number of studies, intervention‐comparisons, patients randomised and patients analysed by outcome

Included in meta‐analysis

Outcome

Studies reporting outcome
N

Studies
N

Intervention‐comparisons
N

Total patients analysed/ randomised (%)

Intervention patients analysed/ randomised (%)

Control patients analysed/ randomised (%)

Overall

20

20

23

2013/2360 (85.3%)

1056/1209 (87.3%)

957/1151 (83.1%)

Overall survival

1

0

0

0

0

0

Breast Cancer recurrence

4

4

4

281/281 (100.0%)

155/155 (100.0%)

126/126 (100.0%)

Change in body weight

19

18

21

1751/2190 (80.0%)

920/1122 (82.0%)

831/1068 (77.8%)

Change in body mass index

15

14

17

1353/1682 (80.4%)

714/862 (82.8%)

639/820 (77.9%)

Change in waist circumference

12

12

13

1193/1541 (77.4%)

634/800 (79.3%)

559/741 (75.4%)

Disease‐free survival

0

0

0

0

0

0

Adverse events

5

3

4

394/446 (88.3%)

205/229 (89.5%)

189/217 (87.1%)

Change in quality of life ‐ overall

8

8

10

867/1148 (75.5%)

447/578 (77.3%)

420/570 (73.7%)

Change in quality of life ‐ physical subscales

7

7

10

1024/1351 (75.8%)

530/679 (78.1%)

494/672 (73.5%)

Change in quality of life ‐ social subscales

4

4

6

389/464 (83.8%)

196/228 (86.0%)

193/236 (81.8%)

Change in quality of life ‐ emotional subscales

5

5

8

498/633 (78.7%)

264/321 (82.2%)

234/312 (75.0%)

Change in quality of life ‐ mental health subscales

3

3

3

355/400 (88.8%)

173/200 (86.5%)

182/200 (91.0%)

Change in quality of life ‐ anxiety depression subscales

3

3

3

669/910 (73.5%)

340/457 (74.4%)

329/453 (72.6%)

Change in insulin

4

4

6

134/192 (69.8%)

79/95 (83.2%)

55/97 (56.7%)

Change in glucose

4

4

6

133/192 (69.3%)

78/95 (82.1%)

55/97 (56.7%)

Change in total cholesterol

5

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in HDL cholesterol

5

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in LDL cholesterol

4

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in triglycerides

4

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in leptin

3

1

3

39/74 (52.7%)

27/35 (77.1%)

12/39 (30.8%)

HDL: high‐density lipoprotein
LDL: low‐density lipoprotein

Assessment of heterogeneity

Heterogeneity between trial results was assessed using the Chi2 test statistic and the I2 statistic. The Chi2 test statistic assesses the amount of variation in a set of trials. Small P values for the Chi2 test statistic suggest that there is more heterogeneity present than would be expected by chance. Chi2 is not a particularly sensitive test: a cut‐off of P value less than 0.10 is often used to indicate significance, but lack of statistical significance does not mean there is no heterogeneity. I2 is the proportion of variation that is due to heterogeneity rather than chance. In conjunction with the Chi2 test, we used the I2 statistic to assess heterogeneity using the rule of thumb guide outlined in the Cochrane Handbook (Higgins 2011) (i.e. an I2 between 0% to 40% might not be important; between 30% to 60% may represent moderate heterogeneity; between 50% to 90% may represent substantial heterogeneity; and between 75% to 100% considerable heterogeneity).

Assessment of reporting biases

In addition to assessing each intervention‐comparison individually for 'selective reporting' using Cochrane's 'Risk of bias' assessment tool (see Assessment of risk of bias in included studies above), publication bias and/or small‐study effects were assessed for outcomes with more than 10 or more intervention‐comparisons. Egger's statistical test was used to formally assess the degree of asymmetry (Egger 1997).

Data synthesis

For binary outcomes, RevMan 5.3 was used to estimate pooled RRs and 95%CIs using the random‐effects Mantel‐Haenszel method. For continuous data, RevMan 5.3 was used to estimate pooled MDs or SMDs and 95%CIs using inverse variance random‐effects analysis. No time to event data were available for pooling.

Subgroup analysis and investigation of heterogeneity

For outcomes with more than 10 or more intervention‐comparisons, four subgroup analyses were performed to determine whether the results differed by the following.

  • Intervention type versus control type:

    • diet versus usual care, wait‐list, written materials or placebo;

    • diet + exercise versus usual care, wait‐list, written materials or placebo;

    • diet + exercise + psychosocial versus usual care, wait‐list, written materials or placebo;

    • diet + psychosocial versus active control (diet + psychosocial);

    • diet + exercise versus active control (diet + exercise).

  • Ethnicity:

    • > 95% African descent;

    • > 95% White/Caucasian;

    • mixed (≤ 95% of any one ethnic group); or

    • unspecified.

  • Menopausal status:

    • 100% postmenopausal women;

    • mix of pre‐ and post‐menopausal women; or

    • unspecified.

  • Duration of follow‐up (months):

    • ≤ 6 months;

    • > 6 months;

    • ≤ 12 months;

    • > 12 months.

Possible subgroup differences were assessed using Chi2 tests. Of the above four subgroup analyses, subgroup analyses of intervention type, ethnicity and menopausal status were 'a priori' subgroup analyses pre‐specified in the review protocol while the subgroup analysis on duration of follow‐up was 'post hoc'.

A number of other subgroup analyses pre‐specified in the review protocol were not performed because data relating to the subgroup characteristic were insufficient or participants characteristics relating to the subgroups varied substantially within studies (making subgroup analysis vulnerable to ecological bias).

Sensitivity analysis

To assess the sensitivity of our primary results to our choice of analytical method, we repeated Analysis 1.2, Analysis 1.3 and Analysis 1.4 (the outcomes with the most intervention‐comparisons) but using fixed‐effect rather than random‐effects methods.

Summary of findings and assessment of the certainty of the evidence

We used the standard GRADE system to rate the quality of evidence relating to the estimated treatment effects on breast cancer recurrence, change in body weight, change in BMI, change in waist circumference, adverse events and change in quality of life (overall scales). GRADE criteria for assessing quality of evidence included 'study design', 'risk of bias', 'inconsistency', 'indirectness', 'imprecision', 'suspected publication bias' and 'other considerations' (Schünemann 2011). Assessments of these criteria and corresponding justifications are provided in two 'Summary of findings' tables largely created using GRADEproGDT. GRADE assessments were performed separately for selected subgroups related to effect estimate heterogeneity (or 'inconsistency' as labelled in the GRADE assessment criteria).

We used GRADEproGDT software (GRADEproGDT) to produce 'Summary of findings' tables to illustrate the main outcomes and grade the quality of the evidence.

Four grades of evidence were used, as recommended by the GRADE Working Group, as follows.

  • High quality: further research is very unlikely to change our confidence in the estimate of effect.

  • Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

  • Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

  • Very low quality: we are very uncertain about the estimate.

Results

Description of studies

Results of the search

SeeFigure 1.

Figure 1
Study flow diagram.

Study flow diagram.

We identified 9283 records from a comprehensive literature search conducted on 17 June 2019 (English database search) and 25 June 2019 (Chinese database search) (Figure 1). A further four records were identified through other sources including correspondence with authors and bibliographic searching. We identified and removed 547 duplicates, leaving 8740 records. Of these records, 8027 were excluded by the title and abstract screen, and 713 records were subsequently full‐text screened. Of these records, 632 were excluded by the full‐text screen, with reasons provided (refer to Characteristics of excluded studies). Of the remaining 81 records, 59 records (corresponding to 20 studies containing 23 intervention‐comparisons) were included in the qualitative and quantitative synthesis (refer to Characteristics of included studies), 17 records (corresponding to 12 studies) were awaiting classification (refer to Characteristics of studies awaiting classification) and five records (corresponding to five studies) were ongoing studies (refer to Characteristics of ongoing studies). No studies from the Chinese searches met the inclusion criteria for our review.

Included studies

In total, 20 studies comprising 23 intervention‐comparisons were included in this review (refer to Characteristics of included studies table). The principal publications of these 20 studies are: Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Dittus 2018; Djuric 2002a; Ferrante 2017; Ghavami 2017; Goodwin 2014; Greenlee 2013; Kwiatkowski 2017; Mefferd 2007; Reeves 2016; Rock 2015; Scott 2013; Shaw 2007; Sheppard 2016; Stendell‐Hollis 2010; Stolley 2017; Swisher 2015; Thomson 2010. There were two intervention‐comparisons for the DAMES (Daughters And MothErS Against Breast Cancer) trial (Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b), and three intervention‐comparisons for the ABC (Weight Loss After Breast Cancer Diet Study) trial (Djuric 2002a; Djuric 2002b; Djuric 2002c). Where applicable, we contacted authors for further data for this review, and received additional data regarding the following intervention‐comparisons: Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Kwiatkowski 2017.

Of the 23 intervention‐comparisons included in this review (Table 1):

  • no time‐to‐event data for pooled hazard ratio (HR) estimation were extractable for 'overall survival', 'disease‐free survival' or 'breast cancer recurrence'. One intervention‐comparison (Rock 2015) reported the number of deaths in each arm during the 24‐month follow‐up period. Four intervention‐comparisons (Greenlee 2013; Mefferd 2007; Sheppard 2016; Kwiatkowski 2017) reported the number of breast cancer recurrences in each arm during their respective maximum follow‐up periods (6, 4, 3 and 36 months, respectively). Hence, we performed a meta‐analysis of the 'breast cancer recurrence' outcome, but with risk ratios (RRs) rather than HRs as the measure of effect.

  • 21 (91%) were included in meta‐analysis for 'change in body weight', 17 (74%) for 'change in BMI' and 13 (57%) for 'change in waist circumference'.

  • 10 (43%) were included in meta‐analysis for 'change in overall QOL', 10 (43%) for 'change in QOL physical subscales', 6 (26%) for 'change in QOL social subscales', 8 (35%) for 'change in QOL emotional subscales' and 3 (13%) for both 'change in QOL mental health subscales' and 'change in QOL anxiety and depression subscales'

  • 6 (26%) were included in each meta‐analysis for 'change in insulin', 'change in glucose', 'change in total cholesterol', 'change in high‐density lipoprotein (HDL) cholesterol', 'change in low‐density lipoprotein (LDL) cholesterol' and 'change in triglycerides', and 3 (13%) for 'change in leptin'.

  • 4 (17%) were included in meta‐analysis for adverse events.

The 23 intervention‐comparisons included in this review can be subdivided according to the types of interventions and controls as follows:

Excluded studies

We excluded 632 reports in the full‐text screen and reasons for exclusion were provided for a subset of reports (i.e. 158 reports; refer to the Characteristics of excluded studies table). Common reasons for exclusion included trials where the aim of the intervention was not weight loss, trials with participants who were not overweight or obese (BMI <25 kg/m2), trials not reporting outcomes relevant to this review (e.g. anthropometric measures), and trials with the wrong study design (e.g. non‐randomised trials).

Risk of bias in included studies

Allocation

Random sequence generation

All 23 intervention‐comparisons were randomised, as this was an inclusion criteria for this review. Of these, 11 intervention‐comparisons provided sufficient details on the method of randomised sequence generation to be judged as having a low risk of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Ferrante 2017; Ghavami 2017; Goodwin 2014; Kwiatkowski 2017; Reeves 2016; Rock 2015; Shaw 2007; Stendell‐Hollis 2010; Stolley 2017) (Figure 2). The remaining 12 intervention‐comparisons did not provide sufficient details on sequence generation and were therefore judged to have an unclear risk of bias (Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Greenlee 2013; Mefferd 2007; Scott 2013; Sheppard 2016; Swisher 2015; Thomson 2010).

Figure 2
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Allocation concealment

Sixteen out of the 23 intervention‐comparisons described their method of allocation concealment and were judged to have a low risk of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Ferrante 2017; Ghavami 2017; Goodwin 2014; Greenlee 2013; Kwiatkowski 2017; Reeves 2016; Scott 2013; Shaw 2007; Stendell‐Hollis 2010; Stolley 2017; Swisher 2015). The remaining seven intervention‐comparisons did not describe their method of allocation concealment and were therefore judged as having an unclear risk of bias (Djuric 2002a; Djuric 2002b; Djuric 2002c; Mefferd 2007; Rock 2015; Sheppard 2016; Thomson 2010).

Blinding

Blinding of participants and personnel

Only one intervention‐comparison was judged to be at low risk of bias for blinding of participants and personnel (Stendell‐Hollis 2010). This was a trial comparing two different types of tea, and was described as being 'double‐blind'. The remaining 22 intervention‐comparisons were judged to be at a high risk of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Ghavami 2017; Goodwin 2014; Greenlee 2013; Kwiatkowski 2017; Mefferd 2007; Reeves 2016; Rock 2015; Scott 2013; Shaw 2007; Sheppard 2016; Stolley 2017; Swisher 2015; Thomson 2010). Blinding of participants would have been difficult due to the nature of most interventions (e.g. one group receiving an active exercise or diet intervention or a psychosocial intervention, whilst the other group received usual care). Blinding of personnel was often not mentioned, although was mentioned to have occurred for some studies, but we still judged these intervention‐comparisons to be at high risk of bias as participants were not blinded (with the exception of Stendell‐Hollis 2010).

Blinding of outcome assessors

Regarding patient‐reported outcomes, 20 intervention‐comparisons were judged to have a high risk of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ghavami 2017; Goodwin 2014; Greenlee 2013; Kwiatkowski 2017; Mefferd 2007; Reeves 2016; Rock 2015; Scott 2013; Shaw 2007; Sheppard 2016; Swisher 2015; Thomson 2010). As outlined above, this related to difficulty blinding participants for many diet, exercise and psychological interventions. Two intervention‐comparisons were judged to have a low risk of bias (Ferrante 2017; Stendell‐Hollis 2010); with the former utilising an active control group and the latter reporting being 'double‐blind'. One treatment comparison was judged to have an unclear risk of bias (Stolley 2017) as no patient‐reported outcomes from this trial were applicable to this review.

Regarding non patient‐reported outcomes, 16 intervention‐comparisons were judged to have an unclear risk of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Ghavami 2017; Goodwin 2014; Greenlee 2013; Kwiatkowski 2017; Mefferd 2007; Rock 2015; Shaw 2007; Sheppard 2016; Thomson 2010). In most cases, this was because the study did not mention blinding of assessors. Four intervention‐comparisons were determined to have a low risk of bias because they reported that assessors were blinded (Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Reeves 2016; Stendell‐Hollis 2010). The remaining three intervention‐comparisons were classified as having a high risk of bias (Scott 2013; Stolley 2017; Swisher 2015).

Incomplete outcome data

Regarding patient‐reported outcomes, nine intervention‐comparisons had a low risk of attrition bias (Arikawa 2017; Ferrante 2017; Ghavami 2017; Greenlee 2013; Reeves 2016; Scott 2013; Sheppard 2016; Stendell‐Hollis 2010; Thomson 2010). These comparisons tended to have a relatively low dropout rate (often with reasons given for dropouts), conducted analyses of completers versus non‐completers and utilised imputation for missing data. Seven intervention‐comparisons were judged to have a high risk of attrition bias (Demark‐Wahnefried 2012a; Dittus 2018; Kwiatkowski 2017; Mefferd 2007; Rock 2015; Shaw 2007; Swisher 2015). These trials tended to have higher dropout rates (which may have been significantly higher in one group) and did not conduct analyses to explore the cause and/or effects of this dropout (e.g. comparing completers and non‐completers). The remaining seven intervention‐comparisons were classified as having an unclear risk of bias (Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Djuric 2002a; Djuric 2002b; Djuric 2002c; Goodwin 2014; Stolley 2017). For Demark‐Wahnefried 2014a and Demark‐Wahnefried 2014b, we contacted authors (and received) data regarding participants with stage I‐III breast cancer, however this did not include the number lost to follow‐up. In Djuric 2002a; Djuric 2002b and Djuric 2002c, nine out of 48 (19%) participants were lost to follow‐up, with eight of these participants allocated to the three intervention groups (four, three and one for the different interventions) and one participant in the control group. The significance of this loss to follow‐up was difficult to determine and therefore classified as 'unclear'. In Goodwin 2014, we could only extract data for participants with BMI ≥ 30 kg/m2 and therefore could not determine the loss to follow‐up in this group. Stolley 2017 did not contain any data regarding patient‐reported outcomes for inclusion in our review.

Regarding non patient‐reported outcomes, 11 intervention‐comparisons were determined to have a low risk of bias due to the relatively low dropout rate and analysis of the impact of these non‐completers (Arikawa 2017; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Ghavami 2017; Greenlee 2013; Reeves 2016; Scott 2013; Sheppard 2016; Stendell‐Hollis 2010). Eight intervention‐comparisons were allocated a high risk of bias, again mostly due to having higher dropout rates without any analysis regarding this loss to follow‐up (Demark‐Wahnefried 2012a; Dittus 2018; Kwiatkowski 2017; Mefferd 2007; Shaw 2007; Stolley 2017; Swisher 2015; Thomson 2010). The remaining four comparisons had an unclear risk of bias (Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Goodwin 2014; Rock 2015). For Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b and Goodwin 2014, we were unable to determine the number lost to follow‐up. In Rock 2015, an analysis found that participants who dropped out were significantly heavier than completers at 12‐month follow‐up, however they were not statistically significantly heavier at baseline and six‐month follow‐up, and the overall dropout rate was relatively low.

Selective reporting

Eighteen intervention‐comparisons were determined to have a low risk of reporting bias, as they conducted pre‐planned analyses as per trial registries or protocols (Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Ferrante 2017; Ghavami 2017; Goodwin 2014; Kwiatkowski 2017; Mefferd 2007; Reeves 2016; Rock 2015; Scott 2013; Shaw 2007; Sheppard 2016; Stendell‐Hollis 2010; Stolley 2017; Swisher 2015; Thomson 2010). The remaining five intervention‐comparisons were determined to have a high risk of reporting bias (Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Greenlee 2013). For Dittus 2018 QOL data were not reported in the manuscript (despite it being mentioned in the registry), but the authors did provide these data on request. For Djuric 2002a; Djuric 2002b; Djuric 2002c, the trial mentioned collecting questionnaire data but did not present these data. For Greenlee 2013 the study mentioned that the main outcomes were reported, however it was unclear what (or why) other outcomes were omitted.

Other potential sources of bias

Twenty‐two intervention‐comparisons were determined to have a low risk of other potential sources of bias (Arikawa 2017; Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Ghavami 2017; Greenlee 2013; Kwiatkowski 2017; Mefferd 2007; Reeves 2016; Rock 2015; Scott 2013; Shaw 2007; Sheppard 2016; Stendell‐Hollis 2010; Stolley 2017; Swisher 2015; Thomson 2010). One intervention‐comparison was determined to have a high risk of other potential sources of bias (Goodwin 2014). For Goodwin 2014 the trial included participants with BMI ≥ 24 kg/m2, and participants with BMI ≥ 25 kg/m2 would be eligible for our review. However, we could only extract data for participants with BMI ≥ 30 kg/m2.

Effects of interventions

See: Summary of findings 1 All weight loss interventions for overweight and obese breast cancer survivors; Summary of findings 2 Weight loss interventions involving diet, exercise and psychosocial support for overweight and obese breast cancer survivors

Overall survival

No time‐to‐event data for pooled hazard ratio (HR) estimation were extractable for 'overall survival'. One intervention‐comparison (Rock 2015) reported the number of deaths in each arm during the 24‐month follow‐up period (0 deaths out of 348 intervention participants and five deaths out of 349 control participants). We are uncertain about the effect of these interventions on overall survival.

Breast cancer recurrence

No time‐to‐event data for pooled HR estimation were extractable for 'breast cancer recurrence'. However, four intervention‐comparisons (Greenlee 2013; Mefferd 2007; Sheppard 2016; Kwiatkowski 2017) reported the number of breast cancer recurrences in each arm during their maximum follow‐up periods (six, four, three and 36 months, respectively), and these data were pooled using meta‐analyses (but with risk ratios (RRs) rather than HRs as the measure of effect). Of these four intervention‐comparisons, all 281 randomised participants were analysed. Fourteen participants had recurrence of their breast cancer, with 10 of these participants being in the intervention group and the remaining four in the control group. We are uncertain about the effect of these interventions on breast cancer recurrence. There was no significant difference in the incidence of cancer recurrence between the intervention and control groups, with a RR of 1.95 (95% confidence interval (CI): 0.68 to 5.60, P = 0.21;low‐quality evidence). There was no heterogeneity identified across trials (P = 0.60; I2 = 0%) (Analysis 1.1).

Change in body weight

Twenty‐two of 23 intervention‐comparisons assessed change in body weight as an outcome, with 21 comparisons providing adequate data for pooling in meta‐analyses. From these 21 intervention‐comparisons, 1751 out of 2190 (80.0%) randomised patients were assessed at follow‐up. The weight loss intervention appears to reduce body weight compared to the control group (mean difference (MD): ‐2.25 kg, 95% CI: ‐3.19 to ‐1.30, P < 0.00001; low‐quality evidence), although there was considerable heterogeneity identified (P < 0.00001; I2 = 69%) (Analysis 1.2;Figure 3; Figure 4).

Figure 3
Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.2 Change in body weight [kg].

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.2 Change in body weight [kg].

Figure 4
Funnel plot 1: Change in body weight [kg]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body weight. The plot does not show substantial asymmetry (Egger's test P value 0.40).

Funnel plot 1: Change in body weight [kg]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body weight. The plot does not show substantial asymmetry (Egger's test P value 0.40).

Subgroup analyses

  • There were no significant differences in change in body weight between the different subgroups of types of intervention (diet, diet + exercise, diet + exercise + psychosocial support) and control groups (active or passive control groups) (P = 0.21; I2 = 32%). However, there was some suggestion that bimodal interventions such as combined diet and exercise interventions might perform better than diet interventions alone, and that multimodal interventions such as combined diet and exercise and psychosocial interventions might perform better than just diet and exercise (diet only: MD: 0.09 kg, 95% CI: ‐3.88 to 4.05, P = 0.006, heterogeneity P = 0.008, I2 = 79%, 3 studies, 72 participants; diet + exercise: MD: ‐2.03 kg, 95% CI: ‐3.48 to ‐0.58, P = 0.006; heterogeneity P = 0.58, I2 = 0%; 3 studies, 93 participants; diet + exercise + psychosocial: MD: ‐2.88 kg, 95% CI: ‐3.98 to ‐1.77, P < 0.00001; heterogeneity P = 0.0001, I2 = 69%; 13 studies, 1526 participants). (Analysis 2.1).

  • There were no significant differences in change in body weight between ethnicity subgroups (P = 0.11; I2 = 54.3%) (Analysis 3.1), between subgroups based on menopausal status (P = 0.67; I2 = 0%) (Analysis 4.1) and between subgroups based on duration of follow‐up (P = 0.76; I2 = 0%) (Analysis 5.1).

Change in body mass index (BMI)

Eighteen of 23 intervention‐comparisons included 'change in BMI' as an outcome, with 17 comparisons providing adequate data for pooling in meta‐analyses. From these 17 intervention‐comparisons, 1353 of 1682 (80.4%) randomised patients were assessed at follow‐up. The weight loss intervention appears to reduce BMI compared to the control group (MD:‐1.08 kg/m2, 95% CI: ‐1.61 to ‐0.56, P < 0.0001; low‐quality evidence) although large heterogeneity is noted (P < 0.00001; I2 = 84%) (Analysis 1.3; Figure 5; Figure 6).

Figure 5
Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.3 Change in body mass index [kg/m2].

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.3 Change in body mass index [kg/m2].

Figure 6
Funnel plot 2: Change in body mass index [kg/m2]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body mass index. The plot does not show substantial asymmetry (Egger's test P value 0.13).

Funnel plot 2: Change in body mass index [kg/m2]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body mass index. The plot does not show substantial asymmetry (Egger's test P value 0.13).

Subgroup analyses

  • There were no significant differences in change in BMI between subgroups of different types of intervention and control groups (P = 0.13; I2 = 46.9%). (Analysis 2.2).

  • There were no significant differences in change in BMI between ethnicity subgroups (P = 0.25; I2 = 26.9%) (Analysis 3.2), between subgroups based on menopausal status (P = 0.21; I2 = 36.4%) (Analysis 4.2) and between subgroups based on duration of follow‐up (P = 0.48; I2 = 0%) (Analysis 5.2).

Change in waist circumference

Fourteen of 23 intervention‐comparisons assessed 'change in waist circumference' as an outcome, with 13 comparisons providing sufficient data for pooling in meta‐analyses. From these 13 intervention‐comparisons, 1193 of the 1541 (77.4%) randomised patients were assessed at follow‐up. The weight loss intervention appears to decrease weight circumstance compared to the control group(MD: ‐1.73 cm, 95% CI: ‐3.17 to ‐0.29, P = 0.02; low‐quality evidence) although considerable heterogeneity was identified (P < 0.0001; I2 = 73%) (Analysis 1.4; Figure 7).

Figure 7
Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.4 Change in waist circumference [cm].

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.4 Change in waist circumference [cm].

subgroup analyses

  • There was a marginally significant difference in change in waist circumference between the different subgroups of types of intervention and control groups (P = 0.04). This difference was most pronounced between the 'diet + exercise' subgroup compared to the 'diet' subgroup (diet + exercise: MD: ‐3.63 cm, 95% CI: ‐5.76 to ‐1.51, within group P = 0.0008; heterogeneity P = 0.31, I2 = 15%; 3 studies, 93, participants; diet: MD: 2.30 cm, 95% CI: ‐1.93 to 6.53, within group P = 0.29; heterogeneity not applicable; 1 study, 39 participants) (Analysis 2.3).

  • There were no significant differences in change in waist circumference between ethnicity subgroups (P = 0.22; I2 = 33.9%) (Analysis 3.3), between subgroups based on menopausal status (P = 0.51; I2 = 0%) (Analysis 4.3) and between subgroups based on duration of follow‐up (P = 0.66; I2 = 0%) (Analysis 5.3).

Disease‐free survival

None of the intervention‐comparisons reported information on disease‐free survival.

Adverse events

Six of 23 intervention‐comparisons recorded adverse events as an outcome, with four comparisons providing sufficient data for pooling in meta‐analyses. From these four intervention‐comparisons, 394 of the 446 (88.3%) randomised patients were assessed at follow‐up. There were no differences between the weight loss program and control groups with regards to adverse events (RR 0.94 events, 95% CI: 0.76 to 1.17, P = 0.59; high‐quality evidence) and there was no evidence of heterogeneity (P= 0.71; I2 = 0%) (Analysis 1.5).

Change in quality of life (QOL)

Change in overall QOL

Fifteen of 23 intervention‐comparisons assessed change in QOL outcomes. Of these, 10 comparisons reported on 'change in overall QOL', and all 10 comparisons had sufficient data for pooling in meta‐analyses. From these 10 intervention‐comparisons, we analysed 867 of the 1148 (75.5%) randomised patients. The weight loss intervention appears to improve quality of life compared to the control group. There was (standardised mean difference (SMD): 0.74, 95% CI: 0.20 to 1.29, P = 0.008; low‐quality evidence) although large heterogeneity was identified (P < 0.00001; I2 = 89%) (Analysis 1.6).

Change in physical QOL subscales

Ten of 23 intervention‐comparisons assessed 'change in physical QOL' subscales as an outcome and provided sufficient data for pooling in meta‐analyses. Of these 10 intervention‐comparisons, 1024 of the 1351 (75.5%) randomised patients were assessed at follow‐up. There was a significant difference favouring the intervention group (SMD: 0.33, 95% CI: 0.10 to 0.56, P = 0.005) although there was evidence of moderate heterogeneity (P = 0.06; I2 = 45%) (Analysis 1.7).

Change in social QOL subscales

Six of 23 intervention‐comparisons assessed 'change in social QOL' subscales as an outcome and provided sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 389 of the 464 (83.8%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in social subscale QOL (SMD: 0.19, 95% CI: ‐0.01 to 0.39, P = 0.07) and there was no evidence of statistically significant heterogeneity (P = 0.49; I2 = 0%) (Analysis 1.8).

Change in emotional QOL subscales

Eight of 23 intervention‐comparisons assessed 'change in emotional QOL' subscales as an outcome and provided sufficient data for pooling in meta‐analyses. Of these eight intervention‐comparisons, 498 out of the 633 (78.7%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in emotional subscale QOL (SMD: 0.11, 95% CI: ‐0.09 to 0.30, P = 0.28) and heterogeneity was minimal (P = 0.40; I2 = 4%) (Analysis 1.9.)

Change in mental health QOL subscales

Three of 23 intervention‐comparisons assessed 'change in mental health QOL' subscales as an outcome and provided sufficient data for pooling in meta‐analyses. Of these three intervention‐comparisons, 355 of the 400 (88.8%) randomised patients were assessed at follow‐up. There was a significant difference favouring the intervention group (SMD: 0.60, 95% CI: 0.17 to 1.02, P = 0.006), although moderate heterogeneity was identified (P = 0.09; I2 = 59%) (Analysis 1.10).

Change in anxiety and depression QOL subscales

Three of 23 intervention‐comparisons assessed 'change in anxiety and depression QOL' subscales as an outcome and provided sufficient data for pooling in meta‐analyses. Of these 3 intervention‐comparisons, 669 of the 910 (73.5%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups in change in anxiety and depression subscale QOL (SMD: 0.63, 95% CI: ‐0.07 to 1.33, P = 0.08) but there was evidence of significant heterogeneity (P = 0.0001; I2 = 91%) (Analysis 1.11).

Change in oestradiol

One of 23 intervention‐comparisons reported data on 'change in oestradiol'. Scott 2013reported the median changes in oestradiol levels was ‐0.5 pg/mL for the intervention group and ‐1.0 pg/mL for the control group (P = 0.58 for difference between groups).

Change in testosterone

One of 23 intervention‐comparisons reported data on 'change in testosterone'. Scott 2013reported the median changes in testosterone levels was 0.0 nmol/L for the intervention group and 0.1 nmol/L for the control group (P = 0.44 for difference between groups).

Change in insulin

Seven of 23 intervention‐comparisons measured 'change in insulin' concentrations as an outcome, with six comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 134 of the 192 (69.8%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to insulin concentrations (MD: ‐1.49 mcU/mL, 95% CI: ‐6.62 to 3.64, P = 0.57) and moderate heterogeneity was identified (P = 0.05; I2 = 54%) (Analysis 1.12).

Change in insulin‐like growth factor 1 (IGF‐1)

Two of 23 intervention‐comparisons reported data on IGF‐1, however this was insufficient for pooling in meta‐analyses because one intervention‐comparison reported medians, while the other reported means. Scott 2013reported that the median changes in IGF‐1 levels was ‐1.7 pg/mL for the intervention group and ‐1.3 pg/mL for the control group (P = 0.84 for difference between groups). Arikawa 2017 reported the mean changes in IGF‐1 levels was 13.6 ng/mL for the intervention group and 5.8 ng/mL for the control group (SDs for change scores and P values for group differences were not reported).

Change in fasting glucose

Six of 23 intervention‐comparisons measured 'change in fasting glucose' concentrations as an outcome, with all 6 comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 133 of the 192 (69.3%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in fasting glucose concentrations (MD: ‐0.46 mg/dl, 95% CI: ‐4.86 to 3.93, P = 0.84), and there was no evidence of significant heterogeneity (P = 0.68; I2 = 0%) (Analysis 1.13).

Lipids profile

Change in total cholesterol

Seven of 23 intervention‐comparisons measured 'change in total cholesterol (TC) concentrations' as an outcome, with six comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons,189 of the 256 (73.8%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in TC concentrations (MD: ‐0.34 mmol/L, 95% CI: ‐0.84 to 0.16, P = 0.18) and there was evidence of moderate heterogeneity (P = 0.07; I2 = 50%) (Analysis 1.14).

Change in high‐density lipoprotein (HDL) cholesterol

Seven of 23 intervention‐comparisons measured HDL concentrations as an outcome, with six comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 189 of the 256 (73.8%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in HDL concentrations (MD: 0.00 mmol/L, 95% CI: ‐0.08 to 0.08, P = 0.98) and there was no evidence of statistically significant heterogeneity (P = 0.70; I2=0%) (Analysis 1.15).

Change in low‐density lipoprotein (LDL) cholesterol

Six of 23 intervention‐comparisons measured LDL concentrations as an outcome, with all six comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 189 of the 256 (73.8%) randomised patients were assessed at follow‐up. There were no significant differences between the intervention and control groups with regards to change in LDL concentrations (MD: ‐0.18 mmol/L, 95% CI: ‐0.59 to 0.22, P = 0.38) and there was evidence of some heterogeneity (P ‐ 0.19; I2 = 32%) (Analysis 1.16).

Change in triglycerides (TG)

Six of 23 intervention‐comparisons measured TG concentrations as an outcome, with all six comparisons providing sufficient data for pooling in meta‐analyses. From these six intervention‐comparisons, 189 of the 256 (73.8%) randomised patients were assessed at follow‐up. There was a significant difference favouring the intervention group (MD: ‐0.26 mmol/L, 95% CI: ‐0.45 to ‐0.07, P = 0.008) and there was no evidence of statistically significant heterogeneity (P = 0.92; I2 = 0%) (Analysis 1.17).

Change in leptin

Six of 23 intervention‐comparisons measured leptin concentrations as an outcome, with three comparisons (all from a single study: Djuric 2002a; Djuric 2002b; Djuric 2002c) providing sufficient data for pooling in meta‐analyses. From these three intervention‐comparisons, 39 of the 74 (52.7%) randomised patients were assessed at follow‐up. There was a significant difference favouring the intervention group (MD: ‐14.67 ng/mL, 95% CI: ‐26.36 to ‐2.98, P = 0.01) and there was no evidence of significant heterogeneity (P = 0.53; I2 = 0%) (Analysis 1.18).

Change in adiponectin

Two of 23 intervention‐comparisons reported data on 'change in adiponectin' concentrations but one intervention‐comparison reported medians while the other reported means. Swisher 2015 reported the median changes in adiponectin levels was ‐0.9 ng/mL for the intervention group and ‐1.3 ng/mL for the control group (P = 0.58 for difference between groups). Arikawa 2017 reported the mean changes in adiponectin levels was 1 ng/mL for the intervention group and 0.6 ng/mL for the control group (SDs for change scores and P values for group differences were not reported).

Change in interleukin‐6

Four of 23 intervention‐comparisons reported data on 'change in IL‐6 concentrations', however the data were insufficient for pooling in meta‐analyses.Swisher 2015 reported the median changes in IL‐6 levels was 0.02 pg/mL for the intervention group and 0.3 pg/mL for the control group (P = 0.26 for difference between groups). Scott 2013reported the median changes in IL‐6 levels was 0.09 pg/mL for the intervention group and 0.18 pg/mL for the control group (P value for difference between groups not reported). Arikawa 2017 reported the mean changes in IL‐6 levels was ‐0.3 pg/mL for the intervention group and ‐0.1 pg/mL for the control group (SDs for change scores and P values for group differences were not reported). Mefferd 2007 reported the mean changes in IL‐6 levels was ‐0.3 pg/mL for the intervention group and ‐0.3 pg/mL for the control group (SDs for change scores and P values for group differences were not reported).

Change in tumour necrosis factor alpha (TNF‐α)

Three of 23 intervention‐comparisons reported data on 'change in TNF‐α' concentrations, however the data were insufficient for pooling in meta‐analyses. Swisher 2015 reported the median changes in TNF‐α levels was ‐0.3 pg/mL for the intervention group and ‐0.8 pg/mg for the control group (P = 0.11 for difference between groups). Scott 2013reported the median changes in TNF‐α levels was 0.03 pg/mg for the intervention group and ‐0.07 pg/mg for the control group (P value for difference between groups not reported). Mefferd 2007 reported the mean changes in TNF‐α levels was ‐0.5 pg/mg for the intervention group and ‐0.8 pg/mg for the control group (SDs for change scores and P values for group differences were not reported).

Change in C‐reactive protein (CRP)

Four of 23 intervention‐comparisons reported data on 'change in CRP concentrations', however the data were insufficient for pooling in meta‐analyses. Swisher 2015 reported the median changes in CRP levels was ‐0.2 mg/L for the intervention group and 0.4 mg/L for the control group (P = 0.78 for difference between groups). Scott 2013reported the median changes in CRP levels was 0.1 mg/L for the intervention group and 0.03 mg/L for the control group (P = 0.80 for difference between groups). Arikawa 2017 reported the mean changes in CRP levels was ‐2.2 mg/L for the intervention group and 0.3 mg/L for the control group (SDs for change scores and P values for group differences were not reported). Thomson 2010 reported the mean changes in CRP levels was ‐0.4 mg/L for both the intervention and control groups (SDs for change scores and P values for group differences were not reported).

Sensitivity analyses

Repeating Analysis 1.2, Analysis 1.3 and Analysis 1.4 using a fixed‐effect model did not appreciably change the pooled effect estimates for 'change in body weight', 'change in BMI' or 'change in waist circumference' (Analysis 6.1; Analysis 6.2; Analysis 6.3).

Discussion

Summary of main results

This review analysed 2028 overweight or obese breast cancer survivors from 23 intervention‐comparisons (from 20 studies) (see Table 1). Participants in the experimental groups received weight loss interventions using different strategies and combinations of modalities such as 'diet', 'diet and exercise', 'diet and psychosocial support' or 'diet, exercise and psychosocial support'. Overall, these different intervention approaches appeared to result in a reduction in body weight and waist circumference, were not associated with an increase in adverse events, and improved overall quality of life (refer to summary of findings Table 1). However, the quality of this evidence (GRADE) was low, with the exception of adverse events (which had a high GRADE score). This was due to substantial heterogeneity for the above outcomes, a relatively high attrition rate (> 20%) at follow‐up and inability to blind participants. Similarly, the interventions as a whole were effective in reducing body mass index (BMI).

Regarding the primary outcomes, no data were available for disease‐free survival or overall survival. There was a relatively small amount of data available for cancer recurrence (281 participants from 4 intervention‐comparisons (4 studies), with 14 recurrence events). Recurrence rates were not significantly different between the intervention and control groups, but this analysis was likely underpowered.

We also conducted a subgroup analysis of multimodal interventions incorporating all three elements of diet, exercise and psychosocial support (refer to summary of findings Table 2). These interventions appeared to result in a reduction in body weight and waist circumference. However, the quality of this evidence was similarly graded to be low and there was substantial heterogeneity. Although there was not a significant difference between subgroups of intervention and control types, the subgroup analyses showed some indication of greater benefits for multimodal interventions, as interventions combining 'diet and exercise' or 'diet and exercise and psychosocial support' generally led to greater decreases in anthropometric measures (weight, BMI and waist circumference) compared to 'diet' only interventions.

Relative to controls, the intervention groups did not significantly impact analysed biomarkers (insulin, fasting glucose, total cholesterol, high‐density lipoprotein (HDL), low‐density lipoprotein (LDL)) with the exception of triglycerides and leptin, however data for leptin were derived from three intervention‐comparisons from a single study (Djuric 2002a; Djuric 2002b; Djuric 2002c), with very wide 95% confidence intervals suggestive of imprecision. Other biomarkers including sex hormones, adiponectin and inflammatory markers were unable to be meta‐analysed due to too few trials reporting the outcome. Although one explanation for these biomarker outcomes is that the duration of these weight loss interventions was not long enough to observe statistically significant differences in these biomarkers, subgroup analysis revealed duration of follow‐up did not significantly affect other outcomes, including anthropometric measures (change in body weight, BMI, and waist circumference).

In terms of quality of life (QOL), the weight loss interventions appeared to improve overall QOL, including improved physical and mental health subscales compared to controls. In contrast, there were no significant differences in emotional, social and anxiety/depression subscales, however the social and anxiety/depression subscales were marginally non‐statistically significant. This discrepancy in the significance between mental health subscales and anxiety/depression subscales suggests the intervention may improve aspects of mental health other than anxiety and depression, however this is difficult to interpret due to the low number of intervention‐comparisons for both outcomes (three studies) and the marginally statistically insignificant P value for the anxiety/depression subscales.

Overall completeness and applicability of evidence

This review included 23 intervention‐comparisons from 20 randomised trials which assessed body weight loss interventions in overweight or obese breast cancer survivors. From these 20 studies, 2360 breast cancer survivors were randomised, and we analysed 2028 (85.9%) of these participants (refer to Table 1). The remaining 332 (14.1%) participants were unable to be analysed primarily due to attrition. The attrition rate is in part explained by our decision to include data on the latest follow‐up provided for each study. The ratio of patients analysed to patients randomised varied considerably by outcome, with ranges of 77.4% to 80.4% for anthropometric outcomes, 52.7% to 73.8% for biomarker outcomes and 73.5% to 92.5% for QOL outcomes. The number of intervention‐comparisons (and studies) analysed per outcome was also highly variable, with several outcomes having an insufficient number of intervention‐comparisons to be included in the meta‐analyses. These outcomes included survival outcomes (overall survival and disease‐free survival), biomarkers including sex hormones (oestradiol and testosterone), insulin growth factor IGF‐1, adiponectin and inflammatory markers ( tumour necrosis factor alpha (TNF‐α), C‐reactive protein (CRP) and interleukin‐6 (IL‐6)). As such, this review was unable to determine the impact of weight loss interventions in overweight or obese breast cancer survivors on these outcomes. In contrast, the anthropometric outcomes had the highest number of intervention‐comparisons (and sample size), and this is explained by our inclusion criteria specifying that the aim of the interventions had to be weight loss. Whilst we had sufficient data to meta‐analyse all QOL outcomes, two of the subscales only had data from three intervention‐comparisons derived from three studies (mental health subscales and anxiety and depression subscales). In addition, all data for leptin was obtained from three intervention‐comparisons derived from a single study (Djuric 2002a; Djuric 2002b; Djuric 2002c), thus further studies are required to investigate this outcome. Despite our extensive search strategy, no eligible trials regarding pharmacological therapy or bariatric surgery met the inclusion criteria for this review.

Regarding external validity, the included studies assessed many different types of weight loss interventions (see Characteristics of included studies table). These interventions were grouped based on the core elements of the intervention such as diet, exercise and psychological components. There are various forms of dietary interventions (e.g. low calorie diets, special diets such as modified‐Atkins diet, use of dietician consultations and commercial dietary programs where mode of delivery varied), exercise interventions (e.g. different type of exercise such as aerobic and/or resistance exercise of varying intensity, conducted in individual or group settings, supervised or home‐based) and psychosocial interventions (e.g. counselling, which could be either individualised in‐person (versus group), mode of delivery such as online or via telephone, and use of techniques such as cognitive behavioural therapy (CBT)). Studies used various combinations of these three core elements (diet, exercise and psychosocial support). A variety of different controls were utilised, including usual care (e.g. written materials regarding healthy living guidelines), wait‐list, placebo and active control groups. Control groups received usual care (or additional care e.g. active control groups) in 20 of 23 (87%) intervention‐comparisons (17 of 20 studies). Of the remaining three studies, two intervention‐comparisons (from two studies) utilised a wait‐list approach where it was not clear whether participants received usual care during the wait‐list period (Demark‐Wahnefried 2012a; Greenlee 2013), and the remaining intervention‐comparison (from one study) utilised a placebo with no mention of usual care (Stendell‐Hollis 2010). This suggests that the observed significant differences between the intervention and control groups in this review are representative of how these active interventions may improve outcomes compared to the current standard of care for overweight and obese breast cancer survivors. This was especially true for combined interventions utilising 'diet and exercise' or 'diet and exercise and psychosocial support' owing to the significantly favourable outcomes compared to controls, and therefore usual care. Trials were conducted in many different countries and in a range of ethnic groups. We conducted subgroup analyses on anthropometric outcomes, and there were no significant differences stratified by ethnicity or menopausal status, suggesting the interventions had similar effects on these populations. Thus, the findings in this review are likely to be applicable to overweight and obese breast cancer survivors in general.

Quality of the evidence

We used the GRADE assessment criteria to describe the quality of the evidence provided for each outcome in the 'Summary of findings' tables, except for breast cancer‐specific survival and disease‐free survival because there were no data available on these two outcomes (refer to summary of findings Table 1; summary of findings Table 2). The quality of the evidence ranged from low to high due to several factors including substantial heterogeneity and a high risk of bias. There was a high risk of performance bias in 96% of intervention‐comparisons due to the nature of weight loss interventions, and difficulty in blinding participants in almost all interventions. This lack of blinding of participants also led to a high risk of bias regarding blinding of outcome assessors for patient‐reported outcomes (87% of intervention‐comparisons). There was also evidence of attrition bias, with anthropometric outcomes having attrition rates ranging from 19.6% to 22.6%.

Regarding the comparison between all types of interventions versus controls, the quality of the evidence for breast cancer recurrence was downgraded by two levels to 'low' due to 'imprecision' because the 95% confidence intervals for the risk ratio estimate were very wide (suggesting that the intervention may reduce the risk of progression by up to 32% or increase the risk of progression by up to 460%), and this was based on only 14 events in 281 patients. The quality of the evidence for change in body weight was downgraded by two levels to 'low' due to the >20% attrition rate, lack of blinding of personnel (and assessment of patient‐reported outcomes) and substantial heterogeneity. Waist circumference and QOL quality of evidence was downgraded two levels to 'low' for similar reasons. In contrast, the quality of evidence for adverse events was graded as 'high'.

Regarding the subgroup of interventions including all three components of diet, exercise and psychosocial support versus controls, we graded the quality of the evidence only for the outcomes 'change in body weight' and 'change in waist circumference' as there was an insufficient number of studies in other outcomes to conduct subgroup analyses. The quality of the evidence and justifications for these two outcomes in this subgroup analysis were similar to the corresponding outcomes in the overall analyses mentioned above (subgroup change in body weight: I2 = 69%, P < 0.0001, subgroup change in waist circumference: I2 = 79%, P < 0.0001).

Potential biases in the review process

This review utilised a comprehensive search strategy involving many different databases including Cochrane Breast Cancer Group's Specialised Register, Cochrane CENTRAL, MEDLINE, Embase, WHO ICTRP, Clinicaltrial.gov and Mainland Chinese academic literature databases. The searches had no language restriction, and potentially eligible trials were translated into English for evaluation by the study authors. This extensive search strategy suggests a low probability that other relevant studies have been missed. Many trials in the full‐text screen would have been eligible for the review except they did not report outcomes in an extractable manner (e.g. not presenting data only for breast cancer patients participants with a baseline BMI ≥ 25 kg/m2). In these cases, we contacted study authors for data which could be analysed in this review, however very few authors provided the required data to be subsequently included in the review (nine of 23 total intervention‐comparisons representing six of 20 total included studies: Demark‐Wahnefried 2012a; Demark‐Wahnefried 2014a; Demark‐Wahnefried 2014b; Dittus 2018; Djuric 2002a; Djuric 2002b; Djuric 2002c; Ferrante 2017; Kwiatkowski 2017). Thus, whilst a minority of contacted authors provided analysable data, the data we did receive represented a significant proportion of our total analysis (39% of intervention‐comparisons and 30% of all included studies). Incorporating these data may have made our results more robust (e.g. as we could include the RENEW trial Demark‐Wahnefried 2012a), but it is unclear whether our results would have differed if we had received data from all contacted authors, and this may therefore have introduced bias. Funnel plots showed no evidence of publication bias for anthropometric outcomes (changes in body weight, BMI and waist circumference, respectively: Figure 4; Figure 6; Figure 8).

Figure 8
Funnel plot 3: Change in waist circumference [cm]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in waist circumference. The plot does not show substantial asymmetry (Egger's test P value 0.20).

Funnel plot 3: Change in waist circumference [cm]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in waist circumference. The plot does not show substantial asymmetry (Egger's test P value 0.20).

Agreements and disagreements with other studies or reviews

This is the first systematic review and meta‐analysis assessing the impact of weight loss interventions on survival outcomes, anthropometric outcomes, QOL and biomarkers in overweight and obese breast cancer survivors. Our lack of randomised data for survival outcomes including disease‐free survival and overall survival is in keeping with a systematic review conducted in 2017 which assessed the relationship between weight loss and mortality (both breast cancer‐specific mortality and all‐cause mortality) in overweight and obese breast cancer survivors (Jackson 2017). This systematic review by Jackson and colleagues identified only five studies assessing mortality and all were observational studies (not randomised and therefore ineligible for our review). However, even within these five non‐randomised studies, only two of these studies included only participants with stages I‐III breast cancer (not ductal carcinoma in situ (DCIS) and/or metastatic breast cancer), and neither of these two studies utilised weight loss interventions. A review article was published in 2018 following a National Cancer Policy Forum (NCPF) workshop which comprehensively summarised evidence regarding weight management throughout cancer care (Demark‐Wahnefried 2018a). There is considerable overlap between both the completed and ongoing studies included in our review and the breast cancer studies mentioned in Demark‐Wahefried and colleagues, with their review also highlighting the paucity of evidence regarding survival outcomes and recurrence in randomised weight loss trials.

Study flow diagram.

Figures and Tables -
Figure 1

Study flow diagram.

Navigate to figure in ReviewOpen in new tab

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Figures and Tables -
Figure 2

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Navigate to figure in ReviewOpen in new tab

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.2 Change in body weight [kg].

Figures and Tables -
Figure 3

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.2 Change in body weight [kg].

Navigate to figure in ReviewOpen in new tab

Funnel plot 1: Change in body weight [kg]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body weight. The plot does not show substantial asymmetry (Egger's test P value 0.40).

Figures and Tables -
Figure 4

Funnel plot 1: Change in body weight [kg]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body weight. The plot does not show substantial asymmetry (Egger's test P value 0.40).

Navigate to figure in ReviewOpen in new tab

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.3 Change in body mass index [kg/m2].

Figures and Tables -
Figure 5

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.3 Change in body mass index [kg/m2].

Navigate to figure in ReviewOpen in new tab

Funnel plot 2: Change in body mass index [kg/m2]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body mass index. The plot does not show substantial asymmetry (Egger's test P value 0.13).

Figures and Tables -
Figure 6

Funnel plot 2: Change in body mass index [kg/m2]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in body mass index. The plot does not show substantial asymmetry (Egger's test P value 0.13).

Navigate to figure in ReviewOpen in new tab

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.4 Change in waist circumference [cm].

Figures and Tables -
Figure 7

Forest plot of comparison: 1 All weight loss interventions vs controls (no subgrouping), outcome: 1.4 Change in waist circumference [cm].

Navigate to figure in ReviewOpen in new tab

Funnel plot 3: Change in waist circumference [cm]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in waist circumference. The plot does not show substantial asymmetry (Egger's test P value 0.20).

Figures and Tables -
Figure 8

Funnel plot 3: Change in waist circumference [cm]. Assessing publication bias and/or small‐study effects. Plot includes all intervention‐comparisons with extractable data for change in waist circumference. The plot does not show substantial asymmetry (Egger's test P value 0.20).

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 1: Cancer recurrence

Figures and Tables -
Analysis 1.1

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 1: Cancer recurrence

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 2: Change in body weight

Figures and Tables -
Analysis 1.2

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 2: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 3: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 1.3

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 3: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 4: Change in waist circumference

Figures and Tables -
Analysis 1.4

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 4: Change in waist circumference

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 5: Adverse events

Figures and Tables -
Analysis 1.5

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 5: Adverse events

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 6: Change in quality of life ‐ overall scales

Figures and Tables -
Analysis 1.6

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 6: Change in quality of life ‐ overall scales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 7: Change in quality of life ‐ physical subscales

Figures and Tables -
Analysis 1.7

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 7: Change in quality of life ‐ physical subscales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 8: Change in quality of life ‐ social subscales

Figures and Tables -
Analysis 1.8

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 8: Change in quality of life ‐ social subscales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 9: Change in quality of life ‐ emotional subscales

Figures and Tables -
Analysis 1.9

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 9: Change in quality of life ‐ emotional subscales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 10: Change in quality of life ‐ mental health subscales

Figures and Tables -
Analysis 1.10

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 10: Change in quality of life ‐ mental health subscales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 11: Change in quality of life ‐ anxiety/depression subscales

Figures and Tables -
Analysis 1.11

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 11: Change in quality of life ‐ anxiety/depression subscales

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 12: Change in insulin

Figures and Tables -
Analysis 1.12

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 12: Change in insulin

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 13: Change in glucose

Figures and Tables -
Analysis 1.13

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 13: Change in glucose

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 14: Change in total cholesterol

Figures and Tables -
Analysis 1.14

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 14: Change in total cholesterol

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 15: Change in HDL cholesterol

Figures and Tables -
Analysis 1.15

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 15: Change in HDL cholesterol

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 16: Change in LDL cholesterol

Figures and Tables -
Analysis 1.16

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 16: Change in LDL cholesterol

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 17: Change in triglycerides

Figures and Tables -
Analysis 1.17

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 17: Change in triglycerides

Navigate to figure in ReviewOpen in new tab

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 18: Change in leptin

Figures and Tables -
Analysis 1.18

Comparison 1: All weight loss interventions vs controls (no subgrouping), Outcome 18: Change in leptin

Navigate to figure in ReviewOpen in new tab

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 1: Change in body weight

Figures and Tables -
Analysis 2.1

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 1: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 2: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 2.2

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 2: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 3: Change in waist circumference

Figures and Tables -
Analysis 2.3

Comparison 2: Subgrouped by Intervention type vs control type, Outcome 3: Change in waist circumference

Navigate to figure in ReviewOpen in new tab

Comparison 3: Subgrouped by ethnicity, Outcome 1: Change in body weight

Figures and Tables -
Analysis 3.1

Comparison 3: Subgrouped by ethnicity, Outcome 1: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 3: Subgrouped by ethnicity, Outcome 2: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 3.2

Comparison 3: Subgrouped by ethnicity, Outcome 2: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 3: Subgrouped by ethnicity, Outcome 3: Change in waist circumference

Figures and Tables -
Analysis 3.3

Comparison 3: Subgrouped by ethnicity, Outcome 3: Change in waist circumference

Navigate to figure in ReviewOpen in new tab

Comparison 4: Subgrouped by menopausal status, Outcome 1: Change in body weight

Figures and Tables -
Analysis 4.1

Comparison 4: Subgrouped by menopausal status, Outcome 1: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 4: Subgrouped by menopausal status, Outcome 2: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 4.2

Comparison 4: Subgrouped by menopausal status, Outcome 2: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 4: Subgrouped by menopausal status, Outcome 3: Change in waist circumference

Figures and Tables -
Analysis 4.3

Comparison 4: Subgrouped by menopausal status, Outcome 3: Change in waist circumference

Navigate to figure in ReviewOpen in new tab

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 1: Change in body weight

Figures and Tables -
Analysis 5.1

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 1: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 2: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 5.2

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 2: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 3: Change in waist circumference

Figures and Tables -
Analysis 5.3

Comparison 5: Subgrouped by duration of follow‐up (months), Outcome 3: Change in waist circumference

Navigate to figure in ReviewOpen in new tab

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 1: Change in body weight

Figures and Tables -
Analysis 6.1

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 1: Change in body weight

Navigate to figure in ReviewOpen in new tab

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 2: Change in body mass index [kg/m2]

Figures and Tables -
Analysis 6.2

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 2: Change in body mass index [kg/m2]

Navigate to figure in ReviewOpen in new tab

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 3: Change in waist circumference

Figures and Tables -
Analysis 6.3

Comparison 6: Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach, Outcome 3: Change in waist circumference

Navigate to figure in ReviewOpen in new tab
Summary of findings 1. All weight loss interventions for overweight and obese breast cancer survivors

All weight loss interventions compared to comparator groups for overweight and obese breast cancer survivors

Patient or population: overweight and obese breast cancer survivors
Setting: various settings (individual or group‐based, in‐person or remote, exercise centre or external location)
Intervention: all weight loss interventions (diet, diet & exercise, diet & psychosocial, diet & exercise & psychosocial)
Comparison: control programs (usual care, wait‐list, written materials, placebo or active control)

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(treatment‐ comparisons)

Quality of the evidence
(GRADE)

Comments

Risk with control programs

Risk with all weight loss interventions

Overall survival

‐

‐

‐

‐

‐

No trials reported this outcome as time‐to‐event data.

Breast cancer recurrence

Study population

RR 1.95
(0.68 to 5.60)

281 (4)

⊕⊕⊝⊝
LOW 1

32 per 1,000

62 per 1,000
(22 to 178)

Change in body weight

The mean body weight loss4 was 1.01 kg

MD 2.25 kg lower
(3.19 lower to 1.30 lower)

‐

1751 (21)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.00001, I2 = 69%

Change in BMI

The mean BMI reduction4 was 0.42 kg/m2

MD 1.08 kg/m2 lower (1.61 lower to 0.56 lower)

1353 (17)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.0001, I2 = 84%

Change in waist circumference

The mean waist circumference4 reduction was ‐0.28 cm (i.e. 0.28 cm increase)

MD 1.73 cm lower
(3.17 lower to 0.29 lower)

‐

1193 (13)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.0001, I2 = 73%

Disease‐free survival

‐

‐

‐

‐

‐

No trials reported this outcome.

Adverse events

Study population

RR 0.94
(0.76 to 1.17)

394 (4)

⊕⊕⊕⊕
HIGH

471 per 1,000

443 per 1,000
(358 to 551)

Change in quality of life ‐ overall scales

‐

SMD 0.74 higher
(0.20 higher to 1.29 higher)

‐

867 (10)

⊕⊕⊝⊝
LOW 2 3

Heterogeneity: P < 0.00001, I2 = 89%; An SMD of 0.74 is a moderate sized effect according to Cohen’s interpretation of effect sizes.

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; MD: mean difference; RR: Risk ratio.

GRADE Working Group grades of evidence

High quality: further research is very unlikely to change our confidence in the estimate of effect.

Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality: we are very uncertain about the estimate.

1Quality of evidence downgraded two levels for 'imprecision' because 95% CI for risk ratio estimate suggests that the intervention might reduce the risk of progression by up to 32% or increase the risk of progression by up to 460% and risk ratio estimate is based on only 14 events in 281 patients.
2Quality of evidence downgraded one level for 'risk of bias' because >20% attrition of randomised participants for measurement of this outcome at follow‐up and because most studies were unblinded for participants, personnel and assessment of patient‐reported outcomes.
3Quality of evidence downgraded one level for 'inconsistency' because I‐squared statistic suggests substantial heterogeneity.
4Calculated as inverse variance weighted average of control group measurements.

Figures and Tables -
Summary of findings 1. All weight loss interventions for overweight and obese breast cancer survivors
Navigate to table in Review
Summary of findings 2. Weight loss interventions involving diet, exercise and psychosocial support for overweight and obese breast cancer survivors

Weight loss interventions involving all three diet, exercise and psychosocial support for overweight and obese breast cancer survivors

Patient or population: overweight and obese breast cancer survivors
Setting: various settings (individual or group‐based, in‐person or remote, exercise centre or external location)
Intervention: weight loss interventions involving the following components: diet, exercise and psychosocial support
Comparison: control programs (usual care, wait‐list, written materials or placebo)

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(treatment‐ comparisons)

Quality of the evidence
(GRADE)

Comments

Risk with control programs

Risk with All weight loss interventions

Overall survival

‐

‐

‐

‐

‐

No trials reported this outcome as time‐to‐event data.

Breast cancer recurrence

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

Change in body weight

The mean body weight loss3 was 0.97 kg

MD 2.88 kg lower
(3.98 lower to 1.77 lower)

‐

1526 (13)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 69%.

Change in BMI

The mean BMI reduction3 was 0.41 kg/m2

MD 1.44 kg/m2 lower (2.16 lower to 0.72 lower)

1187 (11)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 89%.

Change in waist circumference

The mean waist circumference3 reduction was ‐0.33 cm (.i.e. 0.33 cm increase)

MD 1.66 cm lower
(3.49 lower to 0.16 lower)

‐

1021 (8)

⊕⊕⊝⊝
LOW 1 2

Heterogeneity: P < 0.0001, I2 = 79%.

Disease‐free survival ‐ not reported

‐

‐

‐

‐

‐

No trials reported this outcome.

Adverse events

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

Change in quality of life ‐ overall scales

‐

‐

‐

‐

‐

Subroup analyses not performed for this outcome (insufficient number of studies per subgroup).

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence

High quality: further research is very unlikely to change our confidence in the estimate of effect.

Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality: we are very uncertain about the estimate.

1Quality of evidence downgraded one level for 'risk of bias' because > 20% attrition of randomised participants for measurement of this outcome at follow‐up and because most studies were unblinded for participants, personnel and assessment of patient‐reported outcomes.
2Quality of evidence downgraded one level for 'inconsistency' because I2 statistic suggests substantial heterogeneity.
3Calculated as inverse variance weighted average of control group measurements.

Figures and Tables -
Summary of findings 2. Weight loss interventions involving diet, exercise and psychosocial support for overweight and obese breast cancer survivors
Navigate to table in Review
Table 1. Number of studies, intervention‐comparisons, patients randomised and patients analysed by outcome

Included in meta‐analysis

Outcome

Studies reporting outcome
N

Studies
N

Intervention‐comparisons
N

Total patients analysed/ randomised (%)

Intervention patients analysed/ randomised (%)

Control patients analysed/ randomised (%)

Overall

20

20

23

2013/2360 (85.3%)

1056/1209 (87.3%)

957/1151 (83.1%)

Overall survival

1

0

0

0

0

0

Breast Cancer recurrence

4

4

4

281/281 (100.0%)

155/155 (100.0%)

126/126 (100.0%)

Change in body weight

19

18

21

1751/2190 (80.0%)

920/1122 (82.0%)

831/1068 (77.8%)

Change in body mass index

15

14

17

1353/1682 (80.4%)

714/862 (82.8%)

639/820 (77.9%)

Change in waist circumference

12

12

13

1193/1541 (77.4%)

634/800 (79.3%)

559/741 (75.4%)

Disease‐free survival

0

0

0

0

0

0

Adverse events

5

3

4

394/446 (88.3%)

205/229 (89.5%)

189/217 (87.1%)

Change in quality of life ‐ overall

8

8

10

867/1148 (75.5%)

447/578 (77.3%)

420/570 (73.7%)

Change in quality of life ‐ physical subscales

7

7

10

1024/1351 (75.8%)

530/679 (78.1%)

494/672 (73.5%)

Change in quality of life ‐ social subscales

4

4

6

389/464 (83.8%)

196/228 (86.0%)

193/236 (81.8%)

Change in quality of life ‐ emotional subscales

5

5

8

498/633 (78.7%)

264/321 (82.2%)

234/312 (75.0%)

Change in quality of life ‐ mental health subscales

3

3

3

355/400 (88.8%)

173/200 (86.5%)

182/200 (91.0%)

Change in quality of life ‐ anxiety depression subscales

3

3

3

669/910 (73.5%)

340/457 (74.4%)

329/453 (72.6%)

Change in insulin

4

4

6

134/192 (69.8%)

79/95 (83.2%)

55/97 (56.7%)

Change in glucose

4

4

6

133/192 (69.3%)

78/95 (82.1%)

55/97 (56.7%)

Change in total cholesterol

5

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in HDL cholesterol

5

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in LDL cholesterol

4

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in triglycerides

4

4

6

189/256 (73.8%)

116/141 (82.3%)

73/115 (63.5%)

Change in leptin

3

1

3

39/74 (52.7%)

27/35 (77.1%)

12/39 (30.8%)

HDL: high‐density lipoprotein
LDL: low‐density lipoprotein

Figures and Tables -
Table 1. Number of studies, intervention‐comparisons, patients randomised and patients analysed by outcome
Navigate to table in Review
Comparison 1. All weight loss interventions vs controls (no subgrouping)

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1.1 Cancer recurrence Show forest plot

4

281

Risk Ratio (M‐H, Random, 95% CI)

1.95 [0.68, 5.60]

1.2 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Random, 95% CI)

‐2.25 [‐3.19, ‐1.30]

1.3 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Random, 95% CI)

‐1.08 [‐1.61, ‐0.56]

1.4 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Random, 95% CI)

‐1.73 [‐3.17, ‐0.29]

1.5 Adverse events Show forest plot

4

394

Risk Ratio (M‐H, Random, 95% CI)

0.94 [0.76, 1.17]

1.6 Change in quality of life ‐ overall scales Show forest plot

10

867

Std. Mean Difference (IV, Random, 95% CI)

0.74 [0.20, 1.29]

1.7 Change in quality of life ‐ physical subscales Show forest plot

10

1024

Std. Mean Difference (IV, Random, 95% CI)

0.33 [0.10, 0.56]

1.8 Change in quality of life ‐ social subscales Show forest plot

6

389

Std. Mean Difference (IV, Random, 95% CI)

0.19 [‐0.01, 0.39]

1.9 Change in quality of life ‐ emotional subscales Show forest plot

8

498

Std. Mean Difference (IV, Random, 95% CI)

0.11 [‐0.09, 0.30]

1.10 Change in quality of life ‐ mental health subscales Show forest plot

3

355

Std. Mean Difference (IV, Random, 95% CI)

0.60 [0.17, 1.02]

1.11 Change in quality of life ‐ anxiety/depression subscales Show forest plot

3

669

Std. Mean Difference (IV, Random, 95% CI)

0.63 [‐0.07, 1.33]

1.12 Change in insulin Show forest plot

6

134

Mean Difference (IV, Random, 95% CI)

‐1.49 [‐6.62, 3.64]

1.13 Change in glucose Show forest plot

6

133

Mean Difference (IV, Random, 95% CI)

‐0.46 [‐4.86, 3.93]

1.14 Change in total cholesterol Show forest plot

6

189

Mean Difference (IV, Random, 95% CI)

‐0.34 [‐0.84, 0.16]

1.15 Change in HDL cholesterol Show forest plot

6

189

Mean Difference (IV, Random, 95% CI)

0.00 [‐0.08, 0.08]

1.16 Change in LDL cholesterol Show forest plot

6

189

Mean Difference (IV, Random, 95% CI)

‐0.18 [‐0.59, 0.22]

1.17 Change in triglycerides Show forest plot

6

189

Mean Difference (IV, Random, 95% CI)

‐0.26 [‐0.45, ‐0.07]

1.18 Change in leptin Show forest plot

3

39

Mean Difference (IV, Random, 95% CI)

‐14.67 [‐26.36, ‐2.98]
Figures and Tables -
Comparison 1. All weight loss interventions vs controls (no subgrouping)
Navigate to table in Review
Comparison 2. Subgrouped by Intervention type vs control type

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

2.1 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Random, 95% CI)

‐2.25 [‐3.19, ‐1.30]

2.1.1 Diet vs usual care, wait‐list, written materials or placebo

3

72

Mean Difference (IV, Random, 95% CI)

0.09 [‐3.88, 4.05]

2.1.2 Diet + exercise vs usual care, wait‐list, written materials or placebo

3

93

Mean Difference (IV, Random, 95% CI)

‐2.03 [‐3.48, ‐0.58]

2.1.3 DIet + exercise + psychosocial vs usual care, wait‐list, written materials or placebo

13

1526

Mean Difference (IV, Random, 95% CI)

‐2.88 [‐3.98, ‐1.77]

2.1.4 Diet + psychosocial vs active control (diet + psychosocial)

1

40

Mean Difference (IV, Random, 95% CI)

0.40 [‐2.67, 3.47]

2.1.5 Diet + exercise vs active control (diet + exercise)

1

20

Mean Difference (IV, Random, 95% CI)

‐3.70 [‐8.35, 0.95]

2.2 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Random, 95% CI)

‐1.08 [‐1.61, ‐0.56]

2.2.1 Diet vs usual care, wait‐list, written materials or placebo

3

72

Mean Difference (IV, Random, 95% CI)

‐0.15 [‐1.57, 1.26]

2.2.2 Diet + exercise vs usual care, wait‐list, written materials or placebo

2

54

Mean Difference (IV, Random, 95% CI)

‐1.05 [‐1.65, ‐0.45]

2.2.3 DIet + exercise + psychosocial vs usual care, wait‐list, written materials or placebo

11

1187

Mean Difference (IV, Random, 95% CI)

‐1.44 [‐2.16, ‐0.72]

2.2.4 Diet + psychosocial vs active control (diet + psychosocial)

1

40

Mean Difference (IV, Random, 95% CI)

0.00 [‐1.17, 1.17]

2.2.5 Diet + exercise vs active control (diet + exercise)

0

0

Mean Difference (IV, Random, 95% CI)

Not estimable

2.3 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Random, 95% CI)

‐1.73 [‐3.17, ‐0.29]

2.3.1 Diet vs usual care, wait‐list, written materials or placebo

1

39

Mean Difference (IV, Random, 95% CI)

2.30 [‐1.93, 6.53]

2.3.2 Diet + exercise vs usual care, wait‐list, written materials or placebo

3

93

Mean Difference (IV, Random, 95% CI)

‐3.63 [‐5.76, ‐1.51]

2.3.3 DIet + exercise + psychosocial vs usual care, wait‐list, written materials or placebo

8

1021

Mean Difference (IV, Random, 95% CI)

‐1.66 [‐3.49, 0.16]

2.3.4 Diet + psychosocial vs active control (diet + psychosocial)

1

40

Mean Difference (IV, Random, 95% CI)

0.80 [‐3.02, 4.62]

2.3.5 Diet + exercise vs active control (diet + exercise)

0

0

Mean Difference (IV, Random, 95% CI)

Not estimable
Figures and Tables -
Comparison 2. Subgrouped by Intervention type vs control type
Navigate to table in Review
Comparison 3. Subgrouped by ethnicity

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

3.1 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Random, 95% CI)

‐2.25 [‐3.19, ‐1.30]

3.1.1 >95% African descent

3

248

Mean Difference (IV, Random, 95% CI)

‐1.01 [‐2.09, 0.06]

3.1.2 >95% white/caucasian

4

197

Mean Difference (IV, Random, 95% CI)

‐2.93 [‐4.70, ‐1.16]

3.1.3 Mixed (≤95% of any one ethnic group) or unspecified

14

1306

Mean Difference (IV, Random, 95% CI)

‐2.36 [‐3.62, ‐1.09]

3.2 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Random, 95% CI)

‐1.08 [‐1.61, ‐0.56]

3.2.1 >95% African descent

2

56

Mean Difference (IV, Random, 95% CI)

‐0.28 [‐1.38, 0.82]

3.2.2 >95% white/caucasian

2

88

Mean Difference (IV, Random, 95% CI)

‐0.61 [‐2.56, 1.34]

3.2.3 Mixed (≤95% of any one ethnic group) or unspecified

13

1209

Mean Difference (IV, Random, 95% CI)

‐1.31 [‐1.93, ‐0.68]

3.3 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Random, 95% CI)

‐1.73 [‐3.17, ‐0.29]

3.3.1 >95% African descent

3

248

Mean Difference (IV, Random, 95% CI)

‐0.37 [‐1.98, 1.24]

3.3.2 >95% white/caucasian

2

143

Mean Difference (IV, Random, 95% CI)

‐0.65 [‐7.89, 6.60]

3.3.3 Mixed (≤95% of any one ethnic group) or unspecified

8

802

Mean Difference (IV, Random, 95% CI)

‐2.33 [‐3.87, ‐0.79]
Figures and Tables -
Comparison 3. Subgrouped by ethnicity
Navigate to table in Review
Comparison 4. Subgrouped by menopausal status

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

4.1 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Random, 95% CI)

‐2.25 [‐3.19, ‐1.30]

4.1.1 100% postmenopausal women

7

536

Mean Difference (IV, Random, 95% CI)

‐2.53 [‐4.00, ‐1.06]

4.1.2 Mix of pre‐ and post‐ menopausal women or unspecified

14

1215

Mean Difference (IV, Random, 95% CI)

‐2.11 [‐3.39, ‐0.82]

4.2 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Random, 95% CI)

‐1.08 [‐1.61, ‐0.56]

4.2.1 100% postmenopausal women

5

379

Mean Difference (IV, Random, 95% CI)

‐0.70 [‐1.28, ‐0.12]

4.2.2 Mix of pre‐ and post‐ menopausal women or unspecified

12

974

Mean Difference (IV, Random, 95% CI)

‐1.34 [‐2.16, ‐0.53]

4.3 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Random, 95% CI)

‐1.73 [‐3.17, ‐0.29]

4.3.1 100% postmenopausal women

3

94

Mean Difference (IV, Random, 95% CI)

‐2.83 [‐6.68, 1.01]

4.3.2 Mix of pre‐ and post‐ menopausal women or unspecified

10

1099

Mean Difference (IV, Random, 95% CI)

‐1.43 [‐3.06, 0.20]
Figures and Tables -
Comparison 4. Subgrouped by menopausal status
Navigate to table in Review
Comparison 5. Subgrouped by duration of follow‐up (months)

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

5.1 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Random, 95% CI)

‐2.25 [‐3.19, ‐1.30]

5.1.1 ≤6 months

11

438

Mean Difference (IV, Random, 95% CI)

‐1.84 [‐3.64, ‐0.04]

5.1.2 > 6 months, ≤12 months

7

535

Mean Difference (IV, Random, 95% CI)

‐2.23 [‐3.64, ‐0.83]

5.1.3 >12 months, ≤36 months

3

778

Mean Difference (IV, Random, 95% CI)

‐2.68 [‐4.05, ‐1.30]

5.2 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Random, 95% CI)

‐1.08 [‐1.61, ‐0.56]

5.2.1 ≤6 months

9

370

Mean Difference (IV, Random, 95% CI)

‐1.27 [‐2.33, ‐0.20]

5.2.2 > 6 months, ≤12 months

6

343

Mean Difference (IV, Random, 95% CI)

‐1.04 [‐1.85, ‐0.23]

5.2.3 >12 months, ≤36 months

2

640

Mean Difference (IV, Random, 95% CI)

‐0.33 [‐1.50, 0.85]

5.3 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Random, 95% CI)

‐1.73 [‐3.17, ‐0.29]

5.3.1 ≤6 months

8

363

Mean Difference (IV, Random, 95% CI)

‐1.88 [‐3.84, 0.07]

5.3.2 > 6 months, ≤12 months

3

246

Mean Difference (IV, Random, 95% CI)

‐2.86 [‐6.48, 0.77]

5.3.3 >12 months, ≤36 months

2

584

Mean Difference (IV, Random, 95% CI)

0.21 [‐5.26, 5.69]
Figures and Tables -
Comparison 5. Subgrouped by duration of follow‐up (months)
Navigate to table in Review
Comparison 6. Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

6.1 Change in body weight Show forest plot

21

1751

Mean Difference (IV, Fixed, 95% CI)

‐2.04 [‐2.48, ‐1.60]

6.2 Change in body mass index [kg/m2] Show forest plot

17

1353

Mean Difference (IV, Fixed, 95% CI)

‐0.67 [‐0.83, ‐0.50]

6.3 Change in waist circumference Show forest plot

13

1193

Mean Difference (IV, Fixed, 95% CI)

‐2.47 [‐3.09, ‐1.86]
Figures and Tables -
Comparison 6. Sensitivity analyses ‐ analyses 1.1, 1.2, 1.3 repeated but with fixed effect approach
Navigate to table in Review