https://orcid.org/0000-0001-8331-8310
Figures
Abstract
Objective
Neighbourhood deprivation has been found to be associated with many health conditions, but its association with low back pain (LBP) and arthritis is unclear. This study aimed to examine the association between neighbourhood deprivation with LBP and arthritis, and its potential interaction with individual socioeconomic status (SES) on these outcomes.
Methods
Monozygotic (MZ) twins from the Washington State Twin Registry were used to control for genetic and common environmental factors that could otherwise confound the purported relationship. Multilevel models were employed to examine the association between neighbourhood deprivation as well as individual-level SES with LBP/arthritis, adjusting for age, sex, body mass index (BMI) and residence rurality.
Results
There were 6,380 individuals in the LBP sample and 2,030 individuals in the arthritis sample. Neighbourhood deprivation was not associated with LBP (P = 0.26) or arthritis (P = 0.61), and neither was its interaction with individual-level SES. People without a bachelor’s degree were more likely to report LBP (OR 1.44, 95% CI 1.26–1.65) or both LBP and arthritis (OR 1.67, 95% CI 1.14–2.45) than those with a bachelor’s degree, but not for arthritis alone (P = 0.17). Household income was not significantly associated with LBP (P = 0.16) or arthritis (p = 0.23) independent of age, sex, and BMI.
Citation: Feng Y, Bowden JL, Hunter DJ, Ferreira P, Duncan GE (2024) Does neighbourhood deprivation influence low back pain and arthritis: An empirical study using multilevel twin design. PLoS ONE 19(4): e0298356. https://doi.org/10.1371/journal.pone.0298356
Editor: Koji Akeda, Mie University Graduate School of Medicine, JAPAN
Received: August 19, 2022; Accepted: January 23, 2024; Published: April 26, 2024
Copyright: © 2024 Feng et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data supporting the results of the present study are owned by the Washington State Twin Registry (WSTR). Thus, the data cannot be publicly shared as it involves third-party data. However, researchers wanting to gain access to the data can do so by contacting the WSTR and completing the appropriate forms stipulated in the WSTR Policies & Procedures guidelines. Application information can be sent to the Scientific Operations Manager at the following URL (https://wstwinregistry.org/contact-us/) or via email (ws.twinregistry@wsu.edu).
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Musculoskeletal diseases, including low back pain (LBP) and arthritis, are leading causes of disability [1–3]. In 2017, the global prevalence of LBP and arthritis was 7.5% and 4.4% respectively [4, 5]. Global years lived with disability (YLD) reached 823 for LBP and 245 for osteoarthritis per 100,000 population in 2019. LBP and arthritis create an enormous medical and economic burden on individuals and societies, posing major international public health challenges [6, 7]. Improving our understanding and awareness of the risk factors for LBP and arthritis is, therefore, vitally important in developing intervention strategies to reduce the increasing burden of these conditions.
Important risk factors for LBP and arthritis include genetics, age, sex, body mass index (BMI) and socioeconomic status (SES, e.g., education, income) [8–15]. However, increasing evidence has shown that a wide variety of environmental and neighbourhood factors, such as neighbourhood deprivation, access to green space, and rurality of residence can also have a profound effect on one’s health [16–28]. These factors, together with individual-level factors, may jointly affect the health of our communities at a population level [29]. Higher neighbourhood-level deprivation has been associated with higher risks of coronary heart disease, increased depression, and poorer overall health outcomes [4, 30–33], but research on its association with LBP or arthritis remains limited. Some studies indicated a higher prevalence of back pain [34] and arthritis [35–37] in more deprived areas, but whether and how neighbourhood deprivation interacts with individual-level risk factors, jointly affecting the prevalence of LBP and/or arthritis, needs further investigation [35]. Recently, greater consideration and better understanding of the role of social determinants of health for people with osteoarthritis, especially in low-resourced settings, has been noted as a key focus in the Lancet Commission [38]. Assessing which social and environmental determinants of health are important for musculoskeletal care is a growing topic to reduce inequity and ensure equality of care for these conditions [39].
In this study, we hypothesised that higher neighbourhood deprivation would be associated with a higher prevalence of LBP and/or arthritis. We tested the hypothesis through a novel multilevel twin design, controlling for genetic and shared environmental factors that might otherwise confound the relationship between LBP/arthritis and neighbourhood-level deprivation.
Materials and methods
Participants
We used survey data from the community-based Washington State Twin Registry (WSTR) during 2009–2018. The WSTR contains twins over 18 years old, identified by the Washington State Department of Licensing records. Details related to recruitment and data collection can be found elsewhere [31, 40, 41]. Briefly, the surveys were completed by twins independently upon enrolment, with information on a range of socio-demographics and health outcomes.
We collected Monozygotic (MZ) twin data for our analysis to fully control for genetics and shared common environmental factors as MZ Twins share nearly 100% of their genetic and many shared environmental influences.
Outcome variables
We examined two outcome variables, using participants’ answers to the following questions collected from the survey:
- doctor-diagnosed LBP: “Has a doctor or other medical professional ever diagnosed you with low back pain?” (yes or no)
- doctor-diagnosed arthritis: “Has a doctor or other medical professional ever diagnosed you with arthritis?” (yes or no)
We conducted two separate primary analyses for LBP and arthritis using two different samples, and a supplementary analysis on MZ twins containing both conditions, LBP and arthritis, as multi-site pain is emerging as a new geriatric syndrome that could have synergistic effects on patients. Multivariate multilevel models with two response variables will be used in this supplementary analysis.
Socioeconomic status (SES)
The primary exposures of interest included one neighbourhood-level measure of SES and two individual-level measures of SES. Neighbourhood-level SES, or deprivation, was derived by linking the individual’s residential address at the time of the survey [41] to the deprivation levels of the zip codes of residential neighbourhoods, measured in the Singh Index [42]. In the U.S., zip code represents a relatively small neighbourhood area covering on average 85 square miles with a mean population of approximately 8,000 people. The Singh Index is a composite SES measure that draws on 17 indicators for education, employment, income and poverty, and housing at the census tract level. A higher Singh Index equates to higher levels of deprivation or lower SES. The average of the Singh Index from the 2010 and 2015 censuses was used to measure neighbourhood deprivation level.
Individual-level measures of SES included educational attainment and household income. Educational attainment was used as a dichotomous variable indicating having completed a bachelor’s degree or not, as derived from the original five qualifications. Household income was categorised into three categories (less than $40,000, $40,000–$79,999, $80,000 or more), collapsed from the original eight categories.
Confounding variables
Factors considered as potential confounders of the associations between neighbourhood deprivation and LBP/arthritis were individual age, sex, BMI and neighbourhood rurality, as extensive evidence has shown that they are important risk factors for LBP and arthritis [43, 44]. Age was classified into 4 groups (18–24, 25–39, 40–54 and > = 55 years), sex was dichotomised (male/female) and BMI was categorised into four groups according to the WHO criteria [45]: underweight (<18.5 kg/m2), normal weight (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), and obese (≥30.0 kg/m2).
At the neighbourhood level, an urban-rural classification was considered, as some evidence showed that people living in rural areas had a higher prevalence of LBP and arthritis than those living in urban areas [17, 18]. If not considered, rurality might be confounded with the effect of neighbourhood deprivation.
Statistical analysis
Twins form a natural two-level hierarchy, whereby twin individuals (Level-1 units) are clustered within twin pairs (Level-2 units). At the same time, twin individuals do not necessarily reside in the same area, rather they might live in different neighbourhoods. Therefore, people living in the same neighbourhood form another hierarchical clustering structure, creating a cross-classified structure. We used cross-classified multilevel models (MLM) [46] to account for such an overlapping non-hierarchical structure (Fig 1). In addition, MLM does not require complete records from both twins, hence preserving a relatively large sample without the need to remove the records with missing values for just one of the twins.
MLM is a re-parameterisation of the Classical Twin Model (CTM) [47, 48]. Traditionally, the CTM is often used to model twin data. However, MLM has the advantage of estimating the variance components attributable to higher-level clustering, such as the clustering of twins and neighbourhoods. In the CTM, the phenotypic variance is decomposed into Additive genetic effects (A), Common environmental variance (C) and unique or unshared Environmental variance (E) components [49]. The MLM separates the Additive genetic variance (A) in the CTM into orthogonal components: a common additive genetic variance (AC) component shared by members of a twin pair at the higher twin level (Level-2) and a unique additive (AU) component that is unique to each individual within pairs/families at the lower individual level (Level-1). As MZ twins share 100% of genes, the A variance is completely shared by the individuals from the same twin pair and there is no variance unique for the individual (i.e., the variance of unique additive genetic component, AU component is 0). This re-parameterisation decomposes the total phenotypic variance components in MZ twins into unique Environmental (E) parts at Level 1 and common additive genetic (AC) and Common environmental (C) parts at Level 2. The unique environmental part (E) at Level-1 represents within-family influences that are unique to each individual, creating variation among individuals from the same family, while the Level-2 common parts are the factors that make MZ twins similar, representing between-family influences [50]. The Level-2 variance is shared by members of a twin pair from the same family, including both the common additive genetic variance (AC) and Common environmental variance (C). The proportion of Level-2 variance of the total unexplained variance, the intraclass correlation coefficient (ICC), measures the extent of the between-pair and within-pair variances. The higher the ICC, the more effect is attributable to Level-2 common or shared factors, regardless of its source whether from common additive genetic (AC) or environmental (C) influence. Another benefit of MLM is that it is flexible with the number of members per family: it allows just one individual in a family where the other individual has missing values. Therefore, even if data on one of the twins is missing, these records were not excluded, thus preserving a large sample size and increasing statistical power.
For the supplementary analysis, we constructed bivariate MLMs with two responses, LBP and arthritis to investigate the association between neighbourhood SES and multi-site pains, both LBP and arthritis, as such condition is emerging as a new geriatric syndrome that could have synergistic effects on patients.
A multivariate multilevel model offers advantages over a separate model for each outcome. The multivariate multilevel approach allows us to gain clinically meaningful adjusted association parameters and more efficient parameter estimates. Moreover, it provides the correlations among the outcomes at the baseline and follow-up time levels. As Tate and Pituch [51] also noted, the multivariate multilevel model provides more accurate standard errors and powerful tests of the covariates as compared to a separate modelling approach.
Multivariate multilevel logistic model models treat two outcome variables as the ‘within-individual’ measurements at the lowest level, with individuals at the next higher level and twins at the highest level [2]. They offer advantages over a separate model for each outcome as they allow more efficient parameter estimation, more accurate standard errors and more powerful tests of the covariates than a separate model for each outcome [52].
We specified the models following three steps for both the primary and supplementary analyses:
- Model 1 was a two-level MLM that included individual-level variables (age, sex, BMI, educational attainment and income);
- Model 2 added neighbourhood-level variables (Singh Index and rurality) to Model 1;
- Model 3 added cross-level interactions between neighbourhood SES and individual-level SES factors to Model 2.
Each of the MZ twins has a different level of BMI, education and income. They have been included in the models as variables operating at the individual level (Level-1) in all models. Each of the twins may also reside in different areas, hence the neighbourhood Singh Index and rurality. These factors were added to Model 2 and Model 3 as variables operating at the neighbourhood level (Level-2). For each of the models, we trialled adding neighbourhoods as the units that cross-classified with a 2-level twin design and would only proceed if such specification improves the goodness of fit of the model. Otherwise, we would construct a standard 2-level model with individuals (Level-1) nested in twins (Level-2). We also tested the potential correlation between neighbourhood-level factors (deprivation and rurality) and individual-level SES (education and income). All models were fitted in MLwiN (version 3.05) [53] and the results were reported as Odds Ratios (ORs) with 95% confidence intervals (CIs).
Ethics
Written consent was obtained from all participants. All recruitment and data collection procedures were approved by Washington State University’s Institutional Review Board (assurance number: 14515).
Results
A total of 12,590 individual twins were available from the WSTR database, 54.5% were MZ twins (6,864 individuals). The majority (75.5%) have lived or currently live in Washington State; the rest cover all 50 states and the Washington D. C. area of the United States. After excluding records with missing data (137 on LBP, 74 on BMI and 273 on income), we had a total of 6,380 individuals for LBP analyses (Fig 2(a)). The sample contained 35.3% males and was 40.6 years old (SD 17.5) on average, with a mean BMI of 25.7 (SD 5.5).
(a) Flowchart for low back pain (LBP), (b) Flowchart for arthritis analysis.
We followed similar exclusion criteria and obtained 2,030 individuals for the arthritis analysis (Fig 2(b)). The sample for arthritis was relatively smaller primarily because the responses to the question regarding arthritis were only collected from 2010 onwards and the response rates to this question were low. The arthritis sample has a median age of 40.3 (SD 16.6) years old and a mean BMI of 25.6 (SD 5.4), with 32.3% being males.
Descriptive statistics
In our study sample, 29.8% of people reported LBP and 14.8% reported arthritis (Table 1). Participants with LBP or arthritis were much older than their counterparts without the condition (median age: 48 versus 32 for people with and without LBP; 62 versus 32 for people with and without arthritis). People with LBP or arthritis condition also had higher BMI than those without the condition (Median BMI: 25.8 versus 24.0 kg/m2 for people with and without LBP; 26.6 versus 24.3kg/m2 for people with vs without arthritis). A higher percentage of the LBP sample had no bachelor’s degree or higher, regardless of the condition status. In contrast, although there were more people without a bachelor’s degree among those with arthritis (8.2% versus 6.6%), more people had a bachelor’s degree or higher amongst those without arthritis (46.8% versus 38.4%). Overall, more people reported having an annual household income of $80,000 than the other two categories in both LBP and arthritis samples. At the neighbourhood level, participants mostly lived in urban areas, regardless of whether they had been diagnosed with LBP or arthritis. People with LBP and arthritis generally lived in more deprived neighbourhoods (higher Singh scores) than those without the condition.
We did not find a strong correlation between individual confounders with neighbourhood rurality or deprivation. The Spearman’s correlation between age groups and neighbourhood rurality was -0.0448, between income and rurality was 0.0467, and between education and rurality was -0.1053. The correlations between neighbourhood deprivation and individual age, income and education were -0.0004, -0.3018 and 0.2571, respectively.
We started the statistical analyses from the standard 2-level MLM with a twin design as described in the Methods section and then tested whether adding the neighbourhood clustering would improve the model goodness-of-fit from the standard 2-level model. Results showed that this was not the case for both LBP and arthritis samples, therefore, the standard 2-level multilevel twin models were specified for all the subsequent analyses. The results for LBP and arthritis are displayed in Tables 2 and 3, respectively, with the supplementary analysis results in the S1 Table.
Analysis #1: Doctor-diagnosed LBP
The results for the LBP sample showed that LBP was significantly associated with all person-level confounders: age (P<0.001), sex (P = 0.03) and BMI (P<0.001), with an increased likelihood of having LBP amongst older people, males and people with higher BMI. For instance, people 55 years or older are more than 5 times more likely to have LBP (OR 5.40, 95% CI 4.40–6.63 in Model 1) than the 18–24 years old age group. The magnitude of such relationships remained largely unchanged from Model 1 to Model 3 when adding neighbourhood factors and interactional effects between neighbourhood SES and individual SES. Educational attainment was significantly associated with LBP independent of age, sex and BMI across all three models: people without a bachelor’s degree were almost 50% more likely to have LBP (OR 1.44, 95% CIs 1.26–1.65 in Model 3). Household income was significantly associated with LBP in Model 1 before neighbourhood-level factors were considered: people with an annual income of $80,000 or more were 18% less likely to have LBP compared with those with an annual income of $40,000 or less (OR 0.82, 95% CIs 0.70–0.96). However, the association became statistically insignificant in Models 2 and 3 once neighbourhood factors were added (P = 0.23 in Model 2 and P = 0.16 in Model 3).
The association between LBP and neighbourhood rurality (P = 0.55 in both Model 2 & 3) or neighbourhood deprivation (P = 0.35 in Model 2 and P = 0.26 in Model 3), was not significant, neither was the cross-level interaction between neighbourhood deprivation and person-level SES (P = 0.47 for interaction between neighbourhood deprivation and education; P = 0.44 for interaction between neighbourhood deprivation and household income). This was expected as the main effect of neighbourhood deprivation was insignificant. This suggested that the effects of individual education and income levels on LBP did not differ significantly according to the deprivation level of their residential neighbourhoods.
The ICC remained to be 0.14 across all three models, suggesting that 14% of the total unexplained variation was attributable to common or shared factors. These factors made MZ twins more alike to each other, after accounting for the person-level confounders, age, sex and BMI. The remaining 86% of the unexplained variation (86% = 100%−14%) was attributable to unique environmental influences (E) on one member within the same twin pair at Level-1, as there is no unique additive genetic component (AU) variance within MZ twins.
Analysis #2: Doctor-diagnosed arthritis
Similar to the results of LBP, all individual-level confounders, age, sex and BMI, were also significantly associated with arthritis. Women were more than twice as likely to have arthritis compared with their male counterparts (OR = 2.06, 95% CI 1.44–2.95 in Model 3); older adults (P<0.001) and those with higher BMI (P<0.001) were more likely to have arthritis. Person-level education (OR 1.27, CI 0.90–1.78, P = 0.17), household income (P = 0.23), neighbourhood deprivation (OR 0.91, CI 0.62–1.32, P = 0.61) and rurality (OR 1.02, CI 0.70–1.48, P = 0.92) were not significantly associated with arthritis in Model 3. The cross-level interaction between deprivation and education (P = 0.68), and between deprivation and income (P = 0.30) were also not statistically significantly associated with arthritis.
The Level-2 ICC was similar to that for LBP across all models (15%), suggesting that approximately 15% of the residual unexplained variation after accounting for all explanatory variables in the models was attributable to common factors. The rest 85% of the variation was attributable to unique environmental influences or events that happened to only one member within the twin pair.
Analysis #3: Supplementary analysis—multivariate multilevel logistic model models with two response variables (LBP and arthritis) analysed simultaneously
The results from the supplementary analysis (S1 Table) were largely aligned with the findings in the main analyses. Older adults were more likely to have arthritis and LBP than younger adults (P<0.001). Participants without a bachelor’s degree were 67% more likely to have reported arthritis and LBP (OR = 1.67, 95% CI 1.14–2.45). Women were 74% more likely than their male counterparts (OR = 1.74, 95% CI 1.16–2.60) to have arthritis and LBP, which was in line with the arthritis analyses, but opposite to the LBP results. Different from the main analyses for LBP or arthritis, BMI were not significantly associated with the outcomes when LBP and arthritis were considered together.
We did not find significant associations between household Income, neighbourhood rurality, deprivation and the risk of having both conditions and therefore did not proceed with testing the interactions between individual- and neighbourhood-level SES variables.
Discussion
Traditional risk factors for LBP and arthritis are individual age, sex, BMI, and occupation. Growing evidence shows that these individual-level risk factors do not fully account for all the associations observed. Neighbourhood deprivation is increasingly recognized as an important determinant of people’s health and wellness. In this study, we investigated the association between neighbourhood deprivation and the conditions of LBP and arthritis in a large twin dataset through an innovative modelling strategy.
Contrary to our hypothesis, and at least some of the existing literature, we did not find a significant association between neighbourhood deprivation and LBP or arthritis. We also did not find a significant association of neighbourhood rurality, or interaction between neighbourhood deprivation and individual-level SES. The majority of the unexplained variation in the prevalence of the conditions studied (86% for LBP and 85% for arthritis) was attributable to unique environmental influences (E), while only a small part was attributable to common family factors at the twin level (14% for LBP and 15% for arthritis). This suggests that other unique environment exposures rather than neighbourhood deprivation level, or events that occurred to only one twin, contributed to a large proportion of the unexplained variation in LBP and arthritis prevalence.
Comparison with previous studies
The finding of non-significant neighbourhood random effects from our study supports earlier findings [35], but the non-significant association between neighbourhood deprivation and LBP or arthritis is somewhat inconsistent with findings from previous non-twin studies. For example, it was reported that individuals from more deprived or disadvantaged neighbourhoods were more likely to report back pain, arthritis or osteoarthritis [35–37] compared to their counterparts in less deprived neighbourhoods [34], but the understanding of how neighbourhood deprivation affects these conditions remains limited [35]. An Australian study showed that living in a highly disadvantaged area was associated with fewer hip replacements, but that study did not report hip osteoarthritis rate, therefore the results are less comparable [54]. We speculate that the inconsistency might be due to several reasons, including different samples, confounders considered, and modelling techniques used. First, the association between neighbourhood deprivation and LBP/arthritis might be confounded by individual-level factors that were not included or not available in previous studies. For example, Urwin and colleagues [34] did not adjust for age, sex, and individual SES. The differences in back pain prevalence could be associated with these individual-level confounders, leading to the findings of a higher prevalence of back pain amongst people in more deprived areas. Second, the effects of neighbourhood-level SES and individual-level SES were treated as operating at the same level in some of the prior studies [36], which is subject to ecological fallacy arising from confounding of the individual-level relationship due to the heterogeneity of neighbourhood deprivation and individual confounders within the same twin pair. Ignoring the hierarchical structure of the data may underestimate the standard error of the regression coefficient of the aggregate neighbourhood-level factors, leading to an overestimation of the significance of the effect of neighbourhood deprivation on LBP and arthritis. Third, neighbourhood deprivation can be a complex combination of individual-level SES and the socioeconomic context of people’s neighbourhoods. These factors may affect the risk for LBP or arthritis through various complex mechanisms, thus increasing the ‘risk of risks’ or offering protection against LBP or arthritis [55]. Neighbourhood deprivation is often characterised by reduced access to health-related resources, such as preventive health care, health management, fresh food shops and recreation facilities [3]. Neighbourhood deprivation also contributes to individual risk factors and influences lifestyle behaviours, such as physical activity and fresh produce intake. Previous research showed a greater propensity for individuals residing in more advantaged neighbourhoods to be more physically active [56], while neighbourhood walkability has been found to moderate the association between LBP and physical activity [57]. These factors could act as surrogates for neighbourhood deprivation, resulting in a significant association between neighbourhood deprivation and LBP and arthritis as found in previous studies. Indeed, people in neighbourhoods with better access to parks, sidewalks and fresh food stores have been found to have a lower prevalence ratio of knee and low back pain [58] and better management of osteoarthritis, which improves other aspects of health and overall quality of life [29]. Alternatively, if individual-level SES represents the pathway via which neighbourhood deprivation influences the likelihood of experiencing LBP/arthritis, adjusting for individual SES could have attenuated the variation attributable to neighbourhood deprivation [59].
Without a full understanding of the mechanism whereby neighbourhood deprivation affects LBP or arthritis, these explanations remain speculative. Further studies to examine the environmental effects of musculoskeletal conditions are recommended, in particular by using the multilevel modelling approach as in this study. This is because the neighbourhood could influence LBP and arthritis conditions at a different level than individual factors; ignoring this would lead to an ecologic fallacy, which may overestimate the significance of the effects of the aggregate risk factors at the neighbourhood level. To avoid potential overestimation of neighbourhood effects, the MLM approach is highly recommended to consider neighbourhood clustering effects (i.e., individuals nested within neighbourhoods).
Our findings about individual characteristics reinforced the findings from Brennan et al [35], which found arthritis prevalence was greatest for women and older adults 60–65 years. Our results also corroborate the non-significant cross-level interactions between neighbourhood SES and individual SES (education and household income).
Strengths and limitations
Neighbourhood effects on LBP or arthritis are rarely studied, and this is one of the first studies employing a twin cohort design. Our findings on the relationships between neighbourhood deprivation and LBP or arthritis, while non-significant, offer additional evidence to the limited research in this space and add to our knowledge about the potential role of the neighbourhood socioeconomic environment on LBP and arthritis. By applying an innovative modelling method to a large sample of twin data, this study presented a methodology for future research on neighbourhood effects on health outcomes, including twin studies. MLM can handle complex clustering structures: not only twin individuals nested within families but also families within neighbourhoods. It utilizes the unique features of two data and decomposes the phenotypic variance components into a part that is unique to each individual and a part that is common to the twins. Another benefit of MLM is that this modelling technique does not require complete sets of twins, thus preserving a large sample size and increasing statistical power. In addition, by using a relatively large MZ twin sample, this study controlled for common genetics and common environmental factors shared by twins and decomposed the total unexplained variance in the prevalence of LBP and arthritis into the between-pair and within-pair parts, which is not possible to achieve in other cohort or population-based studies.
Limitations of this study also warrant mention. First, the cross-sectional data limits our ability to capture causality in the association between residential environments and LBP or arthritis. Second, neighbourhood deprivation was ascertained at the time of the survey, therefore it could not capture unique environments where people resided prior to the survey. The analyses could be improved if multiple waves of surveys were conducted on the same twins so that any causal relationship and the effects of previous environmental factors can be assessed and addressed through longitudinal data. Additionally, some information such as disease severity, pain locations, intensity and duration, or occupational exposure, was not available for this study, which would have provided further insights for our analysis. These limitations are common in large population-based studies.
Conclusion
In conclusion, our study expands the existing literature on the association between SES measures and LBP/arthritis, which is an under-researched area. By drawing from community-based twin data, we examined important measures of individual SES (education and income) with neighbourhood-level SES to help identify different aspects of SES that could potentially jointly affect LBP and arthritis. Although our empirical study did not find significant associations between deprived neighbourhoods and LBP or arthritis, it contributes to the existing literature on the potential role of neighbourhood factors on LBP and arthritis in twins. More empirical studies using MLMs to examine environmental influences on musculoskeletal conditions would help our understanding of specific neighbourhood environmental factors that may play a role in LBP and arthritis. This information would help to design intervention strategies and prioritise action plans for these conditions.
Supporting information
S1 Table. Results for two response variables: Doctor-diagnosed LBP and arthritis.
https://doi.org/10.1371/journal.pone.0298356.s001
(XLSX)
References
- 1. James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease Study 2017. The Lancet. 2018;392(10159):1789–858. pmid:30496104
- 2. Lee H, Hübscher M, Moseley GL, Kamper SJ, Traeger AC, Mansell G, et al. How does pain lead to disability? A systematic review and meta-analysis of mediation studies in people with back and neck pain. 2015;156(6):988–97.
- 3. Murray CJ, Vos T, Lozano R, Naghavi M, Flaxman AD, Michaud C, et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380(9859):2197–223. pmid:23245608
- 4. Roux AVD, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, et al. Neighborhood of residence and incidence of coronary heart disease. 2001;345(2):99–106.
- 5. Wu A, March L, Zheng X, Huang J, Wang X, Zhao J, et al. Global low back pain prevalence and years lived with disability from 1990 to 2017: Estimates from the Global Burden of Disease Study 2017. Ann Transl Med. 2020;8(6):299. pmid:32355743
- 6. Gore M, Sadosky A, Stacey BR, Tai KS, Leslie D. The burden of chronic low back pain: clinical comorbidities, treatment patterns, and health care costs in usual care settings. Spine. 2012;37(11):E668–77. pmid:22146287
- 7. Hartvigsen J, Hancock MJ, Kongsted A, Louw Q, Ferreira ML, Genevay S, et al. What low back pain is and why we need to pay attention. 2018;391(10137):2356–67.
- 8. Ikeda T, Sugiyama K, Aida J, Tsuboya T, Watabiki N, Kondo K, et al. Socioeconomic inequalities in low back pain among older people: the JAGES cross-sectional study. Int J Equity Health. 2019;18(1):15. pmid:30665404
- 9. Johnson VL, Hunter DJ. The epidemiology of osteoarthritis. Best Pract Res Clin Rheumatol. 2014;28(1):5–15. pmid:24792942
- 10. Lee JY, Han K, Park YG, Park S-H. Socioeconomic Factors and Knee Osteoarthritis: Effects of Education, Income, and Occupation on Prevalence and Symptoms. Research Square; 2021.
- 11. Liechty A, Tsang S, Turkheimer E, Duncan GE. Association between low back pain and body mass index in adult twins: an analysis of monozygotic and dizygotic twins of the Washington State Twin Registry. Spine J. 2020;20(11):1805–15. pmid:32615328
- 12. Silverwood V, Blagojevic-Bucknall M, Jinks C, Jordan JL, Protheroe J, Jordan KP. Current evidence on risk factors for knee osteoarthritis in older adults: a systematic review and meta-analysis. Osteoarthritis Cartilage. 2015;23(4):507–15. pmid:25447976
- 13. Steffens D, Maher CG, Pereira LS, Stevens ML, Oliveira VC, Chapple M, et al. Prevention of Low Back Pain: A Systematic Review and Meta-analysis. JAMA Intern Med. 2016;176(2):199–208. pmid:26752509
- 14. Pinheiro MB, Ferreira ML, Refshauge K, Colodro-Conde L, Carrillo E, Hopper JL, et al. Genetics and the environment affect the relationship between depression and low back pain: a co-twin control study of Spanish twins. Pain. 2015;156(3):496–503. pmid:25679471
- 15. O’Neill TW, McCabe PS, McBeth J. Update on the epidemiology, risk factors and disease outcomes of osteoarthritis. Best Pract Res Clin Rheumatol. 2018;32(2):312–26. pmid:30527434
- 16. Astell-Burt T, Feng X, Kolt GS. Neighbourhood green space and the odds of having skin cancer: multilevel evidence of survey data from 267072 Australians. J Epidemiol Community Health. 2014;68(4):370–4. pmid:24376218
- 17. Biglarian A, Seifi B, Bakhshi E, Mohammad K, Rahgozar M, Karimlou M, et al. Low back pain prevalence and associated factors in Iranian population: findings from the national health survey. Pain Res Treat. 2012;2012:653060. pmid:23024861
- 18. Boring MA, Hootman JM, Liu Y, Theis KA, Murphy LB, Barbour KE, et al. Prevalence of Arthritis and Arthritis-Attributable Activity Limitation by Urban-Rural County Classification—United States, 2015. MMWR Morb Mortal Wkly Rep 2017(66):527–32
- 19. de Vries S, Verheij RA, Groenewegen PP, Spreeuwenberg P. Natural Environments—Healthy Environments? An Exploratory Analysis of the Relationship between Greenspace and Health. Environment and Planning A: Economy and Space. 2003;35(10):1717–31.
- 20. Ghimire R, Ferreira S, Green GT, Poudyal NC, Cordell HK, Thapa JR. Green Space and Adult Obesity in the United States. Ecological Economics. 2017;136:201–12.
- 21. Howell NA, Tu JV, Moineddin R, Chu A, Booth GL. Association Between Neighborhood Walkability and Predicted 10-Year Cardiovascular Disease Risk: The CANHEART (Cardiovascular Health in Ambulatory Care Research Team) Cohort. J Am Heart Assoc. 2019;8(21):e013146. pmid:31665997
- 22. Kamphuis CB, van Lenthe FJ, Giskes K, Brug J, Mackenbach JP. Perceived environmental determinants of physical activity and fruit and vegetable consumption among high and low socioeconomic groups in the Netherlands. Health & Place. 2007;13(2):493–503. pmid:16815073
- 23. Leyden KM. Social capital and the built environment: the importance of walkable neighborhoods. Am J Public Health. 2003;93(9):1546–51. pmid:12948978
- 24. Maas J, Verheij RA, Groenewegen PP, de Vries S, Spreeuwenberg P. Green space, urbanity, and health: how strong is the relation? J Epidemiol Community Health. 2006;60(7):587–92. pmid:16790830
- 25. Martin KR, Shreffler J, Schoster B, Callahan LF. Associations of perceived neighborhood environment on health status outcomes in persons with arthritis. Arthritis Care Res (Hoboken). 2010;62(11):1602–11. pmid:20521309
- 26. Muller G, Harhoff R, Rahe C, Berger K. Inner-city green space and its association with body mass index and prevalent type 2 diabetes: a cross-sectional study in an urban German city. BMJ Open. 2018;8(1):e019062. pmid:29358439
- 27. Pearson AL, Bentham G, Day P, Kingham S. Associations between neighbourhood environmental characteristics and obesity and related behaviours among adult New Zealanders. BMC Public Health. 2014;14(1):553. pmid:24894572
- 28. Rojas-Rueda D, Nieuwenhuijsen MJ, Gascon M, Perez-Leon D, Mudu P. Green spaces and mortality: a systematic review and meta-analysis of cohort studies. Lancet Planet Health. 2019;3(11):e469–e77. pmid:31777338
- 29. Bowden JL, Hunter DJ, Feng Y. How can neighborhood environments facilitate management of osteoarthritis: A scoping review. Semin Arthritis Rheum. 2021;51(1):253–65. pmid:33387921
- 30. Dinescu D, Horn EE, Duncan G, Turkheimer E. Socioeconomic modifiers of genetic and environmental influences on body mass index in adult twins. Health Psychol. 2016;35(2):157–66. pmid:26348497
- 31. Reijneveld SA. Neighbourhood socioeconomic context and self reported health and smoking: a secondary analysis of data on seven cities. J Epidemiol Community Health. 2002;56(12):935–42. pmid:12461115
- 32. Short H, Al Sayah F, Ohinmaa A, Lahtinen M, Johnson JA. The relationship of neighbourhood-level material and social deprivation with health-related quality of life. Qual Life Res. 2018;27(12):3265–74. pmid:30069793
- 33. Strachan E, Duncan G, Horn E, Turkheimer EJPm. Neighborhood deprivation and depression in adult twins: genetics and gene× environment interaction. 2017;47(4):627.
- 34. Urwin M, Symmons D, Allison T, Brammah T, Busby H, Roxby M, et al. Estimating the burden of musculoskeletal disorders in the community: the comparative prevalence of symptoms at different anatomical sites, and the relation to social deprivation. Ann Rheum Dis. 1998;57(11):649–55. pmid:9924205
- 35. Brennan SL, Turrell G. Neighborhood disadvantage, individual-level socioeconomic position, and self-reported chronic arthritis: a cross-sectional multilevel study. Arthritis Care Res (Hoboken). 2012;64(5):721–8. pmid:22232068
- 36. Cleveland RJ, Schwartz TA, Prizer LP, Randolph R, Schoster B, Renner JB, et al. Associations of educational attainment, occupation, and community poverty with hip osteoarthritis. Arthritis Care Res (Hoboken). 2013;65(6):954–61. pmid:23225374
- 37. Reyes C, Garcia-Gil M, Elorza JM, Mendez-Boo L, Hermosilla E, Javaid MK, et al. Socio-economic status and the risk of developing hand, hip or knee osteoarthritis: a region-wide ecological study. Osteoarthritis and Cartilage. 2015;23(8):1323–9. pmid:25819582
- 38. Hunter DJ, March L, Chew M. Osteoarthritis in 2020 and beyond: a Lancet Commission. Lancet. 2020;396(10264):1711–2. pmid:33159851
- 39. Bowden JL, Hunter DJ, Mills K, Allen K, Bennell K, Briggs AM, et al. The OARSI Joint Effort Initiative: Priorities for osteoarthritis management program implementation and research 2024–2028. Osteoarthr Cartil Open. 2023;5(4):100408. pmid:37771392
- 40. Afari N, Noonan C, Goldberg J, Edwards K, Gadepalli K, Osterman B, et al. University of Washington Twin Registry: construction and characteristics of a community-based twin registry. Twin Res Hum Genet. 2006;9(6):1023–9. pmid:17254446
- 41. Duncan GE, Avery AR, Strachan E, Turkheimer E, Tsang S. The Washington State Twin Registry: 2019 Update. Twin Res Hum Genet. 2019;22(6):788–93. pmid:31358074
- 42. Singh GK. Area deprivation and widening inequalities in US mortality, 1969–1998. Am J Public Health. 2003;93(7):1137–43. pmid:12835199
- 43. Roffey DM, Budiansky A, Coyle MJ, Wai EK. Obesity and Low Back Pain: Is There a Weight of Evidence to Support a Positive Relationship? Current Obesity Reports. 2013;2(3):241–50.
- 44. Zheng H, Chen C. Body mass index and risk of knee osteoarthritis: systematic review and meta-analysis of prospective studies. BMJ Open. 2015;5(12):e007568. pmid:26656979
- 45.
World Health Organization. Obesity: preventing and managing the global epidemic. Report on a WHO Consultation (WHO Technical Report Series 894). Geneva, Switzerland: World Health Organization; 2000.
- 46.
Leckie G, Bell A. Cross-Classified Multilevel Models—MLwiN Practical. LEMMA VLE Module 122013. p. 1–60.
- 47.
Falconer DS. Introduction to quantitative genetics: Pearson Education India; 1996.
- 48. Tamimy Z, Kevenaar ST, Hottenga JJ, Hunter MD, de Zeeuw EL, Neale MC, et al. Multilevel Twin Models: Geographical Region as a Third Level Variable. Behav Genet. 2021;51(3):319–30. pmid:33638732
- 49. Eaves LJ, Last KA, Young PA, Martin NG. Model-fitting approaches to the analysis of human behaviour. Heredity (Edinb). 1978;41(3):249–320. pmid:370072
- 50. Austin PC, Merlo J. Intermediate and advanced topics in multilevel logistic regression analysis. Stat Med. 2017;36(20):3257–77. pmid:28543517
- 51. Tate RL, Pituch KA. Multivariate Hierarchical Linear Modeling in Randomized Field Experiments. The Journal of Experimental Education. 2007;75(4):317–37.
- 52. Dessie ZG, Zewotir T, Mwambi H, North D. Multivariate multilevel modeling of quality of life dynamics of HIV infected patients. Health Qual Life Outcomes. 2020;18(1):80. pmid:32209095
- 53.
Rasbash J, Charlton C, Browne WJ, Healy M, Cameron B. MLwiN Version 2.10. Centre for Multilevel Modelling: University of Bristol; 2009.
- 54. Dixon T, Urquhart DM, Berry P, Bhatia K, Wang Y, Graves S, et al. Variation in rates of hip and knee joint replacement in Australia based on socio-economic status, geographical locality, birthplace and indigenous status. ANZ J Surg. 2011;81(1–2):26–31. pmid:21299795
- 55. Paeratakul S, Lovejoy JC, Ryan DH, Bray GA. The relation of gender, race and socioeconomic status to obesity and obesity comorbidities in a sample of US adults. Int J Obes Relat Metab Disord. 2002;26(9):1205–10. pmid:12187397
- 56. Turrell G, Haynes M, Burton NW, Giles-Corti B, Oldenburg B, Wilson LA, et al. Neighborhood disadvantage and physical activity: baseline results from the HABITAT multilevel longitudinal study. Ann Epidemiol. 2010;20(3):171–81. pmid:20159488
- 57. Zadro JR, Shirley D, Pinheiro MB, Bauman A, Duncan GE, Ferreira PH. Neighborhood walkability moderates the association between low back pain and physical activity: A co-twin control study. Prev Med. 2017;99:257–63. pmid:28322877
- 58. Okabe D, Tsuji T, Hanazato M, Miyaguni Y, Asada N, Kondo K. Neighborhood Walkability in Relation to Knee and Low Back Pain in Older People: A Multilevel Cross-Sectional Study from the JAGES. Int J Environ Res Public Health. 2019;16(23). pmid:31756959
- 59. Diez Roux AV. Estimating neighborhood health effects: the challenges of causal inference in a complex world. Soc Sci Med. 2004;58(10):1953–60. pmid:15020010