Altered pain processing and sensitisation is evident in adults with patellofemoral pain: a systematic review including meta-analysis and meta-regression

2.1 Search strategy

PubMed, CINAHL, Web of Science and EMBASE were systematically searched by one reviewer (CB) from inception to April 2019 using a comprehensive and reproducible search strategy outlined in Table 1. Each database was searched using two groups of keyword combinations, relating to PFP and pain processing, to identify all relevant literature. The electronic search was complemented by hand searching the reference lists of retrieved articles and the completion of citation tracking using Google Scholar.

2.2 Inclusion and exclusion criteria

Studies were eligible for inclusion if they investigated pain processing using QST in PFP participants compared to controls. QST measures included pressure pain thresholds, tactile detection thresholds, thermal detection thresholds (warmth and cold detection thresholds), vibration detection thresholds, CPM and TS. PFP participants were required to meet the diagnostic criteria of insidious onset anterior or retropatellar knee pain, exacerbated by activities associated with patellofemoral joint loading including squatting, kneeling, stair ambulation, running and jumping [4]. Mixed-sex and single-sex participant groups of all ages were included. Studies were required to involve only human subjects, full-text cohort, cross-section or case-control studies and published in peer-review journals in the English language.

Studies were excluded if they did not contain a control group or included participants with additional knee pathologies (including internal derangement or ligamentous instability) or previous knee surgery. The exclusion criteria also ruled out unpublished, non-peer reviewed, animal and studies not in the English language, in addition to case studies, reviews, letters, opinion articles, conference proceedings and thesis papers.

2.3 Review process

Studies identified through the search strategy were downloaded into Endnote X7.5 (Thomson Reuters Cooperation, New York, NY, USA), with duplicates subsequently deleted. Titles and abstracts were screened for eligibility with the full-texts of potentially relevant articles obtained for further review by one reviewer (CB). Full-texts were screened where eligibility could not be determined by the abstract alone, with any uncertainties resolved at a consensus meeting with a second reviewer (SL).

2.4 Quality assessment

Study methodological quality was evaluated using a modified version of the Downs and Black checklist [34] and a PFP diagnostic checklist [35]. Downs and Black [34] is a validated tool that is widely used in literature, with good reliability and validity reported [36], [37]. The modified version contained 16 criteria following the removal of criteria 4, 8, 9, 13, 14, 17, 19, 23, 24, 26 and 27 which have previously been deemed inappropriate for non-randomised studies [38]. As described by Hootman et al. [39], studies were scored out of a maximum of 17 points, with the score converted into a percentage to allow banding into low (≤33.3%), moderate (33.4–66.7%) or high (≥66.8%) quality. Studies scoring below 50% were excluded from subsequent analysis to prevent the inclusion of studies with a high risk of bias [35]. The PFP diagnostic checklist is a seven-item scale developed by Barton et al. [35] that identifies key inclusion and exclusion criteria for the diagnosis of PFP. A higher score indicates a higher number of criteria having been reported thus a more comprehensive diagnosis made. Both quality assessment tools were completed by two independent reviewers (CB and BN) with any discrepancies resolved at a consensus meeting. A third reviewer (SL) was available but not required.

2.5 Data extraction

Study details (primary author, year, study design), participant demographics (sample size, sex, age, height, weight, BMI), symptom duration and QST measures of pain processing (PPTs, tactile detection thresholds, vibration detection thresholds, thermal detection thresholds, CPM, TS), both locally and distally, were extracted and analysed from each study. Means and standard deviations were sourced from the papers or through contacting the authors via email. One study [16] presented data as median and interquartile range which were converted into mean and standard deviation following guidance from the Cochrane Handbook [40].

2.6 Data analysis

Data analysis was performed using the review manager software package RevMan 5.3.5 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Meta-analysis was completed where a minimum of two studies recorded comparable QST measures in PFP cases and controls. Data which could not be pooled was summarised in a narrative format and presented within tables. Means and standard deviations were inputted into RevMan to allow calculation of standardized mean difference (SMD) and 95% confidence interval (95% CI). SMDs were calculated for both pooled and unpooled data as all studies presented data as continuous scale.

A random effects model was used for meta-analysis. Calculated SMDs were categorised as small (≤0.59), medium (0.60–1.19) or large (≥1.20) effect sizes, as recommended by Hume et al. [41]. The level of statistical heterogenicity for pooled data was established using χ² and I² statistics with the level of significance set at p<0.05. PPT locations were categorised into local and distal sites, with the following definitions constructed to standardize analysis. Local was defined as either the most painful site reported at the patella or the site closest to the central patella, since this is thought to be the most prevalent site of pain in PFP [23]. Distal was defined as the site most anatomically distant from the primary area of pain [28].

Levels of evidence were established for each finding with the definitions based on the van Tulder Criteria [42].

1. Strong: results derived from multiple studies, including at least two HQ studies, which are statistically homogenous (I²<50%).

2. Moderate: results derived from multiple studies, including at least one HQ study, which are statistically heterogeneous (I²>50%); or from multiple LQ studies which are statistically homogenous (I²<50%).

3. Limited: results derived from multiple LQ studies which are statistically heterogeneous (I²>50%); or from one HQ study.

4. Very limited: results derived from one LQ study.

5. Conflicting: insignificant results derived from multiple statistically heterogeneous studies (I²>50%).

Initial planned analysis was to consider the relationship between PPTs before and after treatment, however, due to a paucity of studies presenting data for both time points this analysis was not possible. Instead, a random effects meta-regression was performed to explore the relationship between the SMDs of individual studies PPTs and both age and symptom duration. Meta-regression was completed using the comprehensive analysis and graphic software programme Prism 8.1.0 (GraphPad, San Diego, CA, USA). An α level was set a priori at <0.05. Calculated correlation coefficients (R²) were categorised as negligible (0.00–0.10), weak (0.10–0.39), moderate (0.4–0.69), strong (0.7–0.89) and very strong (0.90–1.00) correlations, as outlined by Schober et al. [43].

3.1 Search strategy

The results of the database searches are displayed in Fig. 1. The search identified 3,286 relevant papers and following the deletion of duplicates, 2,689 papers remained. After subsequent title and abstract screening, 21 papers remained and full-text assessment for eligibility identified 11 studies meeting the inclusion criteria [15], [16], [17], [18], [19], [20], [21], [22], [23], [32], [44]. The most common reasons for exclusion were a lack of specificity to PFP or the absence of a control group.

Fig. 1:

PRISMA flow diagram.

3.2 Study characteristics

Study characteristics are outlined in Table 2. All 11 included studies investigated pain processing in PFP patients compared to controls using QST measures. Three studies were case-control [18], [23], [32], six were cross-sectional [15], [19], [20], [21], [22], [44] and two were cohort [16], [17].

3.3 Quality assessment

The results of the Downs and Black checklist [34] and PFP diagnostic checklist [35] are presented in Tables 3 and 4. The Downs and Black scores ranged from 64.7% [16], [18] to 88.2% [20], [21], [22], [23], with nine studies rated high quality (HQ) [15], [17], [19], [20], [21], [22], [23], 32], [44] and two studies rated medium quality (MQ) [16], [18]. All studies exceeded the 50% threshold for inclusion. The most common reasons for methodological compromise were lack of detail concerning the population characteristics, lack of assessor blinding and inadequate adjustment for confounders. The PFP diagnostic checklist scores ranged from 3/7 points [17] to 7/7 points [21], [22], [32], [44]. Commonly, studies did not provide a clear definition of the location of pain or state that pain should be of insidious onset, unrelated to trauma.

3.4 Quantitative sensory testing measures

Data pooling for meta-analysis was possible for PPTs, tactile detection thresholds and thermal detection thresholds with forest plots displayed in Figs. 2–4, respectively. Results that could not be pooled are summarised in a narrative format in Table 5.

Fig. 2:

Meta-analysis of local and distal pressure pain thresholds.

Fig. 3:

Meta-analysis of local and distal tactile detection thresholds.

Fig. 4:

Meta-analysis of warmth and cold detection thresholds.

3.5 Relationship between pressure pain thresholds and sex

Three studies (3 HQ [23], [31], [44]) included mixed sex participants measuring local and distal PPTs. Results from these studies were pooled for PFP patients vs. controls (Fig. 5) and female vs. male PFP patients (Fig. 6).

Fig. 5:

Meta-analysis pressure pain thresholds in mixed sex studies.

Fig. 6:

Meta-analysis pressure pain thresholds in female cases compared to male cases in mixed sex studies.

Moderate evidence (3 HQ [23], [31], [44]) indicates no significant difference in PPTs locally (n=222; SMD −0.14, 95% CI −0.92 to 0.64; p=0.72; I²=86%, p=0.0007) or distally (n=222; SMD −0.12, 95% CI −0.51 to 0.27; p=0.56; I²=48%, p=0.15) in the mixed sex PFP patients compared to controls. The overall effect also indicates no significant difference (n=222; SMD 0.13, 95% CI −0.52 to 0.26; p=0.51; I²=74%, p=0.002).

Moderate evidence (3 HQ [23], [31], [44]) indicates no significant difference in PPTs locally (n=91; SMD −0.85, 95% CI −1.72 to 0.01; p=0.05; I²=62%, p=0.07) or distally (n=91; SMD −0.59, 95% CI −1.62 to 0.43; p=0.26; I²=74%, p=0.02) in the female compared to male PFP patients. However, considering the overall effect, moderate evidence of medium effect indicates decreased PPTs in the female patients (n=91; SMD −0.75, 95% CI −1.34 to −0.16; p=0.01; I²=63%, p=0.02).

3.6 Relationship between pressure pain thresholds and age

All nine studies (8 HQ [15], [19], [20], [21], [22], [23], 31], [44] and 1 MQ [18]) assessing local and distal PPTs provided mean ages of patients. Meta-regression was possible to evaluate the relationship between PPTs and age. Findings indicate a moderate correlation between decreasing local PTTs and decreasing patient age (R²=0.556, β=−3.90, 95% CI −6.40 to −1.40, p=0.0211) (Fig. 7). Similarly, findings indicate a moderate correlation between decreasing distal PTTs and decreasing patient age (R²=0.491, β=−2.63, 95% CI −4.54 to −0.728, p=0.0354) (Fig. 8).

Fig. 7:

Meta-regression scatterplot of local pressure pain thresholds and age.

Fig. 8:

Meta-regression scatterplot of distal pressure pain thresholds and age.

3.7 Relationship between pressure pain thresholds and symptom duration

All nine studies (8 HQ [15], [19], [20], [21], [22], [23], 31], [44] and 1 MQ [18]) assessing local and distal PPTs provided mean symptom durations of patients. Meta-regression analysis was possible to evaluate the relationship between PPTs and symptom duration. No significant association was identified between local pressure pain sensitivity and symptom duration (R²=0.170, β=−0.323, 95% CI −1.41 to 0.759, p=0.270) (Fig. 9). No significant association was identified between distal pressure pain sensitivity and symptom duration (R²=0.092, β=−0.318, 95% CI −1.12 to 0.484, p=0.428) (Fig. 10).

Fig. 9:

Meta-regression scatterplot of local pressure pain thresholds and symptom duration.

Fig. 10:

Meta-regression scatterplot of distal pressure pain thresholds and symptom duration.

4.1 Quantitative sensory testing measures

4.2 Relationship between pressure pain thresholds and sex

With the prevalence of PFP twice as high in females compared to males [60], past research has often focused on females. Out of the nine studies assessing PPTs, just three studies recruited both female and male participants [23], [32], [44], with the remaining six studies recruiting female participants only [15], [18], [19], [20], [21], [22]. Interestingly, the mixed sex studies reported the lowest effect sizes, with two studies finding no significant difference in local or distal pressure pain thresholds in the PFP cases compared to controls [32], [44]. When data from these studies were pooled, this was confirmed, with the overall effect indicating no significant difference between the two groups. Extraction of PPT scores for female and male patients separately indicates an overall effect of reduced pressure pain thresholds in the female patients, hence increased pain sensitivity. Females with knee osteoarthritis are well documented to experience greater pain sensitivity to painful stimuli such as pressure pain thresholds when compared to male populations [61], [62], [63], with evidence suggesting less efficient endogenous pain mechanisms in females [64]. Certainly, further research to explore this potential link between increased pain sensitivity and sex is required to confirm findings and provide a more robust evidence base.

4.3 Relationship between pressure pain thresholds and age

A significant association was identified between decreasing PPTs and decreasing patient age, suggesting younger patients are more likely to experience pressure pain sensitivity. Evidence suggests that childhood and adolescence are critical periods where pain experience can prime nociceptors inducing long-lasting effects not seen amongst adults [65]. Consequently, pain sensitivity may be age dependent and amplified in a younger population [21], [22]. These findings further highlight the importance of effectively managing this common pain complaint within younger populations possibly to include components aimed at therapeutic neuroscience education [19] with involvement of the full multidisciplinary team to guide appropriate pain management interventions [66].

4.4 Relationship between pressure pain thresholds and symptom duration

No significant association was identified between pressure pain sensitivity and symptom duration. This remains in common with literature which reports symptom duration to be both positively and negatively associated with manifestations of hyperalgesia [21], [57]. Furthermore, the time taken to develop hyperalgesia varies greatly between conditions, for example, 1 month in whiplash [67] vs. 5 years in rheumatoid arthritis, but this time remains unclear in PFP.

4.5 Limitations and future research directions

This review is not without limitations, which must be considered when interpreting the results. Due to methodological heterogeneity and limited number of studies assessing pain processing in PFP, meta-analysis was not possible for all sensitisation measures. I² values were consistently high indicating significant heterogeneity between studies, however, a random effects model was used to help account for this. Due to a lack of available evidence, meta-analysis of tactile and thermal detection thresholds involved only several studies resulting in a high risk of bias and reduced methodological quality. The Downs and Black methodological quality assessment identified nine HQ studies [15], [17], [19], [20], [21], [22], [23], 32], [44] and two MQ studies [16], [18]. One common reason for methodological compromise was a lack of detail concerning the population characteristics which may introduce the risk of population bias. Furthermore, only four studies blinded the assessor during group assignment increasing the risk of detection bias [15], [20], [22], [44]. But despite these suggestions of bias, no study scored less than 50% during quality assessment requiring exclusion from subsequent analysis. However, there is no formally accepted quality assessment tool recommended for systematic reviews. Whilst Down and Black is a validated and widely used checklist, it is possible that using alternative tools may have generated different levels of evidence. PFP diagnostic checklist scores varied, ranging from 3/7 [23] to 7/7 points [16], [17], [19], [20], indicating large heterogeneity in the definition of PFP. Ensuring that the diagnostic approach is more uniform in future studies would be advisable to avoid the possibility of confounding pathologies. Furthermore, with sensitisation identified as a contributing factor, further research to consider whether manifestations of sensitisation are associated with treatment or prognosis may prove insightful to guide management approaches.

4.6 Conclusion

This systematic review presents moderate evidence to suggest the presence of altered pain processing and sensitisation in patients with PFP. Results indicate decreased PPTs, increased tactile detection thresholds and increased WDTs in PFP patients compared to controls. Furthermore, PPTs were found to be reduced in female patients with a correlation between decreasing PPTs and decreasing patient age. Although the exact aetiology of PFP remains elusive it is plausible that a combination of biomechanical and neuropathic dysfunction is responsible, possibly with the neuropathic elements contributing to the chronicity of PFP. With female patients and younger patients identified as being more likely to experience increased pressure pain sensitivity, this population may benefit most from a more targeted management approach incorporating components aimed specifically at neuropathic pain, including neuroscience education.