The strength of association between psychological factors and clinical outcome in tendinopathy: A systematic review

Objective Tendinopathy is often a disabling, and persistent musculoskeletal disorder. Psychological factors appear to play a role in the perpetuation of symptoms and influence recovery in musculoskeletal pain. To date, the impact of psychological factors on clinical outcome in tendinopathy remains unclear. Therefore, the purpose of this systematic review was to investigate the strength of association between psychological factors and clinical outcome in tendinopathy. Methods A systematic review of the literature and qualitative synthesis of published trials was conducted. Electronic searches of ovid MEDLINE, ovid EMBASE, PsychINFO, CINAHL and Cochrane Library was undertaken from their inception to June 2020. Eligibility criteria included RCT’s and studies of observational design incorporating measurements of psychological factors and pain, disability and physical functional outcomes in people with tendinopathy. Risk of Bias was assessed by two authors using a modified version of the Newcastle Ottawa Scale. High or low certainty evidence was examined using the GRADE criteria. Results Ten studies of observational design (6-cross sectional and 4 prospective studies), involving a sample of 719 participants with tendinopathy were included. Risk of bias for the included studies ranged from 12/21 to 21/21. Cross-sectional studies of low to very low level of certainty evidence revealed significant weak to moderate strength of association (r = 0.24 to 0.53) between psychological factors and clinical outcomes. Prospective baseline data of very low certainty evidence showed weak strength of association between psychological factors and clinical outcome. However, prospective studies were inconsistent in showing a predictive relationship between baseline psychological factors on long-term outcome. Cross sectional studies report similar strengths of association between psychological factors and clinical outcomes in tendinopathy to those found in other musculoskeletal conditions. Conclusion The overall body of the evidence after applying the GRADE criteria was low to very low certainty evidence, due to risk of bias, imprecision and indirectness found across included studies. Future, high quality longitudinal cohort studies are required to investigate the predictive value of baseline psychological factors on long-term clinical outcome.

Yes -all data are fully available without restriction Objective Tendinopathy is often a disabling, and persistent musculoskeletal disorder. 22 Psychological factors appear to play a role in the perpetuation of symptoms and influence 23 recovery in musculoskeletal pain. To date, the impact of psychological factors on clinical 24 outcome in tendinopathy remains unclear. Therefore, the purpose of this systematic review 25 was to investigate the strength of association between psychological and clinical outcome in 26 tendinopathy.   Conclusion Cross sectional studies report similar strengths of association between 43 psychological factors and clinical outcomes in tendinopathy to those found in other 44 musculoskeletal conditions. However, the overall body of the evidence after applying the 45 GRADE criteria was very low certainty evidence, due to risk of bias, imprecision and 46 Introduction 51 Tendinopathy, previously referred to as tendinitis or tendinosis, is a common musculoskeletal 52 condition characterised clinically by pain reported around the affected tendon with loading 53 [1]. Tendinopathy affects both athletic and non-athlectic populations alike. For example, 54 Achilles tendinopathy is reported in up to 50% of runners before the age of 45 years [2]. In 55 general practice, Albers et al. [3] reported lower extremity tendinopathy prevalence rates of 56 11.8 per 1000 person-years, whilst prevalence rates for upper limb tendinopathies have been 57 estimated between 1.3% to 21.0% [4,5,6]. 58 Recommended care for tendinopathy includes exercise interventions involving progressive 59 exercise interventions such as concentric and/or eccentric strengthening [7,8]. Success rates 60 reported using such exercise programmes in Achilles tendinopathy have been shown to vary 61 between 56% and 100%. [9,10] whilst moderate response rates of 41% have been reported 62 for eccentric exercise in people with lateral elbow tendinopathy [11]. Given the often 63 persistent and multifactorial nature of tendinopathy, similar to to other MSK disorders, 64 treatments in tendinopathy may also need to address the multiple factors that contribute to 65 pain, dysfunction and disability experienced.

66
It has been suggested that psychological factors such as fear of reinjury, pain catastrophising, 67 external locus of control and low self efficacy may negatively impact on clinical outcomes in 68 common musculoskeletal disorders [12][13][14][15]. However, to date, the contribution of 69 psychological factors to the pain, dysfunction and disability experienced in tendinopathy and 70 the benefits of focussing upon theses factors as treatment targets remain uncertain. Recent 71 qualitative studies have outlined the negative psychological impact of persistent Achilles 72 tendinopathy, rotator cuff tendinopathy and greater trochanteric pain syndrome [16][17][18][19]. 73 Likewise, a recent cross-sectional study reported greater levels of psychological distress and 74 poorer quality of life among patients with more severe gluteal tendinopathy [20].

75
Previous systematic reviews [21,22] concluded that there was a conflicting evidence-base for     keywords were used (see Table 1). The complete search strategy for each database is shown   The following data were extracted from each study where available (see Table 2  using the Newcastle-Ottawa Scale (NOS) see Table 3. The NOS is a review tool for 179 evaluating risk of bias in non-randomised studies [26,27]. An adapted version of the scale 180 was selected to evaluate all studies because many were single cohort cross-sectional 181 observational studies [28,29]. The tool consists of four domains of risk of bias assessment;

264
Outcome measures 265 The outcome measures utilised across the 10 included studies are detailed in Table 2.

270
The most common cognitive factor measured was kinesiophobia (4 studies, 3 outcome 271 measures), see Table 2. The overall risk of bias for 9/10 studies was low whilst one study was considered to be at 280 moderate risk of bias. Scores ranged from 12/21 to 21/21 (see Table 3). The most common 281 sources of bias were; not adjusting for confounders, (81.8%), inadequate sample size and 282 statistical power (54.5%) and a lack of objective outcome assessment (45.6%).

283
Overall certainty of evidence 284 Due to heterogeneity regarding tendon sites, psychological variables and outcome measures 285 investigated, we were unable to apply the GRADE criteria to measure the certainty of 286 evidence for each individual patient outcome. Consequently, the GRADE criteria was used to 287 establish the certainty of association as it was intended and this was also equivalent to each 288 individual study. Overall, the GRADE criteria demonstrated very low levels of certainty (see 289   Table 4). Limitations mostly related to risk of bias (80%), imprecision (40%) and indirectness 290 (10%). Observational studies included in our review had a greater potential for risk of bias 291 due to a lack of randomisation which increases the possibility of confounding, and selection 292 bias [46]. When considering imprecision, we rated down the evidence quality in four of our 293 studies due to the lack of reporting of 95% confidence intervals.

Associations between psychological factors and clinical outcomes 295
The specific associations investigated for each study are detailed in Table 2. Some

296
representative examples are provided for each tendinopathy/tendon region below.

297
Rotator cuff tendinopathy 298 There was very low certainty evidence from one study of prospective design [38] supporting 299 a weak positive baseline association between catastrophizing and pain (r = 0.32, p<0.01) and 300 disability (r = 0.37, p<0.01). Fear avoidance beliefs measured at baseline appeared to be 301 significantly associated with baseline disability (r = 0.24, p<0.05) but not significant with 302 disability change scores after 3-months.

303
There was very low certainty evidence from three cross-sectional studies supporting a weak 304 positive association between baseline psychological factors (including depression, anxiety 305 and emotional distress) and pain [36,39], disability [36,39] and a negative association with 306 physical function [37]. Thirty four percent (10/29) of the associations investigated were not 307 significant, (refer to Table 2) with some associations differing between clinical outcomes. For 308 example, Kromer et al. [38] showed fear avoidance behaviour was weakly associated with 309 baseline disability (r = 0.237 (95% CI=0.03 to 0.42) p<0.05) but not pain intensity.

310
Furthermore, preliminary cross-sectional associations found between psychological factors 311 and disability were not always evident when confounding variables were considered using 312 multivariate regression models [36] or when measured at long-term follow up [38].  [42]. Two cross sectional studies demonstrated a weak to moderate association between 318 psychological factors (kinesiophobia, catastrophizing, depression and stress) and foot 319 function and pain [40,41]. Forty seven percent (16/34) of associations investigated were not 320 significant (refer to Table 2). Once again, the associations between clinical and psychological 321 factors varied between studies. For example, cross sectional analysis revealed anxiety was not 322 significantly associated with foot pain [40], whereas, a recent prospective cohort study 323 showed anxiety to be the strongest predictor of pain intensity in people with plantar heel pain 324 (β= -0.41, p=0.01) [42].

325
Lateral elbow tendinopathy 326 Very low certainty evidence from one cohort prognostic study suggested psychological 327 factors (depression and kinesiophobia) were not significantly associated with either pain or 328 disability in patients with lateral elbow tendinopathy at 2 or 12-months [43]. relationships [47].

382
It is also noteworthy that many people in the included trials had low baseline scores for the 383 psychological constructs assessed. People with low baseline psychological scores may be less 384 likely to display a relationship between their psychological status and clinical outcomes [52].

385
For example, baseline mean values for kinesiophobia (TSK) were reported as 23.7 points by 386 Coombes et al. [43]; scores between 17-37 points are considered moderate risk of poor 387 outcome [53,54]. In the same study, participants recorded mean scores of 7.6 out of a total of 388 42 on the Hospital Anxiety and Depression Score (HADS), on which a score of 7 points (or 389 lower) is considered not anxious or depressed [55]. Similarly, median baseline 390 catastrophising was reported as 9 points by Kromer et al. [38], whereas scores >30 points are 391 considered clinically relevant [56]. As higher levels of psychological factors at baseline may 392 predict poorer clinical outcomes [54,57], it appears this population may simply be 393 underrepresented in the included studies; perhaps due to sampling bias or strict 394 inclusion/exclusion criteria which often excludes participants with high levels of pain or 395 bilateral pain symptoms [36,38,39].

396
Similarly to a prior review [21], we found substantial heterogeneity in relation to the range of 397 psychological outcome measures across the included studies included in our review. For 398 instance, depression and anxiety was measured in five studies, in which four different 399 outcome measures were used. This lack of concensus restricts the synthesis and pooling of 400 data for meta-analysis.

439
Identifying individuals who may be at low, moderate or high risk of poor outcome would 440 allow clinicians to adapt their management strategies accordingly.