Article Text

Download PDFPDF

Factors that predict a poor outcome 5–8 years after the diagnosis of patellofemoral pain: a multicentre observational analysis
  1. N E Lankhorst1,
  2. M van Middelkoop1,
  3. K M Crossley2,
  4. S M A Bierma-Zeinstra1,
  5. E H G Oei3,
  6. B Vicenzino2,
  7. N J Collins4
  1. 1Department of General Practice, Erasmus MC, University Medical Centre, Rotterdam, The Netherlands
  2. 2School of Health and Rehabilitation Sciences: Physiotherapy, The University of Queensland, Brisbane, Queensland, Australia
  3. 3Department of Radiology, Erasmus MC, University Medical Centre, Rotterdam, The Netherlands
  4. 4Department of Mechanical Engineering, The University of Melbourne, Melbourne, Victoria, Australia
  1. Correspondence to NE Lankhorst, MD, Department of General Practice, Erasmus MC, University Medical Center, P.O. Box 2040, Rotterdam 3000 CA, The Netherlands; n.lankhorst{at}


Background Patellofemoral pain (PFP) has traditionally been viewed as self-limiting, but recent studies show that a large proportion of patients report chronic knee pain at long-term follow-up. We identified those patients with an unfavourable recovery (‘moderate improvement’ to ‘worse than ever’ measured on a Likert scale) and examined whether there is an association between PFP and osteoarthritis (OA) at 5–8-year follow-up.

Methods Long-term follow-up data were derived from 2 randomised controlled trials (n=179, n=131). Patient-reported measures were obtained at baseline. Pain severity (100 mm visual analogue scale (VAS)), function (Anterior Knee Pain Scale (AKPS)) and self-reported recovery were measured 5–8 years later, along with knee radiographs. Multivariate backward stepwise linear regression analyses were used to evaluate the prognostic ability of baseline pain duration, pain VAS and AKPS on outcomes of pain VAS and AKPS at 5–8 years.

Results 60 (19.3%) participants completed the questionnaires at 5–8-year follow-up (45 women, mean age at baseline 26 years) and 50 underwent knee radiographs. No differences were observed between responders and non-responders regarding baseline demographics, and 3-month and 12-month pain severity and recovery. 34 (57%) reported unfavourable recovery at 5–8 years. 48 out of 50 participants (98%) had no signs of radiographic knee OA. Multivariate models revealed that baseline PFP duration (>12 months; R2=0.22) and lower AKPS (R2=0.196) were significant predictors of poor prognosis at 5–8 years on measures of worst pain VAS and AKPS, respectively.

Summary and conclusion More than half of participants with PFP reported an unfavourable recovery 5–8 years after recruitment, but did not have radiographic knee OA. Longer PFP duration and worse AKPS score at baseline predict poor PFP prognosis. Education of health practitioners and the general public will provide patients with more realistic expectations regarding prognosis.

  • Knee
  • Radiograph

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Patellofemoral pain (PFP) that is aggravated by activities such as squatting, stair walking and running is a common condition affecting a large proportion of adolescents and young adults.1–4 Although PFP has traditionally been viewed as self-limiting,5 the proportion of those reporting chronic knee problems varies widely, from 20% after 1-year follow-up, to 91% after 18 years.2 ,6–11

It is important for clinicians to gain a better insight into the natural course of PFP, so that they can identify patients at risk of chronicity, and better inform patients regarding treatment and prognosis. Previous studies have identified baseline factors that are associated with poor PFP prognosis, including longer duration of symptoms, greater pain severity, lower self-reported function, greater height, positive patella apprehension test and crepitation during physical examination.7 ,12–15 However, most of these studies were short term, varying from 3-month to 12-month follow-up. Only one study had a 7-year follow-up,13 but a substantial proportion of the cohort had undergone surgery (22%), and 35% underwent pre-enrolment diagnostic arthroscopy. Surgery may have altered the natural history of PFP and knee osteoarthritis (OA), and does not represent current practice for PFP.

It has been proposed that PFP may be a precursor to patellofemoral osteoarthritis (PFOA).16–19 PFP and PFOA share common characteristics in terms of symptoms and biomechanics, such as lower limb malalignment (patella and knee), hamstring tightness and reduced quadriceps strength.20–23 Furthermore, a history of PFP symptoms and the presence of knee crepitus are associated with MRI features of PFOA.24 However, longitudinal evidence for a temporal relationship between PFP and PFOA is lacking, particularly with respect to high-quality cohort studies of adequate sample size.17

The aim of this study was to conduct a long-term, 5–8-year follow-up of two randomised clinical trials (RCTs), in order to: (1) describe the proportion of people with PFP who report unfavourable recovery; (2) identify whether people with a history of PFP have radiological signs of PFOA; and (3) determine prognostic indicators of poor outcome on self-reported measures of pain, symptoms and function.


Long-term follow-up data were derived from two RCTs performed in Australia (n=179) and the Netherlands (n=131).6 ,8 Both RCTs investigated the effectiveness of physical therapies for PFP, with detailed methodologies published previously.25 ,26 Ethical approval was obtained prior to start of each study (The University of Queensland's Medical Research Ethics Committee;6 The Erasmus Medical University, Rotterdam8). All participants who volunteered for the 5–8-year follow-up provided written informed consent additional to that obtained at baseline.


All participants from the original RCTs were invited by letter (the Netherlands) or email and letter (Australia) to participate in the long-term follow-up study, 5–8 years after baseline testing. Participants were originally recruited into the RCTs via primary care referral (general practitioners (GPs) and sport physicians)8 and self-referral (local advertising).6 Inclusion criteria for both RCTs were insidious onset peripatellar or retropatellar knee pain present for more than 6 weeks, provoked by at least three of the following activities that load the patellofemoral joint (PFJ): stair ambulation, squatting, running, cycling and prolonged sitting with knees flexed. The Dutch RCT included participants from 14 years of age, while the minimum age for the Australian RCT was 18 years. Exclusion criteria for both studies were: (1) age >40 years; (2) other defined knee pathology (eg, OA, patellar tendinopathy, Osgood-Schlatter's disease); (3) previous knee surgery and (4) physiotherapy intervention within the preceding year. The Australian RCT also excluded volunteers if they rated their worst pain severity in the previous week to be less than 30 mm on a 100 mm visual analogue scale (VAS); had used foot orthoses in the previous year; had foot conditions precluding foot orthoses use; or had concomitant pain in the hip or lumbar spine. Both RCTs randomly allocated participants via concealed allocation. Participants in the Australian RCT were assigned to one of four groups: prefabricated foot orthoses (n=46), flat shoe inserts (n=44), physiotherapy (n=45) and foot orthoses with physiotherapy (n=44).6 The Dutch RCT allocated participants to supervised exercise therapy (n=65) or usual care (PFP education typically provided by GPs and sports physicians in the Netherlands; n=66).8

Outcome measures

Baseline, 3-month and 12-month outcomes were collected via paper format for both studies. The same outcome measures were collected at 5–8 years after randomisation across both cohorts, with the Australian cohort completing paper versions, and the Dutch cohort completing online versions.27

Recovery—primary outcome

Perceived recovery was assessed at 5–8 years for both cohorts. The Australian trial used a five-point Likert scale (‘marked improvement’, to ‘marked worsening’), while the Dutch trial used a seven-point Likert scale (‘fully recovered’ to ‘worse than ever’).

Five-year to eight-year patient reported outcomes

Participants rated their usual knee pain severity (pain on average, or pain at rest) and worst knee pain severity (worse pain or pain during activity). This was completed using 100 mm VAS (Australian cohort) or 11-point numerical rating scales (Dutch cohort). The Anterior Knee Pain Scale (AKPS) was used to measure anterior knee pain symptoms and function, ranging from 0 (maximal disability) to 100 (no disability).28 ,29 The Knee Injury and Osteoarthritis Outcome Score (KOOS) was only measured at long-term follow-up.30 This includes five subscales, with a normalised score for each subscale calculated separately, from 0 (extreme knee problems) to 100 (no knee problems).30 Functional limitations were measured with the Functional Index Questionnaire (FIQ), which has eight items relating to activities that are commonly aggravating for PFP.31 An overall score from 0 (maximal disability) to 16 (no disability) was calculated.28 ,31

Radiographic features

At 5–8 years follow-up, weight-bearing anterior-posterior (AP) and lateral radiographs, as well as skyline view radiographs (Hughston view, with knee in 45° flexion) were taken of the study knee (nominated as most painful knee at baseline). AP radiographs were analysed for tibiofemoral joint space narrowing (none (0), doubtful (1), mild (2) or moderate (3)), medial and lateral tibial and femoral osteophytes (none to moderate (1–3)), tibial attrition (present (0) or absent (1)), tibial and femoral sclerosis, and tibial spiking (present (0) or absent (1)). All features on AP radiographs were scored according to Altman and Gold,32 with the exception of tibial spiking.33 Kellgren and Lawrence (K&L) criteria were used to score tibiofemoral joint OA.34 For lateral views, only osteophytes were scored (none to moderate (1–3)).33 On the skyline views, osteophytes, PFJ space narrowing and patellofemoral sclerosis (none (1) to moderate (3)) were scored.33 All radiographs were scored by one trained medical student, who was unaware of the context of the study and has established reliability (interobserver with trained GP reader prevalence bias adjusted κ score: 0.61–0.75).35 ,36

Predictor variables

Baseline variables were evaluated for their prognostic ability for primary outcomes of VAS pain (worst or activity-related) and AKPS score at 5–8 years. All prognostic indicators that were identified in the analyses for the 3 and 12 months follow-up7 were investigated in the long-term follow-up. These were duration of pain (categorised as 1–2, 2–6, 6–12 and ≥12 months), recruitment method (healthcare professional, self-referral), baseline pain VAS (usual/resting, worst/activity-related) and baseline AKPS score.

Statistical analyses

To establish whether there was selective loss to follow-up, characteristics of long-term participants (n=60) and non-responders (n=250), measured at baseline, 3 and 12 months, were compared using t tests for continuous data and Mann-Whitney U tests for non-normally distributed variables (if n<30; p<0.05). For normally and non-normally distributed categorical data, Pearson χ² tests or Fisher's exact test were used, respectively. Descriptive statistics were applied to describe long-term outcomes.

Scores for recovery were dichotomised into favourable recovery (‘completely recovered’, ‘strongly recovered’ or ‘marked improvement’) and unfavourable recovery (‘moderate improvement’ to ‘worse than ever’). The latter was the primary outcome for this study.

Potential predictor variables were entered into multivariate backward stepwise linear regression analyses (p(in) 0.05, p(out) 0.10). In case of selective follow-up (p<0.1 in both cohorts), those variables were also included in the multivariate analyses. Because more women than men participated in the follow-up study, multivariate analyses also included sex. All analyses were performed based on complete case analysis, using the Statistical Package for the Social Sciences V.21.0 (SPSS Inc, Chicago, Illinois, USA). Significance was set at 0.05.


Of the initial 310 participants from the two RCTs, 60 participants (20 (11.1%) from the Australian RCT, and 40 (30.5%) from the Dutch RCT) completed the patient-reported outcomes at 5–8 years. Of these 60 participants, radiographs were obtained from 50 (83%) participants. Of the 10 participants who did not undergo radiographs, 4 did not provide consent for radiographs, 3 were pregnant, 1 had relocated internationally, 1 sustained an acute knee injury after questionnaire completion but prior to undergoing X-rays, and 1 was unable to attend an imaging clinic. Time since baseline ranged from 5 years and 9 months to 8 years and 6 months. Baseline characteristics and 3-month and 12-month outcomes for long-term follow-up participants and non-responders are presented in table 1. The long-term cohort (n=60) contained significantly more women (Pearson χ² test 7.549, df 1, p=0.006). Australian participants who completed the long-term follow-up had a significantly higher baseline FIQ score compared with non-responders (mean difference (95% CI) 0.403 (−1.72 to −0.70)), but this may not be clinically relevant,28 and there were no significant differences at 3 and 12 months.

Table 1

Participant characteristics at baseline, 3 and 12 months (split by responders (participated in follow-up) and non-responders)


On the dichotomised measure of global recovery, 26 out of 60 participants (43.3%) reported a favourable outcome. Thirty-four out of the 60 (56.7%) reported an unfavourable outcome (figure 1). Of the 27 participants who had an unfavourable outcome at 1-year follow-up, 74.1% still reported an unfavourable outcome after 5–8 years of follow-up. Seven out of the 27 recovered participants after 1 year reported an unfavourable outcome after 5–8 years.

Figure 1

Distribution of favourable and unfavourable outcome scores at 3 months, 12 months and 5–8 years (total n=310 for each time point).

Five-year to eight-year outcomes are presented in table 2. The mean (SD) pain severity score (worst or activity-related pain) among the responding participants was 29.9 (27.7) at the long-term follow-up, while the AKPS was 81 (14.5).

Table 2

Five-year to eight-year outcomes (n=60). Values are mean (SD) unless otherwise stated

Radiographic features

Frequencies of radiographic features for knee OA at 5–8 years are presented in table 3. A K&L score of 0 or 1 was scored in 98% of the participants. Tibial osteophytes ≥2 on the AP view were present in two participants. Out of the 50 participants, 2 had patellofemoral osteophytes ≥2 on the lateral view. Only one participant had osteophytes ≥2 on the skyline view (lateral).

Table 3

Frequencies of radiographic features for knee osteoarthritis at 5–8-year follow-up (n=50)

Prognostic indicators

The multivariate analysis for worst or activity-related pain VAS at 5–8 years revealed that baseline symptom duration longer than 12 months was significantly associated with greater pain severity at long-term follow-up (β 2.90, 95% CI 1.14 to 4.65; table 4). The model, including symptom duration, recruitment source and AKPS, explained 21.6% of the total variance.

Table 4

Prognostics factors for pain severity (worst/activity) and Anterior Knee Pain Scale at 5–8-year follow-up

The multivariate model for AKPS at 5–8 years revealed that a lower (worse) AKPS score at baseline was significantly associated with a lower AKPS score at long-term follow-up (β 0.48, 95% CI 0.21 to 0.76; table 4), with the model explaining 19.6% of the total variance.


Consistent with our previous findings at 3 and 12 months,7 we found that a large proportion of people with PFP who responded to our survey experienced symptoms up to 8 years later, and more than half of these reported an unfavourable recovery. However, only two participants demonstrated radiographic signs of knee OA. We also identified that a longer duration of PFP (>12 months) and lower AKPS score at baseline were associated with worse pain and symptom severity, respectively, 5–8 years later.

The proportion of people with PFP who reported an unfavourable recovery increased from 40% (126/310) at 1-year follow-up,7 to 57% (34/60) after 5–8 years. This provides important evidence that PFP is not self-limiting. These findings have important implications for the primary contact setting (family physicians, physiotherapists)—our data suggest that physicians should frame PFP as a condition that has the potential to become chronic.

While previous studies have reported long-term PFP symptoms, the generalisability of their findings to the greater PFP population is limited, as they were attending orthopaedic clinics or half of the cohort had received an arthroscopy or surgery to the knee.2 ,5 ,9 ,10 In comparison, the interventions received by participants in our two trials reflect recent practice, whereby at least 50% of participants received a short-term efficacious intervention (eg, exercise therapy, physiotherapy, foot orthoses).6 ,8 For example, in the UK and the Netherlands, primary care for PFP consists of an information leaflet advising participants that PFP has a good prognosis.37 ,38 Our findings challenge this information, and should be integrated into educational material to provide PFP participants with more realistic expectations. Rathleff et al11 recently showed that adding exercise during school hours to patient education improved outcomes for 2 years in adolescent PFP.

Identifying subgroups more likely to have worse outcomes

The current study found that longer symptom duration (>12 months) at baseline predicts worse pain at 5–8-year follow-up. Considering the chronic nature of PFP, this provides an important extension of our previous findings. Collins et al7 reported that duration of PFP >2 months predicted poor outcomes for pain, symptoms and function after 1 year. Cumulatively, our findings suggest that early recognition and management of PFP, using effective interventions, is warranted to maximise prognosis, although our study design does not allow us to conclude that this will prove effective.

The identification of PFP subgroups that are more likely to have worse outcomes has implications for future research. Our findings suggest that future RCTs evaluating PFP interventions should consider stratifying participants based on symptom duration and worse pain and function scores at baseline. Furthermore, it is plausible that tailoring treatment programmes to PFP subgroups may enhance longer term outcomes. For example, a short intervention period, as used in our RCTs, may be insufficient to manage a chronic condition such as PFP, and those with longer duration and greater severity of symptoms may benefit from a longer treatment period. Physiotherapy interventions, including exercise and patellar taping, have been shown to be effective in reducing pain and improving function in the short term, but long-term (>12 months) effects are largely unknown.39–41 These considerations make the case for future studies to evaluate longer interventions with regular top-up sessions, which could potentially maximise long-term outcomes.

Radiographic OA was not a feature in this cohort

Contrary to the results of Kannus et al,2 whereby 35% of the young PFP participants had signs of OA on MRI, the majority of participants in the current study did not demonstrate signs of knee OA on X-ray. While our findings do not support the proposition that long-term PFP necessarily leads to PFOA,16 ,17 ,19 the radiographic criteria used may have been suboptimal for this population. In this relatively young PFP cohort (age range at baseline 14–40 years), X-ray cannot detect early OA changes. This may have led us to underestimate the percentage of participants with early signs of PFOA with radiography compared with what might be seen using MRI.42 ,43–45 On the other hand, it is also possible that these early OA signs are not yet present in this relatively young PFP population. Only 13 (26%) participants were scored as K&L grade 1, which is the strongest predictor for future definite OA.46–48 Taken with the findings of Kannus et al,2 future studies evaluating long-term PFP outcomes in younger adults should utilise MRI techniques to address the question of whether PFOA exists.

Strengths and limitations

This prospective, longitudinal study evaluated two of the largest PFP cohorts to date. While the proportion of responders to the 5–8-year follow-up was small (approximately 20%), the finding that baseline, 3-month and 12-month characteristics did not differ largely between responders and non-responders reduces the likelihood of responder bias. We observed a significantly greater proportion of women in the long-term follow-up (75%) than the non-responders (56%), a phenomenon common in epidemiological studies. We therefore included sex in the multivariate analyses. However, due to the high percentage of loss to follow-up, the results of the present study should be interpreted with caution as we analysed the results based on complete case analysis, which could introduce bias.49

Despite the smaller sample size, prognostic factors identified for 1-year outcomes7 (duration of PFP, worse AKPS score) remain significant predictors of outcome at 5–8 years, increasing confidence in our findings. However, the smaller sample size at 5–8 years (n=60) limited the number of potential predictor variables that could be included in the multivariate regression analyses. For this reason, we included the same prognostic factors that were found in the 12-month follow-up study.7 The relatively small adjusted R2 values observed indicate that other factors are important in determining PFP prognosis. Factors such as coping strategies and pain sensitivity have been suggested as possible prognostic factors in other musculoskeletal diseases, and are worth investigating in future PFP prognostic studies.50–52

Summary and conclusions

A substantial proportion of young adults with PFP who responded at follow-up still report notable symptoms after 5–8 years, despite initially receiving treatment and education. This supports previous findings that PFP is not a self-limiting condition, and suggests that efficacious interventions may be required on an ongoing basis to maximise longer term outcomes. The majority of participants did not have signs of radiographic knee OA at 5–8-year follow-up. Longer duration of pain and worse symptoms and function, measured at baseline, remain predictors of poor PFP prognosis up to 8 years later. Education of health practitioners and the general public is recommended, to change the long-held belief that PFP is self-limiting and to improve treatment outcomes.53 ,54

What are the new findings?

  • A substantial proportion (>50%) of people with patellofemoral pain (PFP) who responded at follow-up still have notable symptoms 5–8 years after a 6-week programme of treatment and education.

  • The large majority of people with PFP at follow-up did not have signs of radiographic knee osteoarthritis, which does not support the proposition that long-term PFP is associated with structural patellofemoral osteoarthritis changes in younger adults.

How might it impact on clinical practice in the near future?

  • Medical professionals need to educate patients that PFP is not always self-limiting, especially those with long PFP duration (>12 months) and worse symptoms and function.


The authors thank Femke Wagner for scoring the knee radiographs.


View Abstract


  • Funding The Australian long-term follow-up was funded by Commonwealth of Australia National Health and Medical Research Council Program Grant #631717. The Dutch long-term follow-up was funded by The Royal Dutch Society for Physical Therapy (KNGF)/WCF and The Netherlands Organisation for Health Research and Development (ZON-MW) (project number: 945-04-356) and partly funded by a programme grant of the Dutch Arthritis Foundation.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval Ethical approval was obtained prior to start of each study (The University of Queensland's Medical Research Ethics Committee; The Erasmus Medical University, Rotterdam.

  • Provenance and peer review Not commissioned; externally peer reviewed.