Article Text

Download PDFPDF

Extended report
Incidence, progression and sequence of development of radiographic knee osteoarthritis in a symptomatic population
  1. R Duncan1,
  2. G Peat1,
  3. E Thomas1,
  4. E M Hay1,
  5. P Croft1
  1. 1Arthritis Research UK Primary Care Centre, Primary Care Sciences, Keele University, Keele, UK
  1. Correspondence to Dr R Duncan, Arthritis Research UK Primary Care Centre, Primary Care Sciences, Keele University, Keele ST5 5BG, UK; r.c.duncan{at}


Objectives Symptomatic knee osteoarthritis (OA) is a common disabling condition. Attention has tended to focus on the tibiofemoral joint (TFJ). However, there is evidence that the patellofemoral joint (PFJ) is involved in many cases, but its place in the sequence of development and progression of knee OA is unclear. This study estimates the cumulative incidence, progression and inter-relationship of radiographic changes of OA in the TFJ and the PFJ in symptomatic adults.

Methods A population-based observational cohort of 414 adults aged ≥50 years with knee pain who had knee x-rays (weight-bearing posteroanterior semiflexed, skyline and lateral views) in 2002–3 and again in 2005–6 (mean interval 36.7 months) was studied. The outcome measure was the development of incident or progressive radiographic OA.

Results The 3-year cumulative incidences of patellofemoral joint osteoarthritis (PFJOA) and tibiofemoral joint osteoarthritis (TFJOA) were 28.8% and 21.7%, respectively. Corresponding estimates of 3-year cumulative progression were 18.9% and 25.3%. PFJOA at baseline was common and increased the risk of incident TFJOA (adjusted OR 2.2, 95% CI 1.1 to 4.1) but less clearly progression of TFJOA (adjusted OR 1.7, 95% CI 0.3 to 9.0). TFJOA at baseline increased the risk of PFJOA incidence and progression (adjusted OR 3.1, 95% CI 1.2 to 8.4 and OR 4.5, 95% CI 1.8 to 11.2, respectively).

Conclusions These results suggest a common sequence in the development of radiographic knee OA in symptomatic adults beginning in the PFJ, with subsequent addition and progression of TFJOA. It is proposed that isolated symptomatic PFJOA may be one marker for the future development of TFJOA and a target for the early management of knee OA.

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.


The burden of musculoskeletal conditions is rising.1 Osteoarthritis (OA) of the knee is a common condition which is strongly related to age. The prevalence of radiographic osteoarthritis (ROA) in the knee is expected to increase because of the current rise in both longevity and the prevalence of obesity. An understanding of the natural history of knee OA, which would help identify targets for early intervention, is still incomplete, particularly with respect to the role of OA in the patellofemoral joint (PFJ) of the knee, a condition recognised as important but often overlooked in population studies.2 3 Estimates of incidence and progression of ROA of the knee vary between population studies, with the annual incidence estimated to be 2–17%4,,6 and progression 3.6–23.8%.4,,8 This variability is attributable to a number of factors highlighted by Emrani et al,9 including radiographic approach and the definition of progression. Establishing the chronological sequence of development of ROA in the tibiofemoral and the patellofemoral joints, and their inter-relationship will help to identify symptomatic individuals ‘at risk’ of different patterns of onset and progression of OA who may benefit from targeted or prioritised healthcare.

The specific aims of this paper are: (1) to estimate the cumulative incidence and progression of OA in the TFJ and PFJ; and (2) to explore the association between OA in the TFJ and PFJ and the sequence in which it develops.


Study sample

The Knee Clinical Assessment Study (CAS-K) is a prospective observational cohort study of knee pain and OA in the general population. Baseline data were collected in 2002–3. All patients aged ≥50 years registered with three general practices in North Staffordshire were invited to take part in a two-stage postal survey. Respondents who answered positively to the question: ‘Have you had any pain in the last year in or around the knee?’10 were invited to attend a research clinic which included clinical interview, plain x-rays and a self-complete questionnaire. Detailed descriptions and recruitment and retention to the study have been published previously.11 12 After 3 years (2005–6), individuals participating at baseline were recontacted. Postal surveys and research clinic assessments identical to those at baseline were repeated.

Data collection

Clinical data

Data on the baseline severity of knee pain and function were based on self-report using the Western Ontario and McMaster Universities Osteoarthritis Index Likert version 3.13 14 Age, gender, body mass index (BMI) and duration of knee symptoms were recorded.

Plain radiography


We obtained three views of the knee: a weight-bearing posteroanterior (PA) semiflexed,15 a skyline and a lateral. The latter two views were obtained in a supine position with the knee flexed to 45°. The TFJ was assessed using the PA view and the posterior compartment of the lateral view. The PFJ was assessed using the skyline and lateral views.


A single reader (RD) scored all the study films at baseline and 3-year follow-up and was blinded to all clinical and questionnaire data and baseline x-ray results, but not to the sequence of the x-rays. Interobserver and intraobserver repeatability was tested in structured studies and was good, as reported previously.16 A Kellgren and Lawrence (K–L) score was assigned based on the authors' original written description of both the PA and skyline views.17 In the lateral view, superior and inferior patellar osteophytes were scored using a standard atlas.18 Osteophytes on the posterior tibial surface do not appear in the atlas but were judged on the same basis of severity as other osteophytes in the lateral view. Scoring of radiographic severity is shown in table 1. Detailed explanations of radiographic scoring and definitions have been published previously.16 19 20

Table 1

Definitions of the severity of patellofemoral and tibiofemoral ROA

Definition of incident and progressive ROA

We defined the incidence and progression of ROA based on radiographic assessment alone, and included only participants with complete radiographic data in the analysis. Only one knee per individual was analysed (the ‘index knee’), which was the single painful knee in participants with unilateral knee pain and the most painful knee in those with bilateral knee pain. This analysis explores the incidence and progression of radiographic OA in the PFJ and TFJ within a single knee rather than in an individual.

Joint-specific incidence

At 3-year follow-up, incident cases were defined as follows: (1) incident patellofemoral joint osteoarthritis (PFJOA) as the new appearance of mild or moderate/severe PFJOA in a knee that had no definite PFJOA at baseline; and (2) incident tibiofemoral joint osteoarthritis (TFJOA) as the new appearance of mild or moderate/severe TFJOA in a knee that had no definite TFJOA at baseline.

‘Rapidly progressive disease’ refers to a subset of individuals within the incident group who had moderate/severe ROA at 3 years (they developed new incident OA and also progressive OA within the 3 year follow up period.)

Joint-specific progression

At 3-year follow-up, progression was defined as follows: (1) progressive PFJOA as the new appearance of moderate/severe PFJOA in a knee that had mild PFJOA at baseline; (2) progressive TFJOA as the new appearance of moderate/severe TFJOA in a knee that had mild TFJOA at baseline.

Statistical analysis

The 3-year cumulative incidence and cumulative progression rates for PFJOA and TFJOA were calculated separately. Each of these estimates was then stratified by the presence and severity of ROA in the other joint at baseline. Logistic regression was used to estimate the association between baseline TFJOA and incident or progressive PFJOA at 3 years, expressed as crude OR and 95% CI and then adjusted for age, gender and BMI. The analysis was repeated to estimate the association between baseline PFJOA and incident and progressive TFJOA.

Sensitivity analyses

To address the impact of total knee replacements (TKRs) on estimates of incidence and progression, we looked at the effect of including TKRs occurring during the 3-year follow-up period as additional incident or progressive cases. Information regarding TKRs was obtained from the 3-year x-ray and self-report. We also investigated whether a K–L score of grade 1 (‘doubtful’ osteophyte) at baseline was associated with a higher 3-year cumulative incidence of ROA in either the PFJ or TFJ than those with no K–L grade 1 osteophytes at baseline.


Descriptive characteristics

Of 777 potentially eligible participants at baseline, 481 (62%) attended the 3-year follow-up clinic of whom 450 provided full repeat x-ray data. Thirty-six individuals had moderate/severe disease in both TFJOA and PFJOA in the index knee at baseline and hence could not develop either incident or progressive disease at follow-up, leaving 414 eligible participants with full repeat x-ray data. At baseline, 144 had no evidence of definite ROA in either PFJ or TFJ, 109 had isolated PFJOA (25 moderate/severe PFJOA), 19 had isolated TFJOA (5 moderate/severe TFJOA) and 142 had combined PFJOA and TFJOA (36 moderate/severe PFJOA, 65 moderate/severe TFJOA). Details of the response and selection of participants for our analyses are shown in figure 1. The mean (SD) time between the baseline and 3-year clinical attendance for these participants was 36.7 (1.1) months.

Figure 1

Flowchart of participants. PF, patellofemoral; ROA, radiographic osteoarthritis; TF tibiofemoral; TKR total knee replacement; 3YFU, 3-year follow-up.

Table 2 shows the baseline characteristics of participants and non-participants in the longitudinal analysis. Complete follow-up was more likely in men, younger individuals and those reporting a longer duration of knee symptoms.

Table 2

Baseline characteristics of participants and non-participants in the longitudinal analysis

Joint-specific estimates

Patellofemoral joint OA

Cumulative 3-year incidence

One hundred and sixty-three subjects had no PFJOA at baseline and were therefore at risk of incident PFJOA. At 3 years, 44 had mild PFJOA and three had moderate/severe PFJOA, giving an estimated 3-year cumulative incidence of 28.8% including an incidence of ‘rapidly progressive’ PFJOA of 1.8% (table 3).

Table 3

Cumulative incidence and progression of PFJOA and TFJOA and the inter-relationship

Cumulative 3-year progression

One hundred and ninety subjects had mild PFJOA at baseline and were therefore at risk of PFJOA progression. At 3 years, 36 had moderate/severe PFJOA, giving an estimated 3-year cumulative progression of 18.9%.

Tibiofemoral joint OA

Cumulative 3-year incidence

Two hundred and fifty-three subjects had no TFJOA at baseline. At 3 years, 33 had mild TFJOA and 22 had moderate/severe TFJOA, a cumulative 3-year incidence of 21.7% including an incidence of ‘rapidly progressive’ TFJOA of 8.7%.

Cumulative 3-year progression

Ninety-one subjects had mild TFJOA at baseline. At 3 years, 23 had moderate/severe TFJOA, a 3-year cumulative progression of 25.3%.

Sensitivity analysis

Of the 481 subjects attending the 3-year follow-up clinic, 18 (3.7%) had TKR in their index knee in the 3-year period since recruitment. Including this small number of cases did not markedly alter the estimates of 3-year cumulative incidence (PFJOA 29.7%, TFJOA 22.4%) or progression (PFJOA 23.4%, TFJOA 28.4%).

Of the 144 subjects with no definite ROA at baseline, those with K–L grade 1 osteophytes present in either the PFJ or TFJ at baseline had twice the 3-year cumulative incidence of ROA in either the PFJ or TFJ compared with those with no K–L grade 1 osteophytes (58.8% vs 24.5%).

Inter-relationship between PFJOA and TFJOA

The presence of TFJOA at baseline increased the risk of both incident and progressive PFJOA (crude OR incidence 3.2, 95% CI 1.2 to 8.5; crude OR progression 4.1, 95% CI 1.7 to 10.0; table 3). The findings remained almost identical in magnitude and precision after adjustment for age, gender and BMI (adjusted OR 3.1, 95% CI 1.2 to 8.4 and adjusted OR 4.5, 95% CI 1.8 to 11.2, respectively).

The presence of PFJOA increased the risk of incident TFJOA (crude OR 2.2, 95% CI 1.2 to 4.0), a finding which persisted following adjustment for age, gender and BMI (adjusted OR 2.2, 95% CI 1.1 to 4.1). The effect of PFJOA on the risk of progressive TFJOA was less clear, with neither adjusted nor unadjusted estimates reaching statistical significance (adjusted OR 1.7, 95% CI 0.3 to 9.0; unadjusted OR 2.3, 95% CI 0.5 to 10.9).

Sequence of development of knee OA

To assess the likely sequence of development of OA in the knee as a whole, we considered rates of development of OA in the two joints in the context of the ‘population of older knee pain sufferers’. Figure 2 illustrates the larger numbers of isolated prevalent PFJOA at baseline, which is consistent with the larger numbers developing new isolated PFJOA than new isolated TFJOA (24 vs 10) and, because of the larger pool of PFJOA at baseline, larger numbers of new combined PFJOA and TFJOA disease developed from those with isolated PFJOA at baseline than those with isolated TFJOA at baseline (32 vs 10) despite the higher individual risk of this occurring with TFJOA. Even if it is assumed (on the basis of the higher 3-year incidence of ‘rapidly progressive’ TFJOA) that all 13 subjects who started free of both PFJOA and TFJOA at baseline and developed combined PFJOA and TFJOA at 3 years had acquired TFJOA before PFJOA, it would remain true that most combined PFJOA and TFJOA cases arise from a sequence in which PFJOA occurs first followed by the addition of TFJOA.

Figure 2

Illustration of incident knee osteoarthritis. PFJOA, patellofemoral joint osteoarthritis; TFJOA, tibiofemoral joint osteoarthritis.

Linking this with the progression data, PFJOA is thus a risk factor for the onset of TFJOA but not for its subsequent rate of progression, even when PFJOA is severe. By contrast, both the presence and severity of TFJOA increases the risk of PFJOA progression.


Our results are consistent with—although do not prove—the view that OA of the knee is more likely to start in the PFJ,21 but is more likely to progress in severity in the presence of TFJOA.22

The CAS-K study estimated that, in older adults with knee pain, the 3-year cumulative incidence of PFJOA is 28.8% and TFJOA 21.7%. The 3-year cumulative progression of PFJOA is 18.9% and TFJOA 25.3%. Assuming incident and progressive disease to be constant over the period of follow-up, our study yields the following crude annual estimates: incident PFJOA 9.6%, incident TFJOA 7.2%, progressive PFJOA 6.3% and progressive TFJOA 8.4%. Differing methodology used in previous studies presents difficulties for direct comparison but some inferences can be drawn.

Our estimated annual incidence of TFJOA (7.2%) falls in the middle of the ranges derived from previous population studies in which annual incidence has been estimated to be between 2% and 17%.4,,6 Our annual estimate of TFJOA progression was 8.4%. This estimate falls within that reported in the review by Emrani et al9 and is similar to that reported by Spector et al4 (11.2%) but below that of Felson et al8 who reported an annual TFJOA progression of 20.2%. The use of a semiquantitative scale designed to detect minor joint space narrowing by Felson et al may explain some of the variation, as the scale used to detect progression in the CAS-K study may not have been as sensitive. We report the annual incidence of PFJOA to be 9.6%, higher than the annual incidence of 4.4% for PFJOA reported by Thorstensson et al.23 A number of factors might increase our estimate of incident PFJOA. An important difference was that we sampled symptomatic individuals with knee pain. Furthermore, our participants were all aged ≥50 years and both skyline and lateral views were used to image the PFJ. Annual progression of PFJOA in the CAS-K study was 6.3% while Cahue et al24 reported much higher progression of PFJOA. Different inclusion criteria and radiological methodology are likely to be responsible for the differing estimate.

Incident OA occurred more commonly in the PFJ. The presence of TFJOA at baseline significantly increased the risk of developing both incident and progressive PFJOA. The presence of PFJOA at baseline significantly increased the risk of incident TFJOA but did not affect the risk of progression of TFJOA (in contrast to a study by Mazzuca et al25). The results are consistent with a hypothesis that the first radiographic signs of OA often appear in the PFJ, but that the rate of subsequent progression is determined largely by the subsequent addition of TFJOA.

The high incidence of PFJOA in people with knee pain—which leads us to conclude that the majority of combined disease starts in this joint—highlights the potential importance of subjects with symptomatic PFJOA as an ‘at-risk’ group who may need to be prioritised for, and who would benefit from, targeted treatment as a means of preventing future progressive combined disease. However, the optimal non-operative management of PFJOA is yet to be established.26 Patellar taping has demonstrated some short-term benefits,27 28 and a more recent study reported that patellar tape reduced malalignment and pain associated with PFJOA.29 A non-randomised trial of intra-articular hylan G-F20 showed limited benefit.30 The challenge ahead is to develop specific interventions to prevent and slow down PFJOA and thus to reduce the incidence of combined joint disease.

The CAS-K study used a K–L grade of ≥2 to define ROA. However, the possible significance of a K–L score of grade 1 (‘doubtful osteophyte’) has also been highlighted.31 Our analysis has shown that the presence of a ‘doubtful’ osteophyte in the knee at baseline more than doubles the risk of incident OA, a finding reported in other population studies.31 32 These findings suggest that ‘doubtful’ osteophytes are relevant to the development of ROA, but reproducibility can be poor and misclassification common, so it remains to be established whether they should be included in the classification of ROA.33

Our initial incidence and progression analysis excluded those with TKR, but some would argue that these individuals should be included. Subsequent sensitivity analysis demonstrated only a minimal rise in our estimates of incidence and progression. However, by counting individuals who had incident or progressive OA in the PFJ, we have reported an upper estimate as we acknowledge that TKR is more likely to be for TFJOA.

Several aspects of our study deserve detailed critical comment. First, radiographic severity is defined into three categories (table 1). Our definition does not allow progression to be detected in the moderate/severe group so we cannot detect progression in at least one joint for 131 of the 414 participants. The use of plain radiography rather than MRI may have led to small radiographic changes going undetected, but it was felt to be more practical in a study of this size. Second, our definition does not absolutely separate osteophytosis from joint space narrowing, which may be regarded as a qualitatively different type of pathology to ROA. Third, the CAS-K study did not sample asymptomatic individuals and we cannot confidently assume that our results are generalisable to those without knee pain. Fourthly, by comparing baseline and 3-year data, it is apparent that a small number of individuals (n=31) appear to have regressed with respect to their x-ray status. This is unlikely to have influenced our results, since the data for these individuals were grouped with those individuals who did not change. Finally, 62% of our population were followed up at 3 years and loss to follow-up may have affected the size of our observed associations. Reassuringly, however, the pattern of baseline ROA was similar between those participating at 3-year follow-up and those not.

To our knowledge, this is the first paper to specifically report both detailed incidence and progression of OA in the PFJ in a symptomatic population sample. Our results suggest that a common sequence in the development of radiographic knee OA in symptomatic adults is onset in the PFJ, with progression strongly related to the subsequent onset of TFJOA. If confirmed, isolated symptomatic PFJOA may be a marker—even a target—for early management of knee OA. Further research into specific risk factors for PFJOA, the development of targeted treatments and evaluation of their ability to slow or prevent PFJOA are needed and, indeed, have commenced.34


The authors would like to acknowledge the contributions of Professor Krysia Dziedzic, June Handy, Charlotte Clements, Dr Jonathan Hill, Mel Holden, Dr Kika Konstantinou, Liz Mason, Dr Helen Myers, Dr Ross Wilkie and Dr Laurence Wood to aspects of the conception and design of the study and to the acquisition of data; Dr Milisa Blagojevic for aspects of data analysis; Dr Jacqueline Saklatvala, Carole Jackson and the team of radiographers from the Department of Radiography, Haywood Hospital for the acquisition of radiographic data; Professor Chris Buckland-Wright for advice and training on the radiographic techniques and Professor Iain McCall for his contribution to study concept and design. The authors would also like to thank the administrative and health informatics staff at the Arthritis Research UK Primary Care Centre, Keele University and the staff and patients of the participating general practices.



  • Funding This study is supported financially by a Programme Grant awarded by the Medical Research Council, UK (grant code: G9900220), a Programme Grant awarded by the Arthritis Research UK and by Support for Science funding secured by North Staffordshire Primary Care Research Consortium for NHS service support costs.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the North Staffordshire local research ethics committee (references 1430 and 05/Q2604/72).

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