Introduction

A growing body of evidence has pointed out the clinical importance of patellofemoral (PF) osteoarthritis (OA), with a prevalence of ~ 25% among the adult population worldwide, and one of the leading causes of knee pain and disability [1,2,3,4]. Despite the high clinical importance and prevalence, the main risk factors of PF OA and impact of physical activities on the incidence/progression of this medical condition have not been comprehensibly assessed [5,6,7].

A few studies reported an increased risk of tibiofemoral knee OA progression due to prolonged knee-bending activities. Among these activities, a causal relationship was observed between certain occupational activities that involve excessive kneeling with structural worsening of tibiofemoral OA [8,9,10,11,12,13]. However, it is doubtful whether the kneeling activity can be associated with the increased risk of PF OA progression as well. A limited number of researches have focused on this question with an overall conflicting finding. Among these studies, a few reports suggested a higher chance of having PF OA in subjects with frequent kneeling [8], whereas, others found no significant association between kneeling activity and PF OA development [9]. Moreover, all previous studies were performed using the cross-sectional design and defined their exposure as the general knee-bending activities including climbing, squatting, and kneeling (rather than only kneeling activity) [8,9,10,11]. Furthermore, the development of PF OA was mainly assessed using a knee radiograph which is not sensitive to early and subtle OA-related structural damage worsening [8,9,10,11] Magnetic resonance imaging (MRI) has gained much attention in OA research as it provides a unique opportunity to study early changes of PF OA-related structural damages.

Several reports demonstrated that PF morphology abnormalities including patella alta (PA) as significant predictors of PF OA development [14]. PA is defined as a high-riding patella relative to the trochlear groove, which may result in the instability of the PF joint during knee extension and flexion movements [15, 16]. High-riding patella reduces the contact area between the patella and trochlea during knee motion. It can lead to increased articular surface contact pressure which subsequently may result in PF OA [17,18,19]. The Insall-Salvati ratio (ISR) is a validated measurement to determine PA using sagittal-plane MRI, CT, or lateral knee radiograph [20, 21]. Due to the relationship between the PA and biomechanical loading during knee motion as well as its relation to PF OA, it can be hypothesized that the presence of PA might mediate the association between PF OA development and excessive kneeling activities.

The Osteoarthritis Initiative (OAI) cohort provides the opportunity to evaluate the potential impact of kneeling activity on PF OA-related cartilage damage worsening. This study aims to assess whether kneeling activity is associated with the worsening of MRI measures of PF cartilage damage in subjects with and without PA. Understanding how frequent kneeling activity may affect PF joint might provide further insights into the pathogenesis and primary prevention strategies for PF OA.

Methods

Study population

The OAI is an ongoing multicenter cohort of 4796 subjects with or at risk of knee OA (for details: https://www.niams.nih.gov/grants-funding/funded-research/osteoarthritis-initiative). The OAI is a HIPAA-compliant study and has been approved by the institutional medical ethics review boards of the OAI centers (approval number: 10-00532).

We used the data from the Foundation for the National Institute of Health (FNIH) OA biomarker consortium project in this study. The FNIH is a nested case-control of 600 subjects within the OAI. The FNIH study was primarily designed to assess the role of baseline/24-month knee MRI findings in predicting 48-month symptomatic/radiographic knee OA progression (for details: https://oai.epi-ucsf.org/datarelease/FNIH.asp). FNIH biomarker consortium data contains MRI-based OA-related structural damage measurements for PF joint and provides the opportunity to study risk factors of PF OA. In the present study, we included all participants of the FNIH project (one index knee per subject). Demographic characteristics of included participants (age, sex, body mass index (BMI)), the frequency/duration of kneeling activity, 3-T MRI of knee, and MRI-based measurements were extracted and assessed at baseline and 24 months. The timeline of the study is presented in Fig. 1.

Fig. 1
figure 1

Timeline of the study. Before study: Average frequency of kneeling activity (≥ 30 min) per week was evaluated in all enrolled subjects. Baseline: Insall-Salvati ratio (ISR) measurement, baseline MRI assessments for patellofemoral osteoarthritis (OA) including MRI Osteoarthritis Knee Score (MOAKS). 24-month: follow-up MRI assessments of patellofemoral OA

Full size image

Magnetic resonance imaging pulse sequence protocols

MRI was performed using the same 3-T scanners at four OAI site centers (Trio, Siemens Healthcare). Briefly, the MRI protocol included sagittal three-dimensional (3D) dual-echo at steady state (DESS), a coronal two-dimensional (2D) intermediate-weighted (IW) turbo spin-echo (TSE), sagittal IW fat-saturated TSE sequences, and coronal and axial multiplanar reformations of the 3D DESS [22, 23].

Semi-quantitative assessment of patellofemoral joint

OA-related cartilage damage worsening in PF joints was scored by two experienced musculoskeletal radiologists (A.G. and F.W.R.; with 16 and 14 years of experience) using the validated MRI Osteoarthritis Knee Scoring (MOAKS) method, which is a semi-quantitative scoring system of OA-related features [23, 24]. MOAKS cartilage scores (cartilage surface and full-thickness scores) were evaluated in the patellar and trochlear regions (medial and lateral; 4 subregions total). Cartilage surface area and extent of full-thickness articular damage were determined using the same scale, where 0 was the absence of any damage, and 1, 2, and 3 showed damages involving < 10%, 10–75%, and > 75% of cartilage surface or full-thickness areas, respectively (Fig. 2). Previous studies reported the intra-rater and inter-rater reliability of moderate-perfect (0.64–1.0) and moderate-strong (0.62–0.93) for MOAKS measures, respectively [23, 24].

Fig. 2
figure 2

Insall-Salvati ratio and MRI Osteoarthritis Knee Score (MOAKS) assessment in patients with frequent kneeling (≥ 6 days/week), with and without patella alta. Sagittal IW fat-saturated TSE sequence from the left knee MRI of a 59-year-old female with patella alta and frequent kneeling activity. a The Insall-Salvati ratio was measured by dividing the patellar tendon length by the length of the patella on baseline MRI, which is ≥ 1.3. Baseline sagittal (a) and axial (b) MRI showed minimal-moderate PF cartilage defect with surface and full-thickness score of 2 (10–75% of cartilage surface and thickness involved) in semi-quantitative MRI Osteoarthritis Knee Score (MOAKS). Follow-up sagittal (c) and axial (d) MRI of the same patient at 24-month visit showing PF cartilage damage worsening for both surface and thickness scores. New subchondral bone marrow lesion is also detected on follow-up images. Baseline sagittal (e) and axial (f) IW fat-saturated TSE sequence right knee MRI of a 56-year-old female with frequent kneeling but without patella alta (Insall-Salvati ratio < 1.3) in baseline axial MRI (e). The 24-month follow-up, sagittal (G) and axial (H) MRI of the same patient showing no PF cartilage damage worsening in MOAKS scores

Full size image

To assess the worsening of cartilage damage, we used the worsening criteria defined by Runhaar et al [25]. Briefly, scores of cartilage were reported as worsened if interval (baseline to 24 months); increase was reported in surface and/or full-thickness damage scorings (Fig. 2). Minor within-grade changes were also considered worsening. A region was categorized into the worsening group if any of its subregions demonstrated worsening. We categorized knees with any worsening in cartilage damage into the PF cartilage worsening group.

Assessment of kneeling activity

At baseline visit, all FNIH subjects were asked, “How many days per week, you participated in the activities with kneeling ≥30 minutes?” Subjects reported the average frequency of their kneeling activities which last more than 30 min. All self-reported data were extracted from the OAI database and subjects were categorized into four groups of (A) controls, who had no kneeling activities during week; (B) 1 day of kneeling activities during week; (C) 2 to 5 days of kneeling activities during week; (D) frequent activity which is defined as 6 to 7 days of kneeling activities per week. It is pertinent to note that we followed the same concept of intention to treat (ITT) analysis and analyzed the results of our prospective study according to the study groups that were originally assigned at baseline visit. To minimize the possibility of attrition bias (i.e., changing frequency of kneeling activity over time), all clinical data regarding kneeling activity at 12-month and 24-month visits were also extracted and compared with baseline data, which showed no significant difference.

Insall-Salvati ratio measurement

We calculated the ISR on MRI as the length of the patellar tendon (PT) divided by the height of the patella (PH) (PT/PH) (Fig. 2). We defined the length of PT as the distance between the inferior poles of the patella to the insertion of the patellar tendon to the tibial tuberosity. PH was measured using the sagittal plane with the greatest diameter of the patella [26].

Two musculoskeletal radiologists (M.H. and C.S.) with more than 3 years of experience performed the measurements using baseline MRIs (sagittal MRIs, fully extended knees, IW fat-saturated TSE sequences); readers were blinded to the all outcomes. The intra-rater and inter-rater reliability of moderate-strong (0.6–0.85) for MRI-derived ISR were reported by previous studies [27]. Using MRI-derived ISR, knees were categorized into PA positive (+) with ISR of ≥ 1.3 and PA negative (-) group with ISR < 1.3 [26, 28].

Statistical analysis

Baseline characteristics were compared between the study groups using the ANOVA test (all normally distributed) and chi-square test. We assessed the association between the frequency of kneeling activities and longitudinal PF OA-related cartilage damage worsening (PF cartilage worsening) using a multivariable regression model. We defined the exposure as four grades of kneeling activities (groups A–D, explained earlier). We defined the outcomes as PF surface, full-thickness, or any cartilage worsening over 24 months (both for medial and lateral sides). We considered kneeling activity an “ordered categorical” variables with levels ranging from no kneeling activity to ≥ 6 days/week, and fitted the multivariable regression models. Then, to determine the kneeling level/threshold associated with the worsening of cartilage score, we fitted separate multivariate regression models with separate levels of kneeling activity (categories A–D, described above) as exposure. We adjusted all models for potentially confounding variables including age, sex, and BMI in model 1 and additional adjustment for predefined radiographic progression in the medial joint space narrowing (JSN) over 48-month follow-up of the FNIH study, in model 2. In order to assess the potential role of PA as the moderating variables, we performed two sets of regression analyses which include the presence/absence of PA as a covariate. By comparing the results of each fully adjusted regression model, we aimed to assess whether the association between kneeling activity and PF OA is different between subjects with and without PA. We separately calculated odds ratios of 95% confidence interval (ORs 95% CI) of the association between the frequency of kneeling activities and 24-month PF cartilage worsening in each model with (+) and without (-) PA as a covariate. We used SPSS statistical software, version 24 and the R platform (v.3.2.5) to conduct analysis (easyreg and er1 packages). Two-way p values lower than 0.05 were assumed to be statistically significant.

Results

The mean ± standard deviation (SD) of age and BMI in enrolled FNIH subjects was 61.6 ± 8.9 and 30.7 ± 4.8, respectively, and 58.8% of subjects were female. In all FNIH participants, 447 (74.5%) subjects had no kneeling activities during the week, whereas 82 (13.7%), 37 (6.2%), and 33 (5.5%) subjects reported less 1 day, 2–5 days, and ≥ 6 days of kneeling activities during the week, respectively. Table 1 shows the baseline characteristics of the study population. The analysis showed that a higher proportion of subjects with frequent kneeling activity are female (p value: 0.049). However, we detected no significant differences regarding the age and BMI between the study populations (Table 1).

Table 1 Baseline characteristics of the study population
Full size table

Our analysis considering kneeling activity as an ordered categorical variable showed a significant association between the frequency of kneeling activity and cartilage damage worsening over 24 months (model 2 OR 95% CI of 2.33 (1.08–5.06)) (Table 2). Furthermore, multivariable regression model stratified for different levels of kneeling activity (model 2: adjusted for age, sex, BMI, and JSN progression) showed only frequent kneeling activity, which was defined as ≥ 30 min of kneeling activities ≥ 6 days/week, and was associated with the higher odds of PF joint full-thickness (OR 95% CI of 2.95 (1.06–8.19)) and any cartilage worsening (OR 95% CI of 2.74 (1.03–7.27)) over 24 months when compared with subjects with < 6 days/week or no kneeling activities. Despite the trend, 2–5 days/week of kneeling activity was not associated with the worsening of PF cartilage damage worsening when compared with less kneeling activity (< 2 days/week) (Table 2). Assessment of each covariate in the models showed that progression in radiographic JSN is not significantly associated with the cartilage damage worsening (Supplementary Table 1).

Table 2 Frequency of kneeling activity and MRI-derived patellofemoral cartilage damage worsening over 24 months
Full size table

The possible role of PA in moderating impact of kneeling activity on PF joint cartilage damage was assessed in the last part. Overall, 139 (23.2%) patients had PA on baseline MRI (i.e., ISR of ≥ 1.3). Results showed that PA+ patients were more female (66.2% of PA+ and 56.6% of PA- were female; p value = 0.049). However, no significant difference was observed between PA+ and PA- groups regarding age (PA+ mean age = 60.3 ± 8.5, PA- = 61.9 ± 8.9; p value = 0.126) and BMI (PA+ mean BMI = 31.5 ± 5.0, PA- = 30.5 ± 4.7; p value = 0.433). Two sets of regression models (1: without PA and 2: with PA as the covariate) were performed (Table 3). As noted above, regression analysis showed a significant association between frequent kneeling activity and PF joint cartilage damage worsening over 24 months (Tables 2 and 3). When we added the presence of PA as an additional variable to our model, the OR (95% CI) will significantly decrease to 1.26 (0.78–2.04) (Table 3). This result suggested the potential role of PA as the moderator variable.

Table 3 Role of patella alta in moderating impact of kneeling activity on longitudinal MRI-derived patellofemoral cartilage damage worsening
Full size table

Discussion

We showed that frequent kneeling activity is associated with an increased risk of PF joint cartilage damage worsening, especially in subjects with an underlying PA. We also quantified this risk factor and demonstrated that ≥ 6 days/week of ≥ 30-min kneeling activities is associated with a higher risk of MRI-based PF cartilage damage worsening. Therefore, we could suggest that PA-positive subjects with obligated occupational or sport kneeling activity could reduce the frequency of their kneeling activity as a behavioral modification for primary prevention of PF OA.

Regarding the previously published study on this topic, Rytter et al showed that there are no significant differences regarding radiographic-based PF OA progression between occupations involving high versus low amount of kneeling activities [9]. Another study demonstrated that frequent generalized knee-bending activities (including but not limited to kneeling) are associated with the higher odds of having PF cartilage lesions worsening [8]. It should be noted that all these previous studies utilized a cross-sectional design, relied on radiographic measurements rather than MRI, and not focused on kneeling and its frequency/duration as a “potential dose-dependent and modifiable risk factor” for longitudinal PF OA progression [8,9,10,11].

Current treatment approaches for knee OA include palliative therapies with analgesics with no established protective effects in slowing OA progression, and knee replacement surgery for end-stage disease. To date, existing treatments for knee OA demonstrated controversial or little disease-modifying efficacy, and no therapy can be labeled as disease-modifying OA drugs (DMOADs) [29, 30]. Therefore, it has been suggested to minimize the risk factors of knee OA, especially among the at-risk population, to prevent disease progression. However, there are several issues in knee OA risk modifying and prevention. First, the main risk factors of knee OA, especially PF OA, were not established [31, 32]. For instance, the proven predictors of tibiofemoral OA, such as obesity and female gender, were not shown to be the main risk factors of PF OA [5,6,7]. Second, nearly all knee OA risk factors are non-modifiable. For instance, a large body of evidence demonstrated that older age, female gender, some races, positive family history of OA, presence of Heberden’s node, etc. are significant predictors of knee OA (all are not modifiable). Among all these factors, obesity is the only one that can be labeled as a knee OA modifiable risk factor. Therefore, it has been recommended to all patients who are at risk or early-stages of knee OA, to reduce their BMI to prevent OA incidence and/or progression [33].

All taken together, due to the high importance of determining PF OA risk factors and applying behavioral modification, we aimed to evaluate the possible association between PF OA and frequency of kneeling activity, as the potential modifiable risk factor. Our results demonstrated that frequent kneeling activity (≥ 6 days/week of ≥ 30 min) is associated with an increased risk of PF cartilage damage worsening over 24 months. Therefore, for the first time, we introduced kneeling activity as a “potential dose-dependent and modifiable risk factor” for PF OA cartilage worsening. All subjects who are at-risk of PF knee OA worsening, especially patients with daily occupational or sport kneeling activities, could be recommended to reduce their kneeling activity to prevent PF OA worsening.

We also found a higher level of association between the frequent kneeling activity and PF OA in subjects with PA. High-riding patella in PA might lead to the lower contact area between patella and trochlea during knee bending, and prone subjects to PF OA development [17,18,19]. It should be noted that although PA can be easily measured using a lateral radiograph or CT scan, we used MRI in our study since this imaging is only available in the OAI dataset. Therefore, it can be suggested to perform a lateral knee radiograph and evaluate the presence of PA in subjects with frequent kneeling activities and recommend them to reduce the frequency of their activities if evidence of PA is detected.

Several limitations existed in our study. First, the FNIH study outcomes were primarily designed based on tibiofemoral OA. We used available data of the FNIH study and performed ISR measurements. To minimize this limitation, we adjusted regression models for all known confounding variables. However, a more sophisticated analysis of cohort studies which is mainly designed to assess the main risk factors of PF OA should be performed in future works to confirm our initial observation. Second, the sample size and duration of follow-up may not be adequate, and the relationship between kneeling (especially in PA+ subjects) and PF OA needs to be further explored using more extensive databases with more extended follow-up data. Third, PF joint radiograph (lateral knee radiograph) and PF joint–specific symptom outcomes were not available using the OAI dataset. Therefore, PF joint cartilage damage and ISR were assessed using baseline and follow-up MR images of fully extended knees. Fourth, the frequency of kneeling activity was assessed at baseline visit which may lead to attrition bias since the frequency of activities varies over time. Considering the main concept of intention to treat analysis, we defined our exposed and non-exposed groups at baseline visit and analyzed prospectively collected data according to the study groups that were originally assigned. However, changing the frequency of kneeling activity over time may lead to underestimation for the impact of kneeling activity on PF cartilage damage and the overall association might be stronger than our reported estimates. To address this limitation and validate our findings, we have extracted follow-up clinical data regarding kneeling activity over 24 months which showed no significant difference when compared with baseline records.

Finally, our findings are the first exploratory observational report on the association between kneeling activity and MRI-based PF joint cartilage damage worsening using a longitudinal study design. Further studies are needed to be performed to clarify this association and evaluate the role of kneeling activity as a modifiable risk factor for PF OA development and/or worsening.