1 Introduction

Patients with symptoms of knee or hip osteoarthritis are often referred to an orthopedic surgeon for the evaluation of therapeutical options.^1,2 X-ray imaging is a common practice in diagnosing knee and hip osteoarthritis for patients with distinctive features like stiffness and joint pain.^3,4 The Kellgren & Lawrence (KL) system classifies the severity of knee osteoarthritis.⁵ A clear X-ray confirmation of osteoarthritis parallel to the clinical signs and a motivated patient bring up the indication for an arthroplasty.^1,6,7 However, a dilemma arises if a patient experiences the symptoms of osteoarthritis without any radiologic features on X-ray. This discrepancy between clinical and X-ray assessment raises the question which factors are decisive to proceed to arthroplasty after all.⁶

Studies focusing on prognostic patient factors for knee or hip arthroplasty are surprisingly rare, especially for the hip.^8–11 To our knowledge, none of these studies consider minimal osteoarthritis on the X-ray as an inclusion criteria. Besides, the follow-up time of these studies is short in comparison to the chronic and gradual course of osteoarthritis. This calls for an analysis of long-term prognostic predictors of knee and hip arthroplasty without distinctive radiologic features. The European League Against Rheumatism (EULAR) committee stressed the importance of prognostic studies to identify osteoarthritis phenotypes with faster progression.¹²

The Cohort Hip and Cohort Knee (CHECK) study consists of participants with early symptomatic osteoarthritis of the knee and/or hip. A range of clinical, radiographical and biochemical variables have been assessed in order to create a platform for the determination of the course, prognosis and etiology of early symptomatic osteoarthritis. This cohort contains symptomatic patients without outspoken osteoarthritis on the X-ray and has a 10-years follow-up time.¹¹

The aim of this study is to investigate the patient factors predicting arthroplasty of the knee and hip when typical X-ray features are absent. Identification of the prognostic factors would be a step forward in the personalized patient care for symptomatic patients with knee or hip osteoarthritis. Orthopedic surgeons experiencing the discrepancy between clinical and X-ray assessment could use these predictors in their decision-making to prevent or to proceed to arthroplasty. Besides, this study will provide the incidence of arthroplasty in patients with mild knee or hip osteoarthritis.

2 Methods

All participants from the CHECK cohort with KL 0–2 osteoarthritis on the X-ray were considered for inclusion. This cohort included 1002 individuals, aged 45–65 with knee and/or hip pain, from October 2002 until September 2005 in The Netherlands. Baseline data and the dates of these baseline measurements were acquired from the CHECK database. For 10 years, participants were asked by the CHECK organization if they had undergone any knee or hip arthroplasty.¹¹ After 10 years of follow-up, 604 participants gave permission to be approached for new studies. To assess the primary outcome, these participants were contacted by letter for informed consent. Furthermore, they were asked whether they had undergone any knee or hip arthroplasty after the 10-year follow-up. Moreover, information about the side and date of surgery were requested. The letters were sent by the CHECK organization and the investigators only received the anonymously coded forms. If there was no permission to approach a participant or in the case of non-responders, there was still data about arthroplasties, sides and dates of surgery from the 10-year CHECK database. Exact dates of the annual assessments were not available. Therefore, each annual assessment was assumed 365 days from the previous one. This long-term follow-up has been approved by the Medical Ethics Committee of the University Medical Center in Utrecht (UMCU, reference number 16–583/C).

Two databases were created, one for the knee and one for the hip. During the follow-up, an earlier arthroplasty of the initially painless contralateral side meant the censored end of the follow-up of that participant, as this would give a bias based on the experience of surgery.¹³ If the pain was located bilaterally, the first arthroplasty was defined as an event. In the case of a missing day or month of surgery, January 1st of that year was chosen standardly. While the CHECK database contains clinical and radiological variables on a yearly base, only the baseline data was used for the survival analysis. This was because the decision for therapy is made on one fixed point and most osteoarthritis patients do not visit their orthopedic surgeon annually. The potential baseline predictors were chosen based on the one in ten rule, the literature, the applicability and the permanent nature of these variables.

The results were analyzed using SPSS Statistics software package version 25.0 with the Cox proportional hazards regression method to estimate the probability of arthroplasty (event) versus no arthroplasty (censoring). For a valid inclusion in the Cox regression method the assumption of a proportional hazard ratio and non-informative censoring were checked. To satisfy this latter assumption, the study must ensure that the occurrence of censoring is not related to the chance of an event happening. Moreover, the omnibus test was used to determine if any chosen variable should be removed from the analysis. A P-value < 0.05 was considered statistically significant.

3 Results

A total number of 333 filled questionnaires were received. April 19th, 2018 was selected as the end date for received participant questionnaires without any arthroplasty, as all data was collected by then. 830 participants were identified as having knee pain and 579 experiencing pain of the hip. Sex, age, Body Mass Index (BMI) and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) total sum score were chosen as potential predictors for knee arthroplasty. For the hip, the variables were as follows: sex, age, BMI, WOMAC total sum score, contralateral pain and the perception of the own health. The assumption for non-informative censoring was met. Table 1 illustrates the baseline data and the selected variables of these groups.

Two patients in the knee group and 8 in the hip group underwent arthroplasty during the follow-up, while having no pain at the concerning side at the baseline. Two knee and 9 hip participants were excluded for having KL > 2. Missing data consisted of BMI (knees: n = 20, 2.4%, hips: n = 12, 2.1%), WOMAC total sum score (knees: n = 23, 2.8%, hips: n = 17, 2.9%) and health perception (hips: n = 16, 2.8%). Multiple imputation was not used since most missing data originated from the same individuals lacking the majority of variables.

In the knee cohort, the proportional hazards assumption was met for sex, the only categorical variable. The inclusion of all the selected variables made sense using the omnibus tests of the model (P < 0.05). Twenty-five (3.0%) of the cases were dropped due to missing values and 10 (1.2%) due to censoring before the earliest event in a stratum. The analysis is based on 40 (4.8%) events and 755 (91%) censored outcomes. The following variables were found to be statistically significant: Sex HR (Hazard Ratio) 0.207 (95% CI 0.050–0.861) P = 0.030, BMI HR 1.081 (95% CI 1.013–1.153) P = 0.018 and WOMAC total sum score HR 1.022 (95% CI 1.004–1.040) P = 0.017. Age did not predict the outcome (P = 0.079). Fig. 1 illustrates the survival of the participants in the knee cohort.

Close

Fig. 1 Knee cohort survival plot The survival plot for knee arthroplasty shows a grossly linear occurrence of events. As the plot illustrates, most censoring occurred at the end of the follow-up.

Export to PPT

In the hip cohort, the proportional hazards assumption was met for sex and contralateral pain. Again, the inclusion of all the selected variables made sense using the omnibus tests of the model (P < 0.05). Twenty-three (4.0%) of the cases were dropped due to missing values and 8 (1.4%) due to censoring before the earliest event in a stratum. The analysis has been conducted with 58 (10.0%) events and 490 (84.6%) censored outcomes. The following variables were found to be statistically significant: Sex HR 2.103 (95% CI 1.179–3.752) P = 0.012, age HR 1.062 (95% CI 1.009–1.117) P = 0.022 and WOMAC total sum score HR 1.019 (95% CI 1.002–1.036) P = 0.029. BMI (P = 0.576), contralateral pain (P = 0.877) and health perception (P = 0.405) were not significantly predictive for the outcome. Fig. 2 illustrates the survival of the participants in the knee cohort.

Close

Fig. 2 Hip cohort survival plot The occurrence of events in the group with hip pain is relatively linear. Most censoring occurred at the end of the follow-up and a larger proportion of participants underwent arthroplasty compared to the knee cohort.

Export to PPT

4 Discussion

The purpose of this study was to determine the patient factors predicting knee and hip arthroplasty when typical X-ray features of osteoarthritis are absent. Remarkable is the result that women were more likely to undergo knee arthroplasty (79% increase), while men were more likely to undergo hip arthroplasty (110% increase). An increase of one point in the BMI raised the probability of knee arthroplasty with 8.1%, while this was not the case regarding hip arthroplasties (P = 0.576). Being one year older in age, indicates 6.2% more chance undergoing hip arthroplasty, while this was not the case for the knee (P = 0.079). A one point increase in WOMAC total sum score meant 2.2% more chance for knee arthroplasty, and 1.9% more chance for a hip arthroplasty. On the other hand, contralateral pain (P = 0.877) and health perception (P = 0.405) did not predict hip arthroplasty. The incidence in this cohort of undergoing arthroplasty was 5.1% (in 10.2 years) for the knee and 10.2% (in 9.7 years) for the hip. In conclusion, these predictors increase the chance for future arthroplasty in symptomatic patients lacking clear features of osteoarthritis on the X-ray. These predictors could be decisive for physicians reaching the dilemma whether the treatment should be based on the clinical symptoms or the X-ray. Based on the personalized predictors, the decision should be focused on prevention or proceeding to arthroplasty.

The diagnosis knee osteoarthritis does not require X-ray imaging if other knee pathology is unlikely.^14,15 Yet, the use of imaging is extensive in the clinical practice and it is frequently used as an aid in the decision-making for surgical treatment.¹⁶ X-ray imaging loses its supporting role when discrepancy appears due to absent radiological features of osteoarthritis and present clinical symptoms. Magnetic Resonance Imaging (MRI) is more accurate in finding cartilage morphology, but cannot be absolutely predictive in early knee osteoarthritis due to low discriminative ability.¹⁷ A physician could simply wait and see how future X-rays will be.⁶ According to a recent statement by EULAR, follow-up imaging is recommended if there is a rapid and unexpected progression of clinical symptoms.¹⁵ Still, it remains unclear whether the decision to proceed to arthroplasty is based on the new X-ray or the deteriorating clinical symptoms.⁶ On the other hand, a considerable part of patients undergoing arthroplasty have little radiological osteoarthritis.¹⁶ It is desirable to be able to predict progression to arthroplasty and prevent patient suffering by means of sooner surgery or preventive measures.

Previous studies have found risk factors for the radiological and, to a lesser extent, clinical progression of knee or hip osteoarthritis.^8,9 This raises the same question of which aspect is more important.⁶ A few studies have determined predictors for knee and hip arthroplasty. Yet, having no or mild features of osteoarthritis on the X-ray was not the inclusion criterium.^18–22 This might give a bias since the KL grade of osteoarthritis itself is a predictor of arthroplasty.^16,23 To solve this bias, this study only included participants with KL 0–2. Furthermore, the long period of follow-up and relative large cohort formed a platform to answer this question in a more expanded way than the study of Roemer et al.¹⁰ This latter study consisted of a few dozen participants with KL < 2 and investigated MRI findings and pain levels only.¹⁰ Riddle et al. did determine prognostic factors predicting knee arthroplasty in 3 years, however, participants from all KL grades were included.¹⁸ Other studies did only include participants with KL < 2, but they examined predictors of radiologic progression and not knee symptoms or arthroplasty.^4,8,24

This current study has some limitations. First, any experience with arthroplasty might give a bias in deciding whether to proceed to another one.¹³ For this reason, a contralateral arthroplasty meant the end of the follow-up. However, we did not take into account if knee and hip arthroplasties influenced each other. Second, although the WOMAC total sum score covers a range of clinical parameters, it is not clear how its predictive value could be applied in the clinical practice. Third, these results are not validated in another cohort, making them more uncertain.

The predictors found in this study, should be interpreted as predictors for this specific population only. Prognostic predictors for knee and hip arthroplasty are highly contradictory and phenotype dependent.^25,26 Previous studies have focused on prediction of the radiological progression of osteoarthritis. While the radiological progression is associated with symptoms, it is not the same.^8,27 Among the most common predictors of knee and hip arthroplasty are an increased age, female gender, less function, more pain, and the radiographic severity.^4,14,28 In our study, increased age only predicted hip arthroplasty. Besides, being a woman was a predictor for knee arthroplasty while being a man was a predictor for hip arthroplasty. The studies discussed in the National Institute for Health and Care Excellence (NICE) guidelines suggest the other way around.¹⁴ BMI is a known risk factor for osteoarthritis and knee and hip arthroplasty. This current study showed BMI only to be a predictor for knee arthroplasty. This might be due to the more weightbearing character of the knee. Nevertheless, adiposity would also have an inflammatory effect on non-weightbearing joints as well. In addition, orthopedic surgeons might be reluctant to perform surgery due to higher perioperative risks.^14,29–32 Multiple studies have suggested that WOMAC total sum score predicts radiological deterioration of osteoarthritis, while the NICE guideline reports mixed results regarding arthroplasty.^8,14,19,23 On the other hand, arthroplasty is recommended in patients with pain, stiffness and function loss, covered by the WOMAC total sum score.¹⁴ The reason health perception was chosen as a potential predictor is the impact of comorbidities on pain and functioning in knee and hip osteoarthritis. Also, it covers the physiological aspect of health and the 0–100 scale is easy to use in the clinical practice.^33,34

The decision whether to proceed to knee and hip arthroplasty is a complex one.²⁸ It is multi-factorial, phenotype dependent and shared decision-making should play a role in it. There are simply no clear general criteria for arthroplasty and finding them is difficult.^1,6 However, waiting for radiologic features of osteoarthritis with clinical disease progression is not recommendable.^14,15 The experience of symptoms is subjective and personal, and will leave its mark on the shared decision-making process. Arthroplasty is the only definitive and summarizing end point in this complex process. The predictors in this study could be an aid in the decision-making. More research and validation is needed in finding the major predictors, keeping in mind that it is not pure science.

Credit author statement

Danial Zarringam: Conceptualization, Methodology, Formal Analysis, Investigation, Resources, Writing – Original Draft, Writing – Review & Editing, Visualization, Project Administration. Daniel Saris: Conceptualization, Methodology, Writing – Original Draft, Writing – Review & Editing. Joris Bekkers: Conceptualization, Methodology, Investigation, Writing – Original Draft, Writing – Review & Editing, Supervision, Project Administration.

Funding source

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Ethical approval

This long-term follow-up has been approved by the medical ethics comity of the University Medical Center in Utrecht (UMCU, reference number 16–583/C). Patients were approached by letter and provided informed consent.

Informed consent

Informed consent was obtained from all subjects before the study by letter.

Trial registration

Not applicable.

Data statement

The data used is property of the CHECK organization. Please visit check-onderzoek.nl for more information concerning the availability of the data.

Funding declaration

This study has no involvement of study sponsors.