The influence of preoperative determinants on quality of life, functioning and pain after total knee and hip replacement: a pooled analysis of Dutch cohorts

Total knee or hip arthroplasty (TKA/THA) is an effective treatment for most individuals who suffer from pain and loss of function due to end stage symptomatic osteoarthritis of the hip and knee (OA). In 2010, 109 and 153 patients per 100,000 persons received a TKA or THA respectively in Europe [1]. The development and progression of OA are strongly influenced by age and obesity and both occur more frequently in women [1]. Parallel to the rising prevalence of knee and hip OA, due to an ageing society and obesity, surgery rates are rising as well [2‐4].

TKA and THA should not be performed too early since revision rates are higher in younger patients and the length of life of a prosthesis is limited [5]. On the other hand performing a surgery earlier gives more years of productive quality-adjusted life years (QALY’s). However, outcomes after revision surgery are generally worse compared to primary surgery [6]. Current practice shows that preoperative disease severity varies largely among centers and countries [7, 8], suggesting differences in timing. In addition, about 10–20% of the patients is not satisfied after primary TKA/THA [9‐12], possibly caused by unmet expectations of patients or due to suboptimal timing of surgery.

Previous research has identified preoperative determinants that influence outcomes, but these differed between studies and had opposite directions [13]. This may be due to lack of power so that some studies did not find any effect, while other studies did not adjust for confounders. In addition, most registries collect a minimal data set [14] e.g. only the VAS scale for pain. Therefore, pooling the data from available cohort studies may provide more reliable evidence on which determinants influence the outcome after TKA/THA because of the larger sample size than separate studies and a more comprehensive set of questionnaires with the ability to measure each outcome more reliable compared to registry studies.

The present pooled analysis of 1783 knee and 2400 hip OA patients shows that patients with a higher preoperative quality of life or functioning and less pain also have better postoperative outcomes but that they improve less. Furthermore, women and patients with a higher BMI had more postoperative pain and less improvement after both TKA and THA. Higher age and higher BMI was associated with lower postoperative HRQoL and functioning and more pain after a THA. However, preoperative quality of life, functioning and pain seem to be most consistently associated with outcomes after both TKA and THA.

Our results regarding the effect of preoperative status on outcomes are consistent with other studies that also found that patients with worse preoperative functioning had greater improvements [18‐21], but did not achieve the postoperative level of those with higher preoperative functioning [22‐28]. Contrary, other studies showed opposite results regarding the direction and size of the effect of age, gender and BMI. Santaguida et al. [29] performed a systematic review about patient characteristics affecting the prognosis after TKA/THA and concluded that an older age is related to worse functioning, but that age and sex do not influence postoperative pain level. We found that women had more pain after a TKA (4 points on a 100 point scale) and THA (2 points on a 100 point scale), even though this may not be a clinically relevant difference [30]. For TKA no association with age or gender and functioning was found. In addition, a previous review about prognostic determinants in THA reported that preoperative functioning was most consistently associated with better outcomes [13]. In addition, another systematic review on preoperative predictors on outcomes in THA [31] concluded that only patients’ poor preoperative functioning affects the outcome after THA. This was also found for patients with a TKA [32, 33]. Consistent with our finding, Lingard et al. [33] found that patients with severe pain had worse outcomes after a TKA. Other studies also identified other determinants, such as radiological scores, severity of inflammations or comorbidities. A disadvantage of using multiple cohorts with different protocols for data acquisition was that we could not include these determinants. The linear mixed model had to be the equal for each cohort, so that regression coefficients in each cohort have the same meaning. Thus the prognostic determinants found in this present study are not exhaustive; there may be other determinants that have an additional effect on the outcome.

The effect of different preoperative determinants on the postoperative outcomes after TKA and THA may seem to be small on itself, but if taken together they may add up to a clinically relevant effect. However, the scenarios should be interpreted with care, because these are hypothetical examples based on observational data and cannot be interpreted causally. The overall effects of the virtual scenarios which were calculated as examples vary between 1.2 and 6.5 points better postoperative outcomes and between 1.6 and 9 points worse postoperative outcomes. These scenarios provide more insights how small differences may add up or cancel each other out. This probably explains why most effects do not reach a clinically significant difference. Usually a 10% difference (i.e. 10 points on a 0–100 scale [30]) is considered as clinically relevant, but is a 10% difference the right criterion? Postoperative TKA/THA scores increases on average by 20–40 points on a 0–100 scale (results not shown) compared to preoperative scores regardless of the preoperative status. Thus is it realistic to use a difference of 10 points to define whether it is clinically relevant to operate now or wait, based on differences in preoperative determinants?

It is important to realize that the effects found in our study are not only the effect of the surgery, but also the effect of regression to the mean (RTM). RTM occurs because values are observed with random error, such as random fluctuations in a subject [34]. This means that patients with low preoperative scores are more likely to have higher scores during the next measurement and that patients with high preoperative scores are more likely to have lower scores during the next measurement, even without surgery. This results on average in a larger “improvement” for patients with lower preoperative scores compared to patients with higher baseline scores. Although different methods have been proposed to estimate the size of the RTM effect, but no solution is available to distinguish the real change due to surgery from the change due to RTM. Furthermore, we had to standardize different questionnaires measuring the same domain. Ideally, a minimal dataset should be composed, so that is more easily comparable without the need of standardization since standardized regression coefficients are more difficult to interpret [35]. A strength of our study is that we pooled existing cohort studies. Most of these studies collected a comprehensive set of questionnaires. Although national arthroplasty registries are established, these registries differ from clinical studies. Most registries focus on long-term data collection and therefore focus on minimal data sets and collect patient and operative information, but not all registries collect patient-reported outcomes [36]. If registries collect patient-reported outcomes such as HRQoL, function or pain most often short questionnaires are used, with only 1 or 2 questions covering the domain e.g. VAS-scale for pain or the EQ-5D to measure HRQoL. Most of the in our study included cohort studies used more comprehensive questionnaires with the ability to measure each outcome more reliable. On the other hand using questionnaires with composite scores has some weaknesses. Different patients may have very different domain scores but these may still result in the same composite score. In our study we therefore used domain scores of different questionnaires (functioning and pain) besides the overall HRQoL composite score, which may reduce this problem. Another potential problem is that there may be cultural differences between countries in how questionnaires are answered, but this would only influence our results if these cultural differences would affect e.g. elderly patients differently than younger patients thereby resulting in a different relationship of age with the outcomes.

The information regarding the combined effects of preoperative determinants on postoperative outcomes will support orthopaedic surgeons to estimate differences in outcome after a joint replacement for specific patient groups, i.e. poorer outcomes for patients with a worse preoperative status, but with greater postoperative improvement compared to patients with higher preoperative scores. In addition, preoperative status may decline during a long surgical delay period and thereby lead to worse postoperative outcomes if no other non-surgical treatments are started. On the other hand, it may sometimes be better to first optimize the patient’s preoperative condition or to reduce for example their BMI. The present study may support orthopaedic surgeons in their decision making by giving an estimate of the magnitude of the effect for different scenarios. Future studies should combine the results of our study with observational cohort studies among OA patients who did not have surgery yet, specific survival data from medical literature and the effects on survival of the artificial joint to assess optimal timing of surgery. This is needed to assess the long-term impact for the patient of the decision to perform surgery at a certain preoperative state of specific patient groups.