Implant survival - Chronic Kidney Disease in Patients Undergoing Hip or Knee Joint Replacement

6. Discussion

6.5 Implant survival

In the present study, overall implant survival rate was over 96% at median follow-up of 6.7 years with any revision as an endpoint, while that with revision for PJI was 99.1%. This is in line with earlier reports (Jämsen et al., 2013a; Junnila et al., 2016;

Lenguerrand et al., 2018; Lenguerrand et al., 2019; Miric et al., 2014a; Miric et al., 2014b).

Hip replacement patients had similar risk for any revision and revision for PJI

regardless of their renal function. However, the population was too small to reveal any revision operations among hip replacement recipients with CKD 4-5. Earlier studies (Bozic et al., 2012a; Lenguerrand et al., 2018; Miric et al., 2014a; Podmore et al., 2018;

Tan et al., 2016) have reported similar results (Table 3). However, one study (Tan et al., 2016b) demonstrated that patients with eGFR between 45 and 60mL/min/1.73m² had increased risk for PJI after hip or knee replacement. As their analysis did not address hip and knee replacements separately and also lacked a competing risk analysis, the results cannot be compared.

Knee replacement patients with normal renal function (CKD stage 1), had significantly more all-cause revisions than did patients with CKD stage 2-3. In the

improved life years, especially when treating slowly developing knee arthritis.

However, when a patient suffers from painful osteonecrosis of the femoral head, operation is usually indicated even given the poor renal function. As rehabilitation takes several weeks (Hamel et al., 2008) and life expectancy is limited, patients gain less quality adjusted life years. In this patient group, risk of death should be recognized and discussed preoperatively, and other treatment methods such as analgesics and enhanced physiotherapy should be considered as alternative treatment methods to surgery. Patients with CKD stage 3 had slightly reduced but still fairly good survival;

79% at five years. However, when non-CKD patients were compared to CKD

patients, a combination of CKD with other comorbidities led to poorer survival at five years. In patients with a combination of CKD and diabetes, survival was 69% at five years and 67% among patients with CKD and CHF. In patients with both CKD and coronary disease, 72% were still alive five years postoperatively. Thus, patients with CKD stage 3, and especially patients with a combination of CKD and diabetes, coronary disease or CHF, have a significantly higher risk for mortality and this should guide clinicians in decision-making, especially if the indication for surgery is borderline between whether to operate or not. In developed countries, healthcare budgets are vast but not limitless. Limited healthcare funds should be allocated to interventions that produce the most quality adjusted life years rather than to those interventions that donot. However, whether QALYs in joint replacement surgery differ in different CKD stages, is unknown and needs research in the future.

As preoperative CKD was strongly related to postoperative deaths, so was

postoperative AKI. AKI patients had poor survival compared to non-AKI patients.

However, in the cohort, it is not known whether postoperative AKI itself,

characteristics of AKI patients, or possibly both, contributed to high postoperative mortality. This issue also warrants future research. Similar to the present study, in hip replacement patients, postoperative AKI is known to increase in-hospital mortality

8-85

fold (Singh & Cleveland, 2020). In a series of 425 patients (with 67 AKI cases), no deaths were recorded at three-month follow-up (Kimmel et al., 2014). Other studies on postoperative AKI after joint replacement have not reported mortality (Ferguson et al., 2017; Jafari et al., 2010; Nowicka & Selvaraj, 2016; Perregaard et al., 2016).

6.5 Implant survival

Lenguerrand et al., 2018; Lenguerrand et al., 2019; Miric et al., 2014a; Miric et al., 2014b).

Hip replacement patients had similar risk for any revision and revision for PJI

Knee replacement patients with normal renal function (CKD stage 1), had significantly more all-cause revisions than did patients with CKD stage 2-3. In the

stratified analysis, this result was significant only in patients without diabetes and also among non-obese patients. However, according to the literature, implant survival has not been reported separately across all different CKD stages (Table 3). It must be noted that the finding could be a result of selection bias; patients with fewer

comorbidities end up having more revisions due to relative revision indications such as persistent postoperative pain. In contrast, patients with more morbidity burden usually end up in revision only with vital indications, such as periprosthetic fractures or PJI.

In knee replacement recipients with CKD 4-5 incidence and also cumulative probability of any revision and revision for PJI was greater than incidence of those with CKD stages 1-3, but as there were only three revisions altogether and two of them had been performed for PJI, the results were not statistically significant. Also, the two revisions for PJI explain the increased yet insignificant incidence of any revision in patients with CKD 4-5. Unfortunately, it was necessary to exclude patients with CKD stages 4-5 from the Fine and Gray multivariable analysis. Thus, the results concerning revision for PJI in patients with CKD stage 4-5 are from an unadjusted model. A register-based study by McCleery et al. (McCleery et al., 2010) including almost 60,000 knee replacements with over 3,700 renal failure patients (according to the International Classification of Diseases 9-; ICD-9 criteria), found that renal failure patients had significantly increased risk for PJI. Thus, the results of the present study are in concordance, but the population was too small to show significant results. In knee replacement patients undergoing dialysis, McCleery et al, (2010) reported a 4-fold risk for any revision. Instead, in the present study, increased incidence of all-cause revisions was fullyexplained by increased incidence of revision due to PJI. In their study (McCleery et al., 2010), among dialysis patients, there were 14 deep infections during the follow-up and 13 revisions for any reason occurring during follow-up.

Although it was not reported how many patients underwent revision due to PJI, it

could be suspected that many early revisions were due to PJI in that cohort. Still, in most earlier studies, there is no evidence that renal function could affect overall implant survival after knee replacement (Miric et al., 2014b; Podmore et al., 2018; Tan et al., 2016), but none of these studied all CKD stages separately (Table 3). With the exception of CKD stage 1 patients, the present study concurs with these studies.

In the recent literature, most studies (Bozic et al., 2012b; Kuo et al., 2017;

Lenguerrand et al., 2019; Miric et al., 2014b) have compared the risk for PJI after knee replacement between CKD patients (CKD stages 3-5) and patients without CKD (CKD stages 1-2) (Table 3). Some of these studies (Bozic et al., 2012b; Lenguerrand et al., 2019) demonstrated increased risk for PJI among CKD patients, while others (Kuo et al., 2017; Miric et al., 2014b) did not. The present study is the first to compare the risk across every CKD stage and therefore comparison to other studies is difficult.

Also, as most earlier studies were register studies that did not provide demographic data on their study populations, it cannot be determined whether demographic differences play a role in the discrepancy between the results. Also, in register studies, diagnosis of CKD was not based on SCr, but on (ICD) codes within the register, and thus is not comparable with the present study. The present study did not demonstrate any significant differences in number of revisions for PJI between CKD stages.

However, as previously discussed, the study population was too small to show if patients with CKD 4-5 had greater probability for revision due to PJI.

The present study was also the first to consider death as a competing risk when studying effects of renal function into implant survival. However, at eight years postoperatively, there were no major differences when comparing cumulative incidence numbers to cumulative probability of the CIFs that considered death as a competing risk (Tables 18 and 20). Therefore, the results of earlier studies lacking competing risk analysis can be interpreted as valid.

Although it was not reported how many patients underwent revision due to PJI, it

In the recent literature, most studies (Bozic et al., 2012b; Kuo et al., 2017;

However, as previously discussed, the study population was too small to show if patients with CKD 4-5 had greater probability for revision due to PJI.

6.6 Strengths and Limitations

The strengths of the present study included a relatively large cohort comprising all consecutive operations from a certain period of time from a single center. Compared to earlier studies in the field, length of the follow-up was relatively long. The study also used mortality data from an exclusive data pool including more than 95% of revision operations performed. The study also used a big cohort and a long follow-up period. All consecutive operations were included and the national population registers of mortality and revision data are extremely reliable. Also, comorbid conditions can be interpreted as valid due to being based on reimbursement right. However, in Study 4 comorbid conditions were self-reported.

The present study had also limitations. First, as 5% of the patients lacked preoperative SCr values, not all consecutive elective TJAs could be included in the analysis.

Presumably, this does not introduce any significant bias. The excluded group in Studies 1,2 and 3 included more female patients and more patients undergoing knee replacement, both groups being known to have more CKD. Thus, in this population, CKD prevalence may have been slightly underestimated. Still, most of the patients lacking preoperative measurements are probably outside the joint replacement hospital´s laboratory’s catchment area and on the other hand, emergency operation patients were excluded, and they would presumably have had worse eGFR

preoperatively.

Second, when evaluating AKI, SCr was obtained when clinically necessary, as it was in 30% of the patients. This probably reflects the real clinical practice in most

institutions. However, when assessing incidence of AKI based on fluctuations in SCr between pre- and postoperative values, the present study probably missed some cases of AKI, although the criteria were fairly strict. Therefore, in the present study, the low incidence of AKI should be interpreted with caution. Even so, this does not reduce the significance of the risk factor analysis. Conversely, having missed some AKI cases, the connection between reported risk factors and AKI would rather be more powerful than reported in this study.

Third, in Studies 1, 2, and 3 all the comorbid conditions were derived from

reimbursement right for costs of medications granted by the Finnish government to patients fulfilling the criteria and also having a written statement from their clinicians.

Additionally, the reimbursement criteria for coronary disease and CHF are not very strict and they are close to the diagnostic criteria that clinicians use. Even so, the criteria for reimbursement of hypertension and diabetes medications are stricter than the diagnostic criteria. Also, patients do not necessarily request such statements from their physicians and also some of the conditions may have gone unnoticed or

undiagnosed. In Study 4, patients’ comorbid conditions were self-reported. This explains the different prevalence of comorbid conditions in Studies 1, 2 and 3 compared to Study 4 as the population is from a single hospital but with a different time period (2002-2011 in the first studies and 2008-2017 in the latter). Prevalence of hypertension, diabetes, and coronary disease were 27%, 8%, and 9% in the first studies but 56%, 17%, and 9% in the later one. Hence, reimbursement rights revealed only half of the cases with diabetes and hypertension. As a consequence, the HRs of diabetes and hypertension for death reflect those with more severe disease. This highlights the role of CKD as a strong predictor of death when compared to diabetes and hypertension. There was, however, a similar prevalence of coronary disease in both study populations. In the same hospital, the study by Rajamaki T et al. (2015) found that when screening all patients preoperatively, 11% of patients undergoing