A Meta-Analysis of Acute Kidney Injury in Patients Undergoing Hip Fracture Surgery: Prevalence and Risk Factors

Objective: Acute kidney injury(AKI) was a frequent complication following hip fracture surgery, but recent studies reported inconsistent ndings. The purpose of our study was to systematically clarify the prevalence and risk factors of AKI after hip fracture surgery. Methods: Pubmed, Embase, and Web of Science were searched from the inception to March 2020 to identify observational studies investigating the prevalence and risk factors of AKI in patients undergoing hip fracture surgery. Pooled prevalence and odds ratios(ORs) with 95% condence intervals(CIs) were estimated using random-effect model. Publication bias was evaluated with a funnel plot and statistical test. All the statistical analyses were performed using STATA version 12.0. Results: A total of 11 studies with 16,421 patients were included in the current meta-analysis. The pooled prevalence of AKI in patients undergoing hip fracture surgery was 17%(95%CI, 14%-21%). Postoperative serum albumin(OR 1.80; 95%CI, 1.38–2.36) was a signicant predictor for AKI. Age(OR 1.01; 95%CI, 0.95– 1.07) and ACE inhibitors(OR 1.38; 95%CI, 0.92–2.07) were also associated with increased risk of AKI, but the results were not statistically signicant. No signicant publication bias was identied through statistical tests(Egger’s test, p = 0.258 and Begg’s test, p = 0.087). Conclusions: The pooled AKI following hip fracture surgery was approximately 17%. Postoperative serum albumin was a potential signicant risk factor for AKI.


Introduction
The worldwide population is continuously aging, thereby leading to an increase in the prevalence of osteoporotic hip fractures. 1,2 A recent statistical estimation predicts that the patients receiving surgical interventions for hip fracture will double by 2050 owing to the aging of population. 3 Furthermore, hip fracture is a major public health challenge, for it causes serious morbidity, death, and socioeconomic burden. 3 Acute kidney injury (AKI) is a clinical syndrome featured with a sudden impairment in glomerular ltration and correlates with complex etiologies and pathophysiological processes. Every year, about 13.3 million cases are diagnosed with this intractable syndrome all over the world. 4,5 Worse still, it has been estimated that AKI results in a considerably mortality (1.7 million deaths per year) globally. 4,5 Notably, AKI is a common postoperative complication in patients undergoing surgical interventions for hip fracture, which closely correlated with prolonged hospital stay, various morbidity and increased mortality 6,7 Therefore, to adequately understand the landscape of prevalence and risk factors of AKI may largely help to improve outcomes of patients undergoing hip fracture surgery. Nevertheless, the reported prevalence of AKI associated with hip fracture surgery ranged widely from 5-60% currently. This remarkable range change may be caused by different AKI de nition, inconsistent follow-up durations and small sample size, and so on. It is common sense that an inaccurate estimation of AKI prevalence will prevent clinicians to grasp an overview of the disease burden and its natural history, and this makes it hard for them to evaluate whether speci c treatment strategies are effective for preventing AKI. Additionally, positively monitoring and coping with the risk factors for AKI may also do good to AKI prevention. 8,9 Therefore, in this study we performed a meta-analysis and systematic review to summarize the prevalence and associated risk factors for AKI in patients undergoing surgical treatment for hip fracture.

Methods
This meta-analysis was undertaken according to the guideline of the Meta-analysis of Observational Studies in Epidemiology(MOOSE) checklist and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis(PRISMA) statement. 10,11 Search Strategy And Study Selection A systematic literature search of Pubmed, Embase, and Web of Science was conducted from inception to March 2020 to retrieve studies that reported the prevalence and risk factors of AKI in patients undergoing surgical interventions for hip fracture. The systematic search strategy was established using the terms of "acute kidney injury", "hip fracture surgery" and their variants. A manual search for potentially eligible studies was performed as well by screening the references of the included literature.

Selection Criteria
Eligible studies should be cohort studies, case-control studies, or cross-sectional studies, which must provide the data to estimate prevalence and risk factors of AKI in patients undergoing surgery for hip fracture. Only studies published in English were considered. The Retrieved studies were individually evaluated for eligibility by the two investigators independently. Discrepancies in eligible study selection were resolved via discussion and mutual consensus.

Data Abstraction And Quality Assessment
We applied the pre-designed table to extract the following information: rst author, publication year, study period, country, operation type, case number, the number of patients with AKI, AKI de nition and risk factors of AKI. The primary outcome was the prevalence of AKI after surgery for hip fracture. The secondary outcome was the odds ratio (OR) with corresponding 95% con dence interval (CI), which evaluated the relevant risk factors of AKI after surgery for hip fracture. Furthermore, only ORs with CIs generated from the multivariate analysis in the included studies were extracted. Two authors independently assessed the risk of bias of each eligible study according to the Newcastle-Ottawa Scale (NOS) score. 12

Statistical analysis
The prevalences of AKI after surgery for hip fracture were extracted from eligible studies. Pooled estimated prevalences and 95% CIs were calculated with a random effect model. Odds ratios (ORs) with 95% CIs for risk factors were combined with a random effect model when substantial statistical heterogeneity existed across the included studies. Only candidate risk factors reported in two or more eligible studies on multivariable model were subjected for meta-analysis. I 2 statistic was applied to evaluate the statistical heterogeneity across eligible studies was assessed using with P value for Q < 0.05 or statistic I 2 > 50% regarded as substantial heterogeneity. 13,14 Meta-regression analysis on publication time, sample size and NOS score were used to explore the potential source of heterogeneity. Sensitivity analysis was undertaken by deleting one study each step to investigate the in uence of single study on the overall pooled estimated prevalence of AKI after surgery for hip fracture. Subgroup analyses based on region, sample size, study design, AKI De nition, and NOS score for the primary outcome were conducted to explore the prevalence of AKI in sub-populations. Publication bias was evaluated by Begg's and Egger's tests, in which P < 0.05 and asymmetric funnel plot indicated signi cant publication bias. 15,16 A two-sided P < 0.05 was identi ed as statistical signi cance. All the statistical analyses were performed using STATA 12.0 (Stata Corporation, College Station, TX, USA).

Study selection and characteristics
A total of 810 items were identi ed through systematically searching three databases. Furthermore, the full texts of 57 articles were screened for possible eligibility following removing duplicated and irrelative items. Eventually, a total of 11 studies with 16,421 patients were included in the current metaanalysis. 3,6,17−25 The ow chart of study selection was summarized in Fig. 1. The publication time of included studies ranged from 2010 to 2020. Seven studies were performed in Asia 6,18,19,21,22,24,25 , while the other four in Europe 3,17,20,23 . The operation types, study design, and AKI de nition of included studies were also different from each other. The whole NOS score of included studies ranged from 6 to 8 points, which suggested that the quality of included studies were moderate to high level. The detailed baselines characteristics and quality assessment of included studies were showed in Table 1 and Table 2.  All the included studies reported the prevalence of postoperative AKI in patients undergoing hip fracture surgery. The pooled prevalence of AKI following hip fracture surgery was 17% (95%CI, 0.14-0.21) with substantial heterogeneity(I 2 = 95%) (Fig. 2). Meta-regressions were conducted to explore the potential sources of statistical heterogeneity. The results indicated that publication time(p = 0.368), sample size(p = 0.593), and NOS score(p = 0.558) may not be the potential sources of statistical heterogeneity. Also, we performed strati ed analyses to explore the prevalence of AKI in subgroup patients. In subgroup analyses strati ed by region, the prevalence of AKI in Asia(22%) was higher than that in Europe(12%). When strati ed by sample size, the prevalence of AKI in sample size > 500(40%) was higher than that in sample size ≤ 500(15%). In subgroup analysis by study design, the prevalence of AKI in the subgroup of cohort study(14%) was lower than that in other subgroup(43%). Interestingly, the prevalences of AKI in subgroup strati ed by AKI de nition and NOS score were basically the overall pooled prevalence of AKI. The detailed results of subgroup analyses were showed in Table 3. We also undertook the sensitivity analysis to explore the in uence of individual included studies on the overall pooled estimate. We found that the pooled prevalence of POD basically remained stable, which indicated that pooled result was robust and credible (Fig. 3). We further evaluated the potential publication bias using the funnel plot and statistical tests. The funnel plot seemed to be asymmetric, but the statistical results indicated that the publication bias was not statistically signi cant (Egger's test, p = 0.258 and Begg's test, p = 0.087; Fig. 4).

Risk factors for AKI in patients undergoing hip fracture surgery
We also explore the potential risk factors associated with AKI in patients undergoing hip fracture surgery.
A total of seven studies reported AKI-associated risk factors on multivariate or adjusted model (Table 4). 3 Table 5) were also associated with increased risk of AKI in patients undergoing hip fracture surgery, but the results were not statistically signi cant.

Discussion
The current meta-analysis revealed that AKI was a relatively frequent complication in patients undergoing hip fracture surgery with pooled prevalence ranging from 14 to 21%. Additionally, postoperative serum albumin was identi ed to be a signi cant risk factor for AKI following hip fracture surgery.
The current meta-analysis based on 11 observational studies indicated that the overall pooled prevalence of AKI following hip fracture surgery was 17% with substantial heterogeneity. Considering that the signi cant heterogeneity may impair the credibility of the pooled estimate, meta-regression was performed to explore the potential sources of statistical heterogeneity. Furthermore, we identi ed that publication time, sample size, and NOS score may not be associated with signi cant heterogeneity. Furthermore, we conducted subgroup analysis and sensitivity analysis to explore the prevalence of AKI in sub-population. Interestingly, the results of subgroup analysis and sensitivity analysis were basically consistent with the overall pooled effect, which suggested that the pooled estimate was robust and reliable. A previous meta-analysis showed that the overall estimated prevalence rates of AKI in patients undergoing total hip arthroplasties are 6.3%. 26 Obviously, the prevalence of AKI following total hip arthroplasties was lower than that in patients undergoing hip fracture surgery. Regardless of the fact that the exact causes for these differences were largely unclear, but surgical workers should attach more importance to the potential AKI in patients undergoing hip fracture surgery. In the study, we also investigated the risk factors for AKI following hip fracture surgery. Pooled analysis showed that postoperative serum albumin was a signi cant indicator for AKI in patients undergoing hip fracture surgery. Consistent with our results, some previous studies also found that serum albumin level was a potential risk factor for AKI. Thongprayoon et al. revealed that there existed a U-shape correlation between serum albumin levels and AKI in hospitalized patients. 27 Dos Santos and coworker found that low serum albumin concentration was associated with increased risk of AKI in critically ill patients. 28 Mechanically, a recent study found that 5-Lypoxygenase products induced by albumin overload may be responsible for renal tubulointerstitial injury. 29 Other risk factors including age, ACE inhibitors, and eGFR were possible predictors for AKI, although the pooled results were not statistically signi cant. Collectively, perioperative management methods aimed at these risk factors may decrease the risk of AKI after hip fracture surgery.
There also existed several limitations in the current study. Firstly, our meta-analysis showed substantial statistical heterogeneity, which may potentially impair the reliability of the pooled estimate. Subsequently, we performed meta-regression to explore the sources of statistical heterogeneity and none of signi cant factors were identi ed to be responsible for heterogeneity. A possible interpretation is that multiple clinical and methodological difference across included studies, but not individual factor contribute to the signi cant statistical heterogeneity. Irrespective of the statistical heterogeneity, the results of subgroup analysis and sensitivity analysis were basically consistent with the overall pooled effect, which showed the robustness and reliability of the pooled estimate. Secondly, we evaluated the potential publication bias using the funnel plot and statistical tests. The statistical results showed that the publication bias was not statistically signi cant, but the funnel plot seemed to be asymmetric. Considering the inconsistence, the potential publication bias still cannot be excluded, although we performed a systematic literature search in the meta-analysis. Thirdly, some risk factors reported in included studies were not pooled for meta-analyses owing to limited studies, which may bias the authentic effects for AKI. The pooled analysis based on two studies found that ACE inhibitors may not be a signi cant risk factor for AKI following hip fracture surgery. Actually, many studies found that ACE inhibitors was a signi cant predictor for AKI. [30][31][32] Also, many risk factors including chronic kidney disease, intraoperative hypotension, and dementia were reported to be signi cant predictors for AKI, but we did not included for further pooled analyses owing to that they were reported in the limited studies. Therefore, the limited studies may bias the authentic estimates in the current meta-analysis. Accordingly, the results in our meta-analysis may be relatively conservative and should be interpreted in caution.
Take together, the current meta-analysis revealed that the pooled AKI in patients undergoing hip fracture surgery was approximately 17%. Postoperative serum albumin was identi ed to be a potential signi cant risk factor for AKI. Further high-quality studies should be warranted to systematically clarify the prevalence and risk factors of AKI following hip fracture surgery.
Abbreviations AKI: Acute kidney injury; CI: Con dence interval; OR: odds ratio Declarations Figure 1 Flow diagram of the selection of studies for this meta-analysis Forest plot for prevalence of AKI in patients undergoing hip fracture surgery using random-effects mode.

Figure 3
Sensitivity analysis for prevalence of AKI in patients undergoing hip fracture surgery in the meta-analysis.