Risk factors associated with failure of total ankle arthroplasty: a nationwide cohort study

We conducted a nationwide population-based cohort study to identify the risk factors associated with failure of total ankle arthroplasty (TAA). We included 2,914 subjects who underwent primary TAA between January 1, 2010, and December 31, 2016, utilizing the database of the Korean National Health Insurance Service. Failure of TAA was defined as revision TAA or arthrodesis procedures. An increased risk of TAA failure was observed in the < 65 age group versus the ≥ 75 age group [adjusted hazard ratios (aHR) 2.273, 95% confidence interval (CI) 1.223–4.226 in the 60–64 age group; aHR 2.697, 95% CI 1.405–5.178 in the 55–59 age group; aHR 2.281, 95% CI 1.145–4.543 in the 50–54 age group; aHR 2.851, 95% CI 1.311–6.203 in the < 50 age group]. Conversely, the ≥ 65 age group displayed no increase in the risk of TAA failure. The risk of TAA failure was increased in the severely obese group with body mass index (BMI) of ≥ 30 kg/m2 versus the normal BMI group (aHR 1.632; 95% CI 1.036–2.570). This population-based longitudinal study demonstrated that age < 65 years and BMI of ≥ 30 kg/m2 were associated with increased risk of TAA failure.

Covariates. We assessed subject demographics and lifestyle factors through standardized self-reporting questionnaires. Smoking status was classified into non-, ex-, and current smoking categories. Alcohol consumption was divided into three categories: heavy drinker (more than 30 g/day of alcohol), mild drinker (less than 30 g/day of alcohol), and nondrinker (no alcohol at all) 27,28 . Regular physical activity was defined as performing strenuous exercise for at least 20 min three times per week or moderate exercise for at least 30 min five times per week. Income level was dichotomized at the lowest 20%. Blood pressure (BP) and serum levels of glucose, lipid profile, and creatinine were measured after overnight fasting. Baseline comorbidities, such as hypertension and dyslipidemia, were identified based on the combination of the medical history, health examination results, or ICD-10-CM and prescription codes. We defined hypertension as BP ≥ 140/90 mmHg, or at least one claim/ year for an antihypertensive medication prescription under ICD-10-CM codes of I10-I13, I15. Diabetes mellitus (DM) was defined as fasting glucose ≥ 126 mg/dL, or at least one claim/year for an antidiabetic medication prescription under ICD-10-CM codes of E11-E14. Dyslipidemia was defined as a total cholesterol level ≥ 240 mg/ dL or at least one claim per year for lipid-lowering medication (ICD-10-CM code E78). Hospital volume was  www.nature.com/scientificreports/ categorized into five groups on the basis of the annual number of primary total ankle arthroplasties performed in each institute during the study period: < 10, 10-19, 20-29, 30-39, ≥ 40 primary procedures/year.

Statistical analysis.
The baseline characteristics were presented as the mean ± standard deviation or as a number with percentage. The Chi-square test was applied for categorical variables and Student's t test for continuous variables to compare the characteristics between the 'Non-failure' and 'Failure' groups. The incidence rate of TTA failure was calculated from the number of incident cases divided by the follow-up duration in person-years. The cumulative incidence of TTA failure during follow-up according to age and BMI categories was assessed by Kaplan-Meier curves; the differences among the groups were evaluated using the log-rank test. Cox proportional hazard regression analysis was used to evaluate the association of age and BMI with TAA failure, and hazard ratios (HRs) and 95% confidence interval (CIs) were calculated. The results were adjusted for confounders including age, sex, BMI, hospital volume, income, DM, hypertension, dyslipidemia, and health-related behaviors (physical activity, smoking, and alcohol consumption). Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). A 2-sided P value of 0.05 was considered statistically significant.

Results
Baseline characteristics. A comparison of the baseline characteristics between the non-failure and failure groups is presented in Table 1. A total of 2914 subjects who underwent primary TAA were included in this study. The failure of TAA occurred in 248 participants (8.5%) during the study period. The mean age (62.24 ± 8.26 years) of the failure group was significantly less than that (65.08 ± 8.29 years) of the non-failure group (P < 0.0001). The mean BMI (25.99 ± 3.65 kg/m 2 ) of failure group was significantly greater than that (25.42 ± 3.18 kg/m 2 ) of the non-failure group (P = 0.0079). There were no statistically significant differences between the two groups with regard to sex, income level, comorbidities such as DM, hypertension, and dyslipidemia, health-related behaviors (physical activity, smoking, and alcohol consumption), hospital volume, and cardiometabolic parameters such as systolic and diastolic BP, total cholesterol, and fasting plasma glucose.
Incidence and risk of TAA failure according to age and BMI. During a mean follow-up of 3.82 ± 1.99 years (11,144 person-years), 248 TAA failure developed. The cumulative incidence of TAA failure is presented according to age and BMI categories using Kaplan-Meier curves (Fig. 2). The incidence rate of TAA failure significantly differed between age categories (P = 0.0007) and was significantly higher in the age < 65 group than that of age ≥ 65 group (P < 0.0001). The incidence rate of TAA failure did not differ significantly between BMI categories (P = 0.0501) but was significantly higher in the BMI ≥ 30 group than that of BMI < 30 group (P = 0.0037). The HRs (95% CIs) of TAA failure according to age and BMI categories are presented in Table 2. An increase of age by one year was significantly associated with 3.5% decreased HR for TAA failure. An increased risk of TAA failure was observed in the age < 65 group versus age ≥ 75 group [adjusted hazard ratio (aHR) 2.273, 95% CI 1.223-4.226 in the 60-64 age group; aHR 2.697, 95% CI 1.405-5.178 in the 55-59 age group; aHR 2.281, 95% CI 1.145-4.543 in the 50-54 age group; aHR 2.851, 95% CI 1.311-6.203 in the < 50 age group; Fig. 3A]. Conversely, age ≥ 65 group displayed no increase in the risk of TAA failure (aHR 1.394, 95% CI 0.732-2.656 in the 65-69 age group; aHR 1.334, 95% CI 0.700-2.543 in the 70-74 age group).
BMI increase by every 1 kg/m 2 was significantly associated with 4.2% increased HR for TAA failure and the risk was increased in the severely obese group versus the normal BMI group (aHR 1.632, 95% CI 1.036-2.570; Fig. 3B). There was no decrease in the risk in the underweight group (aHR 0.964, 95% CI 0.232-4.002).

Discussion
Even though the popularity of TAA is growing for the treatment of end-stage ankle arthritis along with advances in implant design and surgical techniques, the complication and reoperation rates after primary TAA still remain high compared with those for AA 3,29 . Consequently, numerous studies have focused on identification of the risk factors for TAA failure. However, there are few large studies that have evaluated the risk factors for TAA failure. Furthermore, the reported risk factors have differed between studies 30 . We sought to conduct a nationwide population-based cohort study to identify the risk factors for implant failure after primary TAA.
Majority of the studies reported that the age at the time of initial TAA surgery was predictive of failure, with younger patients having a higher likelihood of requiring revision, although the definition of implant failure and the type of implant varied from study to study 5,18,19,31 . In a study of 684 patients (722 ankles) using the HINTE-GRA prosthesis, age less than 70 years was identified as an independent risk factor for implant failure 5 . In another study with 72 patients (77 ankles) using the STAR prosthesis, the patients who underwent component revision were significantly younger at the time of total ankle replacement (50.4 ± 10.0 years) than those without revision surgery (57.1 ± 14.5 years) 18 . A multivariate regression analysis in 115 patients based on the use of the Agility prosthesis demonstrated that patient age at the time of surgery was predictive of implant failure 19 . Conversely, in a prospective study with a database of 538 ankle replacements, four implants (INBONE I, INBONE II, STAR, and Salto-Talaris) were used and age was not associated with higher failure rates in a multivariable logistic regression analysis 30 . A retrospective study 32 of 811 HINTEGRA total ankles reported that clinical outcomes and the probability of revision surgery after TAA were comparable between young and old patients; the risk of revision surgery is not affected by age.
The impact of obesity on the outcomes after TAA is inconclusive. In a retrospective study 33 of 97 ankle replacements based on the use of four implants (Agility LP, Agility, INBONE I, and Salto-Talaris), patients were separated into a reference group with a BMI less than 30 kg/m 2 and an obese group with a BMI greater than or equal to 30  Similarly, in a study using a large database of 905 patients who underwent TAA, BMI (OR 1.04, P = 0.046) was an independent risk factor for early adverse events following TAA 31 . Conversely, in a retrospective study involving 684 patients (722 ankles) using the HINTEGRA prosthesis, BMI was not associated with failure of the prosthesis in a univariate Cox regression analysis 5 . In a prospective study with a database of 538 ankle replacements, BMI was not associated with higher failure rates in a multivariable logistic regression 30 . A retrospective study of 115 patients with the use of the Agility prosthesis reported no correlation between BMI and prosthesis failure, with the average BMI in the survival group being similar to that in the failure group (28.3 vs 29.4; P = 0.27) 19 . Despite the fact that DM negatively affects clinical outcomes and perioperative complications following TAA, a review of the literature reveals that diabetes has no correlation with implant failure following TAA 19,22,23,30 . A retrospective study involving 173 patients who underwent TAA with the HINTEGRA prosthesis reported that DM, especially when associated with poor glycemic control, negatively affects the short-to mid-term outcome www.nature.com/scientificreports/ after TAA but there was no significant difference between the non-diabetic and diabetic groups regarding the TAA revision 22 . A retrospective study of 115 patients with the use of the Agility prosthesis, also reported no significant difference between the non-diabetic and diabetic groups with regard to TAA failure 19 . Additionally, a retrospective cohort study involving 538 ankle replacements demonstrated that DM was not associated with implant failure in a univariable analysis 30 . Few studies have investigated the effect of cigarette smoking on TAA failure. In a retrospective study of 646 TAA using the INBONE I, STAR, and Salto-Talaris prostheses, active cigarette smokers had a significantly higher risk of wound complications and worse outcomes than nonsmokers and former smokers 21 . However, active www.nature.com/scientificreports/ smoking did not significantly increase the rate of revision surgery. A retrospective study of 115 patients with use of the Agility prosthesis also found no correlation between TAA failure and smoking 19 .
Our most significant finding is that, firstly, age of less than 65 years was identified as an independent risk factor for TAA failure. Additionally, in the younger age groups there was a higher aHR for TAA failure. Secondly, BMI of ≥ 30 kg/m 2 was independently associated with a statistically significant increase in the risk for TAA failure. Concerning the association between DM or cigarette smoking and TAA failure, we concluded that these factors were not associated with a higher risk of TAA failure, which is comparable to the results of previous studies 19,21,22,30 . The strengths of our study include the large sample size encompassing the entire South Korean population, longitudinal follow-up of up to 8 years, and extensive data from the regular health check-ups concerning demographics, socioeconomic factors, comorbidity, laboratory test results, and lifestyle variables.
However, the current study has several limitations. Firstly, the subjects who did not undergo a health checkup in the last two years before the index TAA operation were excluded, increasing the risk of a selection bias. Secondly, the data on health-related behaviors was reliant on self-reporting, which may be highly subjective and unreliable. Thirdly, the baseline comorbidities were defined by the health examination results or the health claim data, but not by the clinical records, and therefore could be subject to the risk of under-or over identification. However, we used the diagnosis code and medication records together, which has been shown to have higher accuracy 34 . Fourthly, all possible confounding factors, including multiple different diagnoses of ankle arthritis or the different types of implants were not considered. During the study, four implants, including HINTEGRA, Mobility, mobile-bearing Salto, and Zenith prostheses, were used in South Korea. However, we could not investigate the effect of implant type on TAA failure because information on implant type was not included in the database. Finally, the database provided by the South Korean NHIS does not include the medical records, thus we could not investigate the causes of TAA failure such as infection, aseptic loosening, osteolysis, preoperative deformity, or implant malalignment. Technical errors may also play a significant role in TAA failure. However, an analysis of technical errors was impossible in this study because we could not obtain clinical data from the database.
In conclusion, this population-based longitudinal study demonstrated that age < 65 years and BMI of ≥ 30 kg/ m 2 were associated with increased risk of TAA failure. Consequently, these factors should be considered when the surgeon and patient discuss possible surgical treatment options for end-stage ankle arthritis.  www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.