Serum cystatin-C and all-cause mortality in patients with hypertrophic cardiomyopathy: a retrospective cohort study

Study population

This study was a retrospective, single-center, longitudinal cohort analysis registered at http://www.chictr.org.cn (registration ID: ChiCTR2000029352). Between December 2008 and March 2018, 538 adult patients diagnosed with HCM were enrolled from the cardiology inpatient department at West China Hospital of Sichuan University, Chengdu, China. Patient demographics, clinical data, laboratory results, and treatment records were comprehensively documented in the database. Following 2014 guidelines of the European Society of Cardiology (Elliott et al., 2014), the diagnosis of HCM was based on a wall thickness ≥ 15 mm (or ≥13 mm for patients with a family history of HCM) in at least one myocardial segment of the left ventricle as measured by echocardiography or cardiac MRI and where the thickness could not be explained solely by loading conditions.

At baseline, 39 patients were excluded for invalid data (including demographic information, clinical, laboratory, and treatment data, n = 34) or age below 18 years old (n = 5). In terms of loss to follow-up, another 43 patients were excluded. Finally, a total of 456 adult patients were included in the present analysis (Fig. 1).

The study complied with the principles of the Declaration of Helsinki and was approved by the Biomedical Research Ethics Committee, West China Hospital of Sichuan University (approval number: 2019-1147). Individual patient consent was collected. Other detailed information about those patients has been reported elsewhere (He et al., 2019; Liao et al., 2020a).

Measurement and data collection

A standard two-dimensional transthoracic echocardiography evaluation was conducted by experienced doctors in the baseline for the whole cohort during the patients’ first hospitalization from 12/2008 to 03/2018 according to the recommendations of the American Society of Echocardiography and European Association of Echocardiography (Lang et al., 2015). Venous blood specimens collected at the time of hospital admission were sent for tests (Cobas 8000 c702 module) to obtain blood biochemicals. The reference range of serum cystatin-C in electronic medical records in our hospital was 0.51−1.09 mg/L. Baseline eGFR was determined according to patients’ serum creatinine (MDRD formula and CKD-EPI formula) (Levey et al., 2006; Levey et al., 2009).

Baseline characteristics including demographic information, clinical, laboratory, and treatment data were collected from the patients’ electronic medical records in our hospital by experienced physicians. The twice-entry method was used for data entry: when the information of the two entries was consistent, the data would enter the database, otherwise it would be checked. In the present study, we picked data of basic information, medical history, family history, treatment, as well as examination including blood urea nitrogen (BUN), cystatin-C, creatinine, uric acid, triglyceride, total cholesterol, high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), left atria, left ventricular end-diastolic dimension (LVEDD), interventricular septum, left ventricular posterior wall (LVPW), and left ventricular outflow track obstruction.

Endpoint events and follow-up

In this study, all-cause mortality was defined as the endpoint. Follow-up was conducted through clinical consultations, review of medical records, and telephone interviews, spanning from the initial evaluation to either the occurrence of the endpoint or the administrative censoring date, which was set for February 7, 2020.

Statistical analysis

Baseline serum cystatin-C was divided into two groups by the median value for the clear risk stratification. For descriptive results, variables were expressed as the mean ± standard deviation (SD), median and interquartile range, or counts (n) and percentages (%) as appropriate. Baseline characteristic differences between the two groups were analyzed using independent two-sample t-tests for variables with a normal distribution and Wilcoxon rank-sum tests for non-normally distributed variables. Interactions among categorical variables were assessed using either the chi-square test or Fisher’s exact test, depending on suitability.

Cumulative incidences were presented by Kaplan–Meier curves with log-rank tests used for comparison between groups. Univariable and multivariable Cox proportional hazard regression models were constructed to assess the relationship between serum cystatin-C and all-cause mortality. To ensure parsimony of the multivariable models, a forward stepwise approach was considered to keep significant covariates showing a univariable relationship (p < 0.100) with all-cause mortality. In multivariable analyses, models 1-5 were constructed to adjust the covariables of different categories with age and gender forced to fit the models: (1) Model 1, the basic model, adjusted for age and gender; (2) Model 2 adjusted for the basic model plus medical history including chronic obstructive pulmonary disease (COPD), prior thromboembolism events, New York Heart Association (NYHA) III/IV, atrial fibrillation; (3) Model 3 adjusted for the basic model plus medicine including warfarin, beta-blocker, and diuretic; (4) Model 4 adjusted for the basic model plus laboratory measurements including BUN, estimated glomerular filtration rate (eGFR), creatinine, uric acid, triglyceride, total cholesterol, high density lipoprotein cholesterol, and low density lipoprotein cholesterol; (5) Model 5 adjusted for the basic model plus echocardiographic data including left atria, left ventricular end-diastolic diameter (LVEDD), left ventricular posterior wall (LVPW) and ejection fraction. The final model (Model 6) included all the covariates in Model 1-5. The proportional hazards assumption with the help of the cox.zph() function shipped with the survival package or Schoenfeld’s global test in R-program was tested when a Cox proportional hazard regression model was fitted. Time-dependent multivariable adjusted Cox proportional hazards models were constructed in case the effect of the covariate was time-varying with the proportional hazards assumption of the Cox regression model not holding (Zhang et al., 2018).

Variance inflation factor (VIF) values and Pearson’s correlation coefficients were used to evaluate the degree of multicollinearity among the independent variables. A high correlation of the variables was defined as a VIF of >5 (Kim, 2019). A correlation coefficient of <0.7 between two independent variables indicates no multicollinearit (Liao, Yin & Fan, 2020b). Restricted cubic spline analysis with four knots at the 5th, 35th, 65th, 95th percentiles of serum cystatin-C was used to visualize the relationships with risk of all-cause mortality after adjustment for covariates: where there was evidence of non-linearity, a two-line piecewise linear model with a single change point was estimated by trying all possible values for the change point and choosing the value with highest likelihood. Time-dependent area under the curves (AUC) at each time of the follow-up was plotted to evaluate the accuracy of serum cystatin-C in the prediction of all-cause mortality (Kamarudin, Cox & Kolamunnage-Dona, 2017). A generally accepted approach suggests that an AUC of less than 0.60 reflects poor discrimination; 0.60 to 0.75, possibly helpful discrimination; and more than 0.75, clearly useful discrimination (Alba et al., 2017). Serum creatinine and eGFR are common kidney biomarkers in routine clinical use and have independent associations with mortality (Mooney et al., 2019; Sundin et al., 2017), so we compared the predictive ability of serum cystatin-C with that both of eGFR and serum creatinine.

Subgroup analyses with interaction tests were performed to examine whether the relationship differed by sex, age, or eGFR.

For statistical analysis, R version 3.6.3 was used. All tests were two-sided, and p values less than 0.05 were regarded as statistically significant.

Baseline characteristics

The present study consists of 456 (54.61% males) adult HCM patients (Fig. 1) with an average age of 55.78 ± 15.53 years. The values of serum cystatin-C at baseline in our study varied from 0.00 to 9.49 mg/L with the median estimated at 1.01 mg/L (interquartile range: 0.89−1.17). Table 1 summarizes the baseline characteristics, comparing the two groups stratified by cystatin-C levels. Patients in the higher serum cystatin-C group were older, more likely to suffer from several comorbidities, including New York Heart Association (NYHA) III/IV, hypertension, vascular diseases, and atrial fibrillation. The uses of warfarin, statins, angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs), and diuretics were also more common in the higher serum cystatin-C group. Additionally, left atrial diameter, LVPW, BUN, serum creatinine, and uric acid were found to increase significantly in the higher serum cystatin-C group, while the eGFR and ejection fraction decreased significantly.

Outcome

During a median follow-up period of 4.67 years (range: 0.10–10.75 years), 90 patients (19.74%) died. Of these, 64 deaths were attributed to cardiovascular causes, including heart failure (n = 26), stroke (n = 11), sudden cardiac death (SCD) (n = 25), myocardial infarction (n = 1), and hypertension (n = 1). Mortality rates per 100 person-years were 2.14 (95% CI [1.31–2.97]) and 6.75 (95% CI [5.17–8.34]) in lower serum cystatin-C group and higher serum cystatin-C group, respectively (p < 0.001).

Prognostic value of serum cystatin-C

Kaplan–Meier curves show significantly increased cumulative incidence of all-cause mortality in higher serum cystatin-C group with log-rank p < 0.001 (Fig. 2). As Table 2 shows, the crude hazard ratio (HR) for all-cause mortality in the higher serum cystatin-C group was 3.08 (95%CI [1.94–4.89], p < 0.001) compared with the lower serum cystatin-C group. Baseline variables showing a univariable relationship with all-cause mortality (p < 0.100, presented in Table S1) were chosen into multivariable models. The association that higher serum cystatin-C group had significantly increased risk of all-cause mortality remained stable and consistent in Model 1-5 (all p < 0.010). The same was true in the final model (Model 6) that higher serum cystatin-C group was associated with 2.11-fold (adjusted HR: 2.11, 95% CI [1.30–3.42], p = 0.003) risk of all-cause mortality compared with lower serum cystatin-C group after adjusting potential covariates. As Fig. 3 shows, the values of time-dependent AUC of serum cystatin-C at different time points mainly fluctuated between 0.70 and 0.80, indicating an important prognostic role in predicting all-cause mortality.

A positive association between serum cystatin-C and risk of all-cause mortality could be seen after adjustment for covariates by restricted cubic spline analysis (p for non-linearity = 0.530). The median value of serum cystatin-C in the present study was 1.01, above which there was a gradual rise in risk of all-cause mortality with cystatin-C increasing (Fig. 4). While below 1.01 the association was insignificant. Based on Figs. 3 and 4, in the clinical context, baseline higher cystatin-C (for example, ≥ 1.01 mg/L) can predict a worse prognosis to a large extent at various time intervals afterward.

Additional analyses

In contrast, neither eGFR nor creatinine remained independently associated with all-cause mortality in the final model, let alone the time-dependent AUCs, both of which were assessed mainly below 0.70 and even 0.65 (Figs. S1–S3), showing less accurate predictive strengths than cystatin-C. Other covariates including NYHA III/IV, COPD, diuretic, BUN, total cholesterol, LVEDD, left atria, LVPW, and ejection fraction also showed independent associations with all-cause mortality. All the correlation coefficients between pairs of independent variables were <0.7 (Fig. S4) and the VIF values were close to 1, indicating no collinearity among the independent variables.

Subgroup analyses

Table 3 presents the analyses in different subgroups. Interaction tests indicated the relationship between cystatin-C group and all-cause mortality was not affected by sex, age, or GFR. Similar results to the main findings could be found, although in certain subgroups higher cystatin-C showed no significant difference from lower cystatin- C in all-cause mortality risk. Higher cystatin-C remained to be independently associated with all-cause mortality in female, in the patients ≥ 58 years old, and in the patients with eGFR ≥ 60 mL/min/1.73 m². The values of time-dependent AUC for all-cause mortality of the three subgroups mainly fluctuated between 0.70 and 0.80, which was consistent with the overall. Moreover, an overall ascending trend of time-dependent AUC in the subgroup consisting of patients with eGFR ≥ 60 mL/min/1.73 m² could be seen, indicating the predictive strength of serum cystatin-C was strengthening with the extension of time (Fig. S5).