Relation between lymphocyte to monocyte ratio and survival in patients with hypertrophic cardiomyopathy: a retrospective cohort study

Study population

This is a retrospective study where all HCM patients greater than 18 years of age who have been treated at Sichuan University West China Hospital, which is a tertiary referral center in Southwest of China were consecutively enrolled, forming the inception cohort of 508 patients. The enrollment started at Dec 2008 and ended at May 2016. The diagnosis of HCM was based on the presence of increased left ventricular wall thickness (≥15 mm), identified by echocardiography, cardiac magnetic resonance or computed tomography, that was not solely explained by abnormal loading conditions (Zamorano et al., 2014). Patients with other inherited metabolic diseases or syndromic causes of HCM (5 cardiac amyloidosis, 2 restrictive cardiomyopathy, 1 dilated cardiomyopathy and 1 myocarditis) were excluded from the study. For the present study, other exclusion criteria were as follows: (1) patients with comorbidities which would affect the white blood cell count (WBCC) were excluded (n = 91), such as infections, hematological system diseases and taking corticosteroids; (2) patients with age under 18 (n = 5) and patients with incomplete data (n = 26) were further excluded from the study; (3)We also excluded patients who were lost to follow-up after the first evaluation (n = 23). Finally, the present study included 354 adult HCM patients for the final analysis.

This study was approved by the Biomedical Research Ethics Committee, West China Hospital of Sichuan University (approval number: 2019-1147). This study was performed in keeping with the criteria set by Declaration of Helsinki. Due to the retrospective nature of the study, informed consent was waived. Other detailed information has been reported in the recently published studies (Wang et al., 2020).

Baseline clinical evaluation

Complete blood count (CBC) was measured by a Sysmex XN-9000 analyzer (Sysmex Corporation, Kobe, Japan) at the time of hospital admission, and the laboratory parameters of CBC included number of WBCC and percentage of lymphocyte and monocyte, and so on. The normal range for WBCC was 4–10*10⁹/L in this system. The LMR was calculated based on the results of CBC. All patients underwent 2D transthoracic echocardiography examinations by standard techniques at baseline, and the data and other detailed information of baseline characteristics were collected from electronic medical records.

Study endpoint

The endpoint of the study was all-cause mortality. Follow-up was carried out by medical records review or telephone contacting with patients or their family members. All patients were followed until the date of death or the end of follow-up at Feb 2020, whichever came first.

Statistical analysis

To quantify in a simple form the association between LMR and all-cause mortality, the patients were divided into three groups according to baseline LMR: tertile 1 (<3.8), tertile 2 (3.8–5.4) and tertile 3 (≥5.5) (Wang et al., 2017; Kiris et al., 2017; Mabikwa et al., 2017). For each group, continuous variables were presented as mean ± standard deviation (SD) and median with interquartile range (IQR) where appropriate, and categorical variables as number (percentage). The baseline characteristics among the three groups were analyzed by using the analysis of variance or Kruskal-Wallis tests for continuous variables, and the chi-square or Fisher exact tests for categorical variables.

The cumulative mortality in each group was calculated by Kaplan–Meier method, and the log-rank test was used for comparison. Cox proportional hazard regression analysis was used to assess the role of LMR as an independent predictor of mortality. Variables for inclusion were carefully chosen to ensure parsimony of the multivariate models. Age and gender were forced into the multivariate models, and baseline variables that showed a univariate relationship with mortality (p < 0.05) or that were considered clinically relevant (Liu et al., 2017) were included for adjustment in multivariate models. Additionally, we explored the association between LMR and all-cause mortality by smooth curve fitting after adjustment for potential confounders, and a two-piecewise Cox model was also applied to examine the threshold effect of LMR on mortality. Additionally, a time-dependent receiver operating characteristic (ROC) curve was generated to evaluate the accuracy of LMR in the discrimination of mortality at different prediction times (Kamarudin, Cox & Kolamunnage-Dona, 2017). A generally accepted approach suggests that the area under curves (AUCs) of less than 0.60 reflects poor discrimination; 0.60 to 0.75, possibly helpful discrimination; and more than 0.75, clearly useful discrimination.

The robustness of these findings was assessed in multiple sensitivity analyses. First, stratified analysis assessed the consistency of association between LMR and mortality in different groups. Second, we explored the potential for unmeasured confounding between LMR and all-cause mortality by calculating E-values (Haneuse, VanderWeele & Arterburn, 2019). The E-value quantifies the required magnitude of an unmeasured confounder that could negate the observed association between LMR and all-cause mortality, and the E-value was calculated on the web calculator (https://www.evalue-calculator.com/). Third, we also assessed the relation between LMR and mortality in the subgroup with a normal WBCC range as another sensitivity analysis.

All analyses were performed with R version 3.6.3 including the “compareGroups”, “rms”, “survminer”, “tidyverse”, “survival” “timeROC” and “base” packages (http://www.R-project.org). All tests were two sided, and p values <0.05 were considered statistically significant.

Baseline characteristics

Table 1 shows the baseline characteristics of the study cohort. The average age was 56 (IQR: 44.0–66.0) and male patients accounted for 55.1%. The LMR distribution ranged from 0.2–17.4. There were 115 patients in title 1, 116 patients in tertile 2 and 123 patients in tertile 3. The lymphocytes increased significantly across the LMR tertiles. On the contrary, WBCC, neutrophils and monocytes decreased significantly across the LMR tertiles. Besides, patients with higher LMR values tended to be younger and have significantly lower left atria diameter, as well as significantly higher hemoglobin concentration and left ventricular ejection fraction. Other clinical features did not show significant difference among the three groups.

Survival analysis

During the follow-up, 44 patients reached the endpoint of all-cause mortality. The specific causes of deaths were as follows: 7 sudden cardiac deaths, 12 heart failure-related deaths, 9 stroke-related deaths, 2 HCM-related postoperative deaths, 1 other cardiac death and 9 non-cardiac deaths. The cause of death could not be confirmed in 4 patients. The cumulative incidence of all-cause mortality was 24.8% for the whole cohort.

Based on the baseline LMR, there were 23 all-cause mortality in the first terile, 12 in the second tertile and 9 in the third tertile. The corresponding mortality rates were 7.3 (95% CI [4.4–10.1]), 2.2 (95% CI [1.0–3.5]) and 1.0 (95% CI [0.3–1.6]) per 100 person years, respectively. Kaplan–Meier analysis also demonstrated gradually decreased cumulative incidence of all-cause mortality across the three tertiles with log-rank p = 0.004 (Fig. 1).

Relation of LMR to all-cause mortality

Univariate Cox regression analysis revealed that dyspnea, atrial fibrillation, age and left atria diameter were significant risk factors of all-cause mortality, while hemoglobin, lymphocyte, triglycerides, low-density lipoprotein cholesterol were significant protective factors (Table 2). In comparison of the top tertile versus the bottom tertile of LMR, the crude hazard ratio (HR) of all-cause mortality was 0.32 (95% CI [0.15–0.69]; p = 0.004). Two models were constructed to examine the effect of comorbidities, laboratory parameters, echocardiographic parameters and clinically relevant prognostic factors on the association between LMR and all-cause mortality risk. The association did not change materially after adjusting potential confounding factors. The fully adjusted HRs of all-cause mortality were 0.43 (95% CI [0.20–0.91], p = 0.027) for patients in tertile 2 versus tertile 1 and 0.39 (95% CI [0.17–0.90], p = 0.028) for patients in tertile 3 versus tertile 1, respectively (Table 3).

Two-piecewise analysis and time-dependent AUCs

Due to the similar all-cause mortality rate in tertile 2 and tertile 3, we speculated that there might be a nonlinear association between LMR and all-cause mortality. Therefore, we further assessed the association between LMR and all-cause mortality by smooth curve fitting after adjustment for potential confounders, and a threshold effect was observed with a two-stage change and an inflection point. For LMR < 6.5, per SD increase resulted in a significantly decreased risk of all-cause mortality by 62% (adjusted HR: 0.38, 95% CI [0.21–0.68]), and the all-cause mortality risk did not progressively increase for the patients with a LMR ≥ 6.5 (Fig. 2A).

The time-dependent AUCs shown in Fig. 3A were developed on the basis of time-dependent ROC curves, which were calculated every four months. The AUCs indicated that LMR had a helpful discrimination for all-cause mortality in the whole cohort, with AUC ranged from 0.594 to 0.703 (Fig. 3A).

Sensitivity analysis

Stratified analysis suggested the consistency of association between LMR and all-cause mortality remained in all subgroups, including age (<60 and ≥60 years), gender (male and female), AF (yes and no), beta-blocker (yes and no), hemoglobin (<139 and ≥139 g/L), WBCC (<6.3 and ≥6.3 *10⁹/L), left atria diameter (<40 and ≥40 mm), left ventricular ejection faction (<50 and ≥50%) and left ventricular outflow tract gradient (yes and no), and there were no interactions between LMR and the above-mentioned variables.

Furthermore, we generated an E-value to assess the sensitivity to unmeasured confounding in the whole cohort. Based on the model 4 in the Table 3, the E-value was 2.97 for the tertile 2 of LMR, and the upper limit of the confidence interval was 1.34. For the tertile 3, the E-Value was HR = 3.22 (upper limit 1.36). For per SD increase of LMR (<6.5), the E-values for the point estimate and upper confidence bound for all-cause mortality were 3.29 and 1.94, respectively. Combined with the present multivariate results (Table S1) and the published data (Park et al., 2018), it is not likely that an unmeasured or unknown confounder would have a substantially greater effect on all-cause mortality than these known risk factors by having a HR exceeding about 3.0.

Finally, another sensitivity analysis including patients with normal WBCC (4–10*10⁹/L) revealed similar results with the main analyses. Kaplan–Meier curve showed a gradually higher mortality risk with lower LMR (log-rank = 0.002). The adjusted HRs were significantly higher in the tertile 2 (HR = 0.39, 95% CI [0.18–0.86], p = 0.019), and in the tertile 3 (HR = 0.31, 95% CI [0.12–0.77], p = 0.012) compared with the tertile 1. In addition, adjusted HR of all-cause mortality was 0.35 (95% CI [0.18–0.67]) for per SD increase in the patients with a LMR <6.5, and the all-cause mortality risk did not significantly increase when LMR ≥ 6.5 (Fig. 2B). Figure 3B shows that LMR had a helpful discrimination for mortality, with AUC ranged from 0.617 to 0.720.