Different effect of hypercholesterolemia on mortality in hemodialysis patients based on coronary artery disease or myocardial infarction

In this nationwide, population based study, we demonstrated that in patients under maintenance hemodialysis, different patterns of association between TC and mortality exist. In those without MI/CAD, an increased mortality was observed when TC is higher than 250 mg/dL. In addition, patients without MI/CAD who had TC levels 250 mg/dL in the first and third years had a 64% increased risk for mortality.

Although CV mortality was significantly higher in patients under dialysis compared with the general population [14], the correlation of TC values with the major outcomes in the HD subgroups showed conflicting results. There are no studies that compared the differential influence of lipoproteins on HD patients with regard to the presence of MI or CAD. Patients with previous CAD were more likely to receive statins and other antiplatelet medication such as aspirin, which have a protective effect could be one of the confounding factors. A study by Shoji et al. including 45,390 HD patients from Japan renal registry reported that non-HDL was associated with increased risk of CVD incidents, but the mortality after new onset of CAD was not associated with dyslipidemia [7].

The positive correlation of cholesterol and mortality when TC is above 250 mg/dL in the group of HD without previous MI/CAD is similar to that in the general population. Despite that only a few individuals with total cholesterol 250 mg/dL (N = 1975 and 394 in non-MI/CAD and MI/CAD groups seperately) are on regular HD, the mortality still correlated with the higher TC levels in those without previous MI history. In the general population, statin therapy for lipid-lowering is fundamental and “the lower the better” is the strategy for many physicians. According to the current KDIGO guideline, lipid-lowering agents such as statins should be used for all CKD patients with high CVD risk [15]. However, the lipid-lowering agents are under-prescribed for CKD patients than non-CKD patients with dyslipidemia in clinical practice [16]. The results of large randomized clinical trials (4D, AURORA, SHARP) on cardiovascular benefits of LDL lowering therapy in HD patients have been inconclusive [17–20]. The benefit of lowering serum LDL cholesterol could be achieved only after 3 to 5 years of statin therapy [21]. Therefore, the initiation of statins is no longer recommended in dialysis patients [22]. In the SHARP study which excluded patients with history of coronary artery disease [19], it showed beneficial effect in reducing major atherosclerotic events by reducing LDL levels with daily simvastatin 20 mg plus ezetimibe 10 mg. In our study, we found a U-curve relation between TC and mortality in HD patients without MI or CAD from a large population dataset which indicates that cholesterol lowering therapy might be beneficial, but the goal and candidate of treatment may not be the same as the non-dialysis patients with high CV risk.

Although a large body of evidence suggests the positive correlation of TC with mortality in general population [3], this correlation does not always exist in patients with HD. Some authors proposed the reverse epidemiology concept, where lower cholesterol is associated with worsened CVD outcome by itself or maybe confounded by malnutrition or inflammation when individual is on maintenance dialysis therapy [5, 6]. Dialysis patients with hypoalbuminemia interplayed with malnutrition that led to proinflammatory condition and cytokine activation, which is suggested to have higher prevalence in cardiovascular diseases [23–25]. There are other possible explanations of reverse epidemiology other than malnutrition inflammation complex syndrome (MICS). Since these dialysis patients underwent special selection for renal replacement therapy and survived, they had survival bias and had different characteristics when compared to the general population. Hypercholesterolemia was not associated with increased mortality in HD patients with MI/CAD probably because the prevalence of comorbidities was high. Moreover, HD patients with MI/CAD tend to have shorter survival period, the harmful insult from hyperlipidemia was considered in the long term, which is not readily observed in this subgroup. The association of low cholesterol with higher mortality is also found in the group of chronic heart failure patients as in those with chronic renal failure [26]. The possible explanation proposed is that lipoproteins may potentially contribute to the reduction of endotoxin bioactivity and consequently decrease cytokine production [27–30]. It was also observed that cholesterol is a protective factor in patients with chronic disabling disease or old age [31].

The relationship of cholesterol and mortality in HD patients was the same with general population in some observational study. Liu et al. reported in a prospective study of 823 patients that in the absence of inflammation/malnutrition, a positive correlation exists between cholesterol and all-cause mortality [6]. However, in our study, the markers of malnutrition and inflammation, albumin and WBC counts, were found to be similar between MI/CAD and non-MI/CAD group. After adjusted for these parameters, TC level lower than 150 mg/dL still had an increased risk for all-cause mortality in both MI/CAD and non-MI/CAD group.

Patients in the lowest cholesterol group in the first year and the third year had the highest HR in the non-MI/CAD or MI/CAD group, although the effect was more significant in the non-MI/CAD group. This is similar to the observation in non-demanding people older than 70 years from a study of the Honolulu Heart Program, in which the same patients being in the lowest group of cholesterol from two examination periods had the worst CV outcome [31]. However, patients without previous MI/CAD who remained in the group of TC 250 mg/dL in the first three years of dialysis showed a 64% increased risk in mortality in our study gave us a hint that statin treatment maybe beneficial.

There were several limitations of this study. First, there was no detailed data regarding anti-platelet, anti-hypertension and most importantly, statin prescription which may have effect on the mortality. The lower TC level may have been due to the effect of medication. However, it was noticed that even though the majority of patients had TC level between 150 and 250 mg/dL, a substantial proportion of the population had TC level 150 mg/dL, with increased risk of mortality; this suggests that the treatment goal should be TC 250 mg/dL but not 150 mg/dL. Second, the actual cause of mortality can’t be identified from the TWRDS dataset. The cause of death of patients with higher or lower total cholesterol level maybe different in patients under maintenece dialysis and need further investigation. Third, the lack of completeness of plasma lipoprotein profile in this dataset was noted. However, we found there was no association between plasma traiglyceride and all cause mortality in this cohort (data not shown).