Cholesterol variability: a marker for increased risk or a risk factor?

This editorial refers to ‘Cholesterol variability and the risk of mortality, myocardial infarction, and stroke: a nationwide population-based study’†, by M.K. Kim et al., on page 3560. Variability is the law of life, and as no two faces are the same, so no two bodies are alike, and no two individuals react alike and behave alike under the abnormal conditions which we know as disease Sir William Osler Intra-individual variability in haemodynamic parameters [such as blood pressure (BP)] is a sign of life, since BP varies from heartbeat to heartbeat, from morning to evening, from winter to summer, from sleeping to awake, from sitting to standing, with physical activity and emotional stimuli. Despite this physiological variability, a number of epidemiological studies have shown that excess short- and long-term visit-to-visit variability in BP (systolic or diastolic) portends worsening prognosis. Studies have shown that greater short-term BP variability is associated with greater cardiac and vascular damage, with an increase in left ventricular mass1 and carotid intima medial thickness.2 In those with hypertension, greater long-term BP variability is associated with increased risk of coronary artery disease (CAD), stroke, and death, independent of achieved BP.3 , 4 Similarly, in those with CAD, greater long-term BP variability is associated with increased risk of a coronary event, cardiovascular event, myocardial infarct, stroke, and death.5 Even in a general cohort without pre-existing cardiovascular disease (CVD) and largely not on antihypertensive agents, greater long-term BP variability has been shown to contribute to the risk of death from CVD.6 Similarly to BP, intra-individual variability in metabolic parameters (such as blood glucose or lipid parameters) can also be considered a sign of life, since these parameters vary from meal to meal and change with activity and emotional stimuli. Despite this physiological variability, studies have shown that excess long term visit-to-visit variability in cholesterol portends worsening prognosis, but, unlike the data on BP variability, the data on cholesterol variability and prognosis are limited. In the Treating-to-New-Targets (TNT) trial in patients with known CAD, greater LDL-cholesterol (LDC-C) variability was a strong predictor of adverse outcomes including coronary events, myocardial infarction (MI), stroke, and death, and this was independent of statin dose and achieved LDL-C levels.7 In patients with previous MI enrolled in the IDEAL trial, we showed that greater LDL-C variability was a strong predictor of adverse outcomes including coronary events, MI, stroke, and death, and this was independent of statin dose and achieved LDL-C levels.5 In both the above studies, LDL-C variability was less with high-intensity statins when compared with standard intensity statins. However, whether variability in cholesterol impacts prognosis in a general cohort without pre-existing CAD or MI is not known. In this issue of the journal, Kim et al. evaluated total cholesterol variability and the risk of events in >3.6 million subjects without prior MI or stroke.8 They found that high variability in total cholesterol was associated with a significant increase in death, MI, and stroke which was independent of mean total cholesterol levels and the use of lipid-lowering agents. The relationship was seen consistently even in subjects not on lipid-lowering therapy and in a number of sensitivity and subgroup analyses. The strengths of the analyses are the large number of patients included and the extensive number of sensitivity and subgroup analyses conducted. However, there are a number of limitations as appropriately acknowledged by the authors. The authors concluded that these findings suggest that lipid variability is an important ‘risk factor’ in the general population. Before we accept the authors’ conclusion, it is important to evaluate whether variability parameters are ‘causal’ or are merely a marker of increased risk. In 1965, Sir Austin Bradford Hill, an English epidemiologist, established the Hill’s criteria for causation.9 These are a group of nine principles that can be useful in establishing epidemiological evidence of a causal relationship, although this has been debated. The below section discusses lipid variability and the association with worse outcomes in the context of Hill’s criteria for causation (Figure 1). Hill’s criteria of causality applied to the relationship between cholesterol variability and adverse outcomes. Criteria in green circles indicates general support of the criteria and red indicates criteria not fulfilled by the present data. Strength of association. A number of studies (as described above) have shown strong correlation between variability measures and the risk of adverse cardiovascular outcomes and death. In the current study, when compared with patients in the lowest quartiles of cholesterol variability, those in the highest quartile had a 26% increased risk of death, 8% increased risk of MI, and 11% increased risk of stroke, thus showing a modest to strong effect size for this outcome. Consistency (repeated observation). The above studies all showed a consistent increase in the risk of outcomes with greater cholesterol variability—an observation seen in different cohorts, by different people across different countries. It is not known, however, if there could be a potential publication bias, with negative studies unlikely to be published. Specificity. The criterion of specificity is also fulfilled based on the published studies, as increased variability has been linked with worse cardiovascular outcomes and death. Temporality (exposure precedes effect). Sensitivity analyses in the current study and other studies have shown similar association of worse outcomes when variability parameters are calculated for a period of time and outcomes evaluated after the period of time, thus suggesting that the effect does indeed occur after the exposure. Biological gradient (dose–response relationship). As shown in this study and other studies, there was a graded and linear relationship between increased variability and worse outcomes, suggesting a dose–response relationship. Plausibility. Although we and others have suggested plausible mechanisms of increased risk with increased variability, none of them has been proven currently. Coherence. Coherence between epidemiological and laboratory findings increases the likelihood of an effect. However, there are no such laboratory findings (for example in animal models) where such increased cholesterol variability increases the risk of adverse outcomes. Experiment. There is no randomized trial that has tested the effect of increased cholesterol variability on outcomes. In fact, it may indeed not be possible to design such randomized trials, as there are currently no therapies or mechanisms to control (increase or decrease) cholesterol variability without affecting mean cholesterol values. Analogy with other causal relationships. The relationship seen between cholesterol variability and outcomes is analogous to that seen with BP. However, the ‘causal’ relationship between BP variability and outcomes itself remains unproven. As seen above, although the relationship between cholesterol variability and outcomes does satisfy many of the Hill’s criteria, this is far from perfect. A number of unanswered questions remain. How should one measure cholesterol variability? What is an accepted ‘normal’ value for cholesterol variability? What is the exact causal link between increased cholesterol variability and outcomes? The studies have variable length and frequency of cholesterol measurements. As such, these parameters are difficult to use in clinical practice. Moreover, the clinical implications are not known. At present, increased variability may represent potential medication non-compliance and, if not, perhaps consideration should be given to use a high-intensity statin given data showing reduced cholesterol variability. However, it is not known if reducing cholesterol variability will indeed impact prognosis. Before any of these are incorporated into clinical practice, more studies are needed to test whether this is ‘causal’ or merely an epidemiological association. Conflict of interest: none declared.