Published in The British Medical Journal - 12th June 2016

Study selection

Our search gave 2894 hits. We excluded 160 studies, which were not in English, and 2452 studies because, judged from the abstract, it was obvious that they were irrelevant.

The rest of the papers were read in full; 263 of these studies were excluded for the following reasons: (1) the participants did not represent the general population; (2) LDL-C was not measured at the start; (3) follow-up information was not given for the elderly separately; or (4) no information was present about mortality during the observation period (figure 1). One of the studies6 was excluded because it included the same individuals as in a previous study.7

Figure 1

Flow Chart. CV, cardiovascular; LDL-C, low-density lipoprotein cholesterol.

Study characteristics

The remaining 19 studies including 30 cohorts with a total of 68 094 participants met the inclusion criteria (figure 1). All-cause mortality was recorded in 28 cohorts. In 16 of these cohorts (representing 92% of the individuals), the association was inverse and with statistical significance in 14; in 1 of the cohorts, the association was mirror-J-formed with the lowest risk in the highest quartile; in the rest of the papers, no association was found. CV mortality was recorded in nine cohorts; in one of them, the association was almost U-shaped with the lowest risk in the highest quartile (curvilinear fit: p=0.001); in one of them, the association was mirror-J-formed and also with the lowest risk in the highest quartile (curvilinear fit: p=0.03); in the other seven cohorts, no association was found (table 1).

Table 1

Association between LDL-C and all-cause mortality and CVD mortality, respectively, in 19 studies including 30 cohorts with 68 094 individuals from the general population above the age of 60 years

Risk of bias across studies

One explanation for the increased risk of mortality among people with low cholesterol is that serious diseases may lower cholesterol soon before death occurs. Evidence to support this hypothesis may be obtained from 10 of the studies in which no exclusions were made for individuals with terminal illnesses. However, in four of the studies, participants with a terminal illness or who had died during the first observation year were excluded. In one of those studies,8 LDL-C was not associated with all-cause mortality; in the three others,16 ,20 ,24 which included more than 70% of the total number of participants in our review, LDL-C was inversely associated with all-cause mortality and with statistical significance. Thus, there is little support for the hypothesis that our analysis is biased by end of life changes in LDL-C levels.

It is also potentially relevant that all studies did not correct for the same risk factors, and some of them did not inform the reader about which risk factors they corrected for. However, taking all studies together, 50 different risk factors were corrected for in the Cox analyses (table 2).

Table 2

Factors corrected for in the multifactorial analyses of each study

It is worth considering that some of the participants with high LDL-C may have started statin treatment during the observation period. Such treatment may have increased the lifespan for the group with high LDL-C. However, any beneficial effects of statins on mortality would have been minimal because most statin trials have had little effect on CVD and all-cause mortality, with a maximum reduction of mortality of two percentage points. It is therefore relevant that the 4-year mortality among those with the highest LDL-C in the included cohorts was up to 36% lower than among those with the lowest LDL-C. Furthermore, in the largest study20 that included about two-thirds of the total number of participants in our study, the risk was lower among those with the highest LDL-C than among those on statin treatment.

It is also possible that those with the highest LDL-C were put on a different diet than those with low LDL-C. However, this potential bias in mortality outcomes could have gone in both directions. Some of the individuals with high LDL-C may have followed the official dietary guidelines and exchanged saturated fat with vegetable oils rich in linoleic acid. In a recent study, the authors reported that among participants who were older than 65 at baseline, a 30 mg/dL decrease in serum cholesterol was associated with a higher risk of death (HR 1.35, 95% CI 1.18 to 1.54).26 If applied to the general population, this finding suggests that the conventional dietary treatment for high cholesterol with vegetable oil replacing saturated fat may actually increase mortality in those individuals with high LDL-C. Thus, the lack of an association between LDL-C and mortality may have been even stronger than reported since the dietary intervention may have been counterproductive.

Finally, it is potentially relevant that we limited our literature search to PubMed. In preliminary searches with PubMed, OVID and EMBASE, we identified 17 relevant studies in PubMed, but only 2 in OVID and EMBASE, and these 2 studies were found in PubMed as well. Therefore, it is highly unlikely that there are studies with findings with divergent results from those we have reported here, as all of them reported either no association or an inverse association between LDL-C and mortality.

back