How Low Is Safe? The Frontier of Very Low (<30 mg/dL) LDL Cholesterol

Angelos D. Karagiannis; Anurag Mehta; Devinder S. Dhindsa; Salim S. Virani; Carl E. Orringer; Roger S. Blumenthal; Neil J. Stone; Laurence S. Sperling


Eur Heart J. 2021;42(22):2154-2169. 

In This Article

Safety and Tolerability of Achieving Very Low LDL-C

Nervous System

Given that cholesterol is a major membrane component of brain cells, there were initial concerns about the effect of intense LDL-C lowering on neurocognitive function. The OSLER study showed that patients receiving evolocumab/statin treatment reported statistically more neurocognitive events (delirium, cognitive and attention disorders, dementia and amnestic conditions, disturbances in thinking and perception, and mental impairment disorders) compared to patients on statin monotherapy (0.9% vs. 0.3%); the risk of neurocognitive events, though, did not significantly vary between patients achieving very low LDL-C and patients who did not.[70]

EBBINGHAUS, a sub-study of FOURIER, investigated the effect of evolocumab on neurocognitive function and found no difference in cognitive function between the evolocumab/statin vs. statin-only groups at a median of 19 months. The primary endpoint of spatial working memory strategy index of executive function was −0.21 ± 2.62 in the evolocumab vs. −0.29 ± 2.81 in the placebo group (P < 0.001 for noninferiority). Also, there was no difference between subgroups stratified based on attained LDL-C (including patients in the <25-mg/dL subgroup).[71] However, given EBBINGHAUS study's limitations including short follow-up (a median follow-up of 19 months), enrolment of relatively young patients (mean age of 63 years old), and exclusion of patients with known dementia or mild cognitive impairment, further long-term monitoring for possible memory/cognition worsening in high-risk patients is necessary. Additional data from PCSK9 inhibitor trials and sub-studies with participants achieving very low LDL-C have not shown any correlation between marked LDL-C reduction and neurocognitive impairment.[6,7,15,16,18] It is noteworthy that brain cholesterol regulation is primarily dependent upon local de novo cholesterol synthesis in the brain rather than levels of circulating plasma cholesterol.[72] Therefore, lipid-lowering agents, which decrease peripheral LDL-C, are unlikely to have significant impact on brain cholesterol levels.

Several studies have reported a possible correlation between lower LDL-C and haemorrhagic stroke incidence or associated mortality. The SPARCL study showed that patients with prior stroke treated with atorvastatin 80 mg had a higher incidence of intracranial haemorrhage compared to placebo (55[2.3%] vs. 33[1.4%]; P = 0.02).[73] In addition, a CTT meta-analysis showed evidence of a small increased rate of haemorrhagic stroke among patients treated with statins (5–10 haemorrhagic strokes per 10 000 patients in whom LDL-C is reduced by 39–77 mg/dL for 5 years); the excess risk, though, was small and much less frequent than the reduction in ischaemic strokes and was not associated with an increase in mortality.[74]

One small study of 88 consecutive patients found that those with LDL-C < 70 mg/dL had higher 90-day mortality following an episode of intracerebral haemorrhage compared to patients with >70 mg/dL.[75] Low LDL-C was correlated with elevated risk of death due to intraparenchymal haemorrhage in a Japanese population.[76] Patients with LDL-C ≥ 70 mg/dL had a lower intracerebral haemorrhage incidence compared to patients with LDL-C of 50–69 and <50 mg/dL, respectively, in a nine year follow-up cohort study {adjusted hazard ratio 1.65 (95% CI 1.32–2.05) for LDL-C 50–69 mg/dL and 2.69 (95% CI 2.03–3.57) for LDL-C < 50 mg/dL}.[77] In a prospective cohort study among women, after multivariable adjustment, women with LDL-C < 70 mg/dL had 2.17 times the risk (95% CI; 1.05, 4.48) of a haemorrhagic stroke compared to women with LDL-C 100–129.9 mg/dL over a mean of 19.3 years follow-up.[78] Although results were not significant due to limited statistical power, one meta-analysis showed that haemorrhagic stroke incidence was somewhat higher among patients with LDL-C < 50 mg/dL compared to patients with moderately low levels,[79] and another meta-analysis found that lower LDL-C concentration was associated with a higher risk of haemorrhagic stroke.[80]

Other meta-analyses, however, did not find any association between haemorrhagic stroke and statin use or lower LDL-C.[81–83] An analysis of ODYSSEY OUTCOMES results showed that alirocumab significantly decreased the risk of any stroke and the risk of ischaemic stroke without increasing the risk of haemorrhagic stroke [HR, 0.83; 95% CI, 0.42–1.65] in patients with recent acute coronary syndrome and persistent dyslipidaemia despite intensive statin therapy.[84] Importantly, the same analysis found that there was no adverse association between incidence of haemorrhagic stroke and lower attained LDL-C even in patients achieving LDL-C < 25 mg/dL, over a median of 2.8 years follow-up.[84] A recently published large cohort study found that initiation of statin in patients with prior stroke (either haemorrhagic or ischaemic) did not increase the risk of intracerebral haemorrhage in 10 years follow-up.[85] A trial of French/Korean population showed that post ischaemic stroke patients with an LDL-C goal target <70 mg/dL had no significant increase in haemorrhagic stroke compared to patients in the 90–110-mg/dL group (HR, 1.38; 95% CI, 0.68–2.82) over a median follow-up of 3.5 years.[86] Data from people with congenital low LDL-C, sub-studies of patients with very low LDL-C and prospective studies involving thousands of patients attaining low LDL-C have shown no correlation between lower LDL-C and haemorrhagic stroke.[5–7,15,19,50,84,87–90]

Given conflicting literature data, it is still unclear if lower LDL-C levels are associated with higher incidence of haemorrhagic stroke. Also, given haemorrhagic stroke risk may differ among the use of different lipid-lowering agents (e.g. statins vs. PCSK9 inhibitors) or among race (Asian population studies appear more consistent on an association between lower LDL-C and higher haemorrhagic stroke rates),[76,77] an individualized, lipid-lowering approach should be considered based on each patient's risk.

A secondary analysis of JUPITER found increased incidence of insomnia in patients with LDL-C < 30 mg/dL compared to patients with ≥30 mg/dL.[17] However, other sub-studies with patient groups achieving very low LDL-C have not supported higher rates of insomnia.[15,18]

Large prospective cohort studies have found an association of lower LDL-C levels and increased Parkinson's disease incidence,[91,92] including a cohort study with further Mendelian randomization.[93] Long-term surveillance for Parkinson's disease in patients achieving very low LDL-C might be reasonable.

Endocrine System

Statin use is associated with a modest increased incidence of new-onset diabetes.[94,95] A meta-analysis of RCTs with statins including 91 140 participants showed that treatment of 255 patients with statins for 4 years resulted in the reduction in 5.4 major coronary events and 1 extra case of diabetes.[94] Overall, cardiovascular benefit outweighs the increased risk for new-onset diabetes mellitus and clinical practice in patients with multiple cardiovascular risk factors or existing cardiovascular disease should not be modified.[94,95] Also, studies showed that statins moderately increase fasting plasma glucose (FPG)/HbA1c soon after statins initiation, but since the FPG/HbA1c values then remain stable through time, it is mostly the patients with pre-diabetes and elevated body mass index who are more prone to develop statin-induced new-onset diabetes.[96,97] Furthermore, although data from a mendelian randomization study support that variants in PCSK9 and HMGCR genes were correlated with higher diabetes risk per unit decrease in LDL-C,[46] so far, patient studies have not shown any effect of PCSK9 inhibitors on increasing the FPG or new-onset diabetes incidence. Whether new-onset diabetes incidence is increased among patients achieving very low serum LDL-C is still unclear. It is also uncertain whether new-onset diabetes is purely a statin-associated side effect and may not be associated with lowering LDL-C per se or with other lipid-lowering agents use.

In a pooled analysis of data from five statin trials, intensive-dose statin therapy was associated with a higher risk of new-onset diabetes compared with moderate dose therapy (odds ratio 1.12, 95% CI, 1.04–1.22).[98] A cross-sectional study found that individuals not on statin treatment with LDL-C < 60 mg/dL had a higher rate of type 2 diabetes compared to individuals with LDL-C levels between 90 and 130 mg/dL.[99] A sub-study of JUPITER demonstrated that rosuvastatin-treated patients attaining LDL-C < 30 mg/dL had increased risk for developing new-onset diabetes compared to those with LDL-C ≥ 30 mg/dL.[17] However, this was not related to the degree of LDL-C lowering as there was no increased risk of diabetes in patients with large (≥70) percent reductions in LDL-C.[17]

The FOURIER and ODYSSEY OUTCOMES trials, with a combined pool of 46,488 patients (many of whom achieved very low LDL-C), did not show any significant increase in new-onset diabetes rates in patients on PCSK9 inhibitor/statin combination (FOURIER; hazard ratio, 1.05; 95% CI, 0.94–1.17. ODYSSEY OUTCOMES; 9.6% in alirocumab group vs. 10.1% in placebo group).[6,7] A pre-specified analysis of FOURIER, designed to investigate the effect of evolocumab on glycaemia and risk of developing diabetes showed that evolocumab neither increased the risk of new-onset diabetes in patients without diabetes or pre-diabetes at baseline [(hazard ratio 1.05, 95% CI 0.94–1.17) and (hazard ratio 1.00, 95% CI 0.89–1.13), respectively], nor did it worsen glycaemia over a median of 2.2 years of follow-up.[100] Similarly, a prespecified analysis of ODYSSEY OUTCOMES showed that treatment with alirocumab targeting an LDL-C of 25–50 mg/dL did not increase new-onset diabetes rates in patients with pre-diabetes or normoglycaemia at baseline when compared to placebo group (HR 1.00; 95% CI 0.89–1.11) over a 2.8 median follow-up.[101] The same study demonstrated that in the alirocumab treatment group, targeting an LDL-C of 25–50 mg/dL, resulted in twice the absolute reduction in cardiovascular events among patients with diabetes (2.3%, 95% CI 0.4–4.2) compared to those without diabetes (1.2%, 95% CI 0.0–2.4), respectively.[101]

A meta-analysis of RCTs with statins and statin/PCSK9 inhibitors as interventions in 163,688 nondiabetic patients investigated the association of intense vs. less intense lipid-lowering therapy and new-onset diabetes. Among patients in groups treated with more intensive management (defined as groups receiving the most potent pharmacological strategy at each trial), there was no significant association between 1 mmol/L reduction in LDL-C and new-onset diabetes incidence (risk ratio: 0.95; 95% CI, 0.87–1.04).[102] The same meta-analysis demonstrated that patients on PCSK9 inhibitor/statin combination treatment did not have a higher risk of developing diabetes.[102] Sub-studies with patients on PCSK9 inhibitors achieving very low (<30 mg/dL) LDL-C have shown no association with higher risk for developing diabetes.[16,18]

Cholesterol is a precursor of the five major classes of steroid hormones: progestogens, glucocorticoids, mineralocorticoids, androgens, and oestrogen.[103] However, there appears to be no significant effect on cholesterol based hormone levels or function in patients achieving very low LDL-C.[18]

Gastrointestinal System

In a sub-study of JUPITER, patients with LDL-C < 30 mg/dL had significantly higher incidence of hepatobiliary diseases (all causes), driven by increased rates of unspecified biliary diseases and cholelithiasis/choledocholithiasis. The median follow-up in JUPITER, however, was only 1.9 years. Of note, four other sub-studies in patients with very low LDL-C did not find any significant correlation between very low LDL-C and higher incidence of hepatobiliary conditions.[15,16,18,19]

Robinson et al.[18] showed that patients with LDL-C < 25 mg/dL treated with alirocumab do not have increased risk for developing hepatic steatosis, a complication commonly found in patients with familial hypobetalipoproteinemia.

Ocular System

In the HOPE-3 trial, individuals in the 10-mg rosuvastatin per day group had a statistically significant greater number of cataract surgeries than in the placebo group.[104] Pooled data from 14 alirocumab trials (though not including ODYSSEY OUTCOMES' data) showed that the rate of cataracts was higher in patients with LDL-C < 25 mg/dL (2.6%) vs. ≥25 mg/dL (0.8%; hazard ratio: 3.40; 95% CI: 1.58–7.35).[18] However, other sub-studies in patients with very low LDL-C did not find any significant correlation between very low LDL-C and cataract development.[15–17,19] These include results from the FOURIER trial with its large number of subjects with LDL-C < 20 mg/dL[16] and the IMPROVE-IT trial notable for its follow-up of over 6 years.[15] In ODYSSEY OUTCOMES, the cataract incidence for the alirocumab and the placebo groups was 1.3% vs. 1.4%, respectively, which also supports that dramatic LDL-C lowering does not increase cataracts.[7] Analysing ODYSSEY OUTCOMES' data of those patients who achieved very low LDL-C may further clarify whether there is a possible association between very low LDL-C and greater cataract risk.


The SEAS study that investigated the effect of ezetimibe/simvastatin on major cardiovascular events in patients with asymptomatic aortic stenosis found slightly higher cancer (not specific type) rates in the treatment group, although not associated with the degree of LDL-C lowering.[105] However, an observational follow-up study of the SEAS treatment population demonstrated no increased cancer incidence in the 21-month period after completion of the original study (hazard ratio 0.55, 95% CI 0.27–1.11).[106] A univariate analysis of the long-term follow-up of the IMPROVE-IT study found the rate of new, worsening, or relapsing malignancies to be increase in patients from highest to lowest achieved LDL-C level, but this correlation was no longer statistically significant after adjusting for baseline variables.[15] A Mendelian randomization analysis showed lower LDL-C to be associated with higher risks of endometrial cancer of all histologies combined.[107]

The CTT meta-analysis that included multiple statin treatment trials did not find any correlation between statin therapy and any site-specific cancer or overall cancer incidence [relative risk (RR) 1.00, 95% CI 0.96–1.04].[11] There are no sub-studies where patients have attained very low LDL-C or sub-studies with long-term safety follow-up that have shown any association with higher cancer rates.[12,15–19]

Other Systems

A sub-study of JUPITER found a higher rate of physician-reported haematuria in patients with LDL-C < 30 mg/dL.[17] However, the same sub-study did not show higher haematuria rates as determined by protocol-driven laboratory measurements. All other sub-studies in patients with very low LDL-C have shown no increase in haematuria incidence.[15,16,18,19]

Robinson et al.[18] using pooled data from 14 alirocumab trials showed that patients with LDL-C < 25 mg/dL had no significant difference in levels of gonadal hormones, or fat-soluble vitamins A, D, and K and had elevated vitamin E levels when adjusted for LDL-C values.


Few observational and prospective cohort studies that suffer from confounding have examined potential association of lower LDL-C levels with an overall mortality. Lower LDL-C and total cholesterol levels, in the elderly, not on lipid-lowering agents, Chinese, Finish, and Italian population, respectively, paralleled with higher mortality.[108–110] Patients with LDL-C < 77 mg/dL hospitalized for acute myocardial infarction had higher in-hospital mortality[111] and patients with LDL-C ≤ 21 mg/dL who were admitted with severe community-acquired pneumonia requiring intensive care unit hospitalization had decreased inpatient survival rate compared to control patients with higher LDL-C levels.[112]

Figure 3 summarizes all the potential benefits and side effects associated with very low LDL-C.

Figure 3.

Benefits and potential side effects of very low (<30 mg/dL) LDL-C. An increasing number of patients are expected to attain very low LDL-C (<30 mg/dL) given the advent of potent lipid-lowering agents. The clinical impact of achieving very low LDL-C remains unclear. Figure 3 summarizes potential benefits and side effects of attaining very low LDL-C.