Influence of cholesterol on survival after stroke: retrospective study

BMJ 1997; 314 doi: (Published 31 May 1997) Cite this as: BMJ 1997;314:1584
  1. Alexander G Dyker, lecturer in stroke medicinea,
  2. Christopher J Weir, statisticiana,
  3. Kennedy R Lees, clinical director, Acute Stroke Unita
  1. a Acute Stroke Unit, University Department of Medicine and Therapeutics, Gardiner Institute, Western Infirmary, Glasgow G11 6NT
  1. Correspondence to: Dr Dyker
  • Accepted 14 March 1997


Objective: To investigate the association between serum cholesterol concentration and cerebrovascular disease.

Design: Retrospective study.

Setting: Acute stroke unit of inner city general hospital.

Subjects: 977 patients with acute stroke.

Main outcome measures: Serum total cholesterol concentration, type of stroke investigated by computed tomography or magnetic resonance imaging, three month outcome (good (alive at home) or bad (dead or in care)), long term mortality.

Results: After adjustment for known prognostic factors, higher serum cholesterol concentrations were associated with reduced long term mortality after stroke (relative hazard 0.91 (95% confidence interval 0.84 to 0.98) per mmol/l increase in cholesterol) independently of stroke type, vascular territory and extent, age, and hyperglycaemia. Three month outcome was also influenced independently by serum cholesterol (P=0.024).

Conclusions: Our data suggest an association between poor stroke outcome and lower serum cholesterol concentration. Until a prospective controlled study has confirmed the benefits of lowering cholesterol concentration in elderly subjects, the application of cholesterol lowering guidelines cannot be justified as secondary prevention of acute stroke.

Key messages

  • Although the link between cholesterol and stroke is controversial, the balance of evidence suggests higher cholesterol is associated with an increased risk of atherothrombotic stroke but a reduced risk of intracerebral haemorrhage

  • Trials of lipid lowering drugs have concentrated on middle aged patients with a relatively low incidence of stroke events

  • We investigated the association between serum cholesterol and cerebrovascular disease in 977 patients hospitalised with acute stroke and found that higher serum total cholesterol concentrations were associated with a reduction in long term mortality after stroke

  • This relation was independent of the type or extent of the stroke, vascular territory, age, and hyperglycaemia–all factors known to influence survival independently after stroke

  • Until conclusive benefit is shown in elderly patients with cerebrovascular disease, the routine application of lipid lowering treatment after stroke cannot be justified


The association between total serum cholesterol concentration and coronary artery disease is well established, but the relation between stroke and cholesterol remains elusive. Epidemiological studies in Japanese1 2 and Japanese Americans3 failed to associate cerebral infarction with raised cholesterol but found an inverse relation with the incidence of intracerebral haemorrhage. A study of middle aged men in the United States noted a positive association between raised cholesterol concentration and ischaemic stroke and a negative association with haemorrhagic events.4

A meta-analysis of the prospective follow up of 450 000 patients within 45 separate cohorts over an average period of 16 years detected no relation between cholesterol concentration and the overall incidence of stroke. In most cohorts there was no differentiation between stroke type; thus, cerebral infarction or intracerebral haemorrhage and competing associations with infarction and a negative association with haemorrhage could not be excluded. Postmortem examination of Japanese stroke patients suggested that low serum cholesterol concentration was associated with cerebral haemorrhage while high cholesterol was linked with large carotid vessel atherothrombotic infarction. There was no association between subcortical lacunar events and raised cholesterol.5 It is therefore likely that cholesterol influences certain types of stroke, and any effect is likely to be diluted if strokes are not divided into the appropriate diagnostic grouping.

Meta-analyses of published trials of lipid lowering treatment and incidence of stroke are fraught with potential bias and pitfalls, but few trials support the hypothesis that lowering cholesterol concentration effectively reduces the incidence of stroke. In the west of Scotland coronary prevention study treatment of a middle aged population with pravastatin reduced cholesterol concentration and the primary incidence of coronary events but not stroke.6 A further secondary prevention study in patients who had survived a myocardial infarction did, however, suggest a significant reduction in stroke events (3.8% in placebo group v 2.6% in the group given pravastatin, P=0.003).7 In the Scandinavian simvastatin survival study patients given simvastatin (as secondary prevention of coronary heart disease) also had a significantly reduced number of cerebrovascular events.8

There are several reasons why a reduction in the incidence of stroke has not been consistently reported in studies of lipid lowering treatment. These large trials were designed to detect a reduction in coronary deaths in middle aged patients with and without coronary artery disease. Coronary events are 10 times more common than stroke in this age group, and as a result relatively few stroke end points were reported in each study. The few studies that have been conducted in elderly patients do, however, suggest a reduction in the incidence of stroke with lipid lowering treatment.

As yet there are no published data regarding the effect of cholesterol concentration on survival after stroke. We therefore investigated the effect of cholesterol in patients presenting with acute stroke.

Patients and methods

The admission criteria and protocol of the acute stroke unit of the Western Infirmary, Glasgow, have been described in detail elsewhere.9 Briefly, all patients within a well defined geographical region suffering a new focal or global neurological deficit are admitted, regardless of age or severity of neurological deficit. Computed tomography or magnetic resonance imaging is performed routinely within 72 hours of admission. Serum cholesterol is measured in blood samples taken from fasting patients on the morning after admission. Details of each patient's risk factors, presenting complaints, neurological examination, results of investigations, and final diagnosis are prospectively recorded and transferred to a computerised database.

The patients included in this study represent a series of consecutive admissions to this unit. Patients whose symptoms were found to be caused by a condition other than stroke were excluded from the analysis.

Outcome measures

Serum total cholesterol concentration was measured by a standard cholesterol oxidase method. Types of stroke (primary intracerebral haemorrhage or infarction) were diagnosed from the early computed tomograms or magnetic resonance images. Stroke was classified according to the system used by the Oxford Community Stroke Project.10 This describes patients as having total anterior circulation infarction, partial anterior circulation infarction, posterior circulation infarction, or lacunar infarction.

Follow up was by record linkage11 to death records from the registrar general of Scotland and to hospital discharge records to obtain information on medical events after stroke. This technique has been validated in an epidemiological study of hypertension12 and has been used for monitoring end points in a large clinical trial.13 The method of record linkage is reliable, but, admissions to private hospitals or institutions outside Scotland are not detected. Outcome is categorised as alive at home, alive in care, or dead at two, three, six, or 12 months after stroke. We used only the three month data in our study, with good outcome defined as alive at home and bad outcome as alive in care or dead. This subdivision at three months is a marker for three month functional outcome, an end point commonly used in trials of therapeutic agents in stroke.

Statistical analysis

We assessed the effect of serum cholesterol on survival using a Cox proportional hazard model, in which we controlled for known confounding factors affecting outcome–stroke type (haemorrhage v infarction); duration of symptoms (transient ischaemic attack or reversible neurological deficit v sustained stroke); Oxford classification of stroke; plasma glucose concentration; age; packed cell volume; presence of ischaemic heart disease; and smoking history. Since stroke type may not satisfy the proportionality assumption of the Cox proportional hazard model, we stratified the analysis into five groups for type of stroke and Oxford category (primary intracerebral haemorrhage, total anterior circulation infarction, partial anterior circulation infarction, posterior circulation infarction, and lacunar infarction). Age and serum cholesterol were entered as continuous independent variables. Plasma glucose did not satisfy the proportionality assumption and was entered as a dichotomous variable (≤8 mmol/l v >8 mmol/l).

Estimates of survival rates at 1000 days in example cases were provided from the Cox proportional hazard model.14 The model predictions were checked by comparison with Kaplan-Meier survivorship functions for selected subgroups of patients. Variables which predict dichotomous outcome at three months were assessed by logistic regression, using the same independent variables as above (that is, age, Oxford classification, serum cholesterol, and serum glucose). We performed the statistical analysis with Statistica for Windows Version 5.0 (StatSoft, Tulsa, OK, USA). Statistical significance was declared at P<0.05.


Of 1392 patients in the database, 26 had inadequate data for analysis because of immediate death, etc, 140 had a final diagnosis other than cerebrovascular disease, 80 could not be classified according to the Oxford categories or died before computed tomography, 97 had transient ischaemic attacks (symptoms lasting <24 hours), 43 had reversible neurological deficits (symptoms lasting 1-3 days), and in nine the duration of symptoms was not noted. Thus, complete data were available for 997 patients with acute stroke. Their mean (SD) age was 69.9 (12.5) years (range 23-95), their mean serum total cholesterol concentration was 5.93 (1.41) mmol/l, and their plasma glucose concentration was 7.3 (2.7) mmol/l.

The mean follow up of survivors was 895 days (range 105-2032), with total mortality during follow up of 39.4%. Median survival time could not be estimated, but the 25th centile was 254 days. The 1000 day survival of the 109 patients with primary intracerebral haemorrhage was 38%, of the 198 with total anterior circulation infarction it was 38%, of the 326 with partial anterior circulation infarction 63%, of the 91 with posterior circulation infarction 71%, and of the 287 with lacunar infarction 64%. Plasma glucose concentration also significantly influenced mortality: 1000 day survival was 60% in normoglycaemic subjects (concentration ≤8 mmol/l) compared with 43% in hyperglycaemic subjects (concentration >8 mmol/l). Age significantly influenced outcome, with 1000 day survival being 71% in patients aged ≤69 years and 46% for older patients. Serum cholesterol concentration did not correlate with plasma glucose (r2=0.002) or with age (r2=-0.004).

Figure 1) shows the Kaplan-Meier cumulative survival curves for patients grouped according to their cholesterol concentrations. In the Cox proportional hazard model (after adjustment for stroke type and Oxford group) higher plasma glucose, lower serum cholesterol, and greater age were independent predictors of mortality after stroke (table 1). The relative hazard was 9% lower for each 1 mmol/l rise in serum cholesterol concentration. The effect of cholesterol seemed similar for each type of stroke, though it reached significance only in the largest subgroup of patients, those with partial anterior circulation infarcts (table 2). Table 3) shows examples of the predicted 1000 day survival for patients with different stroke types, serum cholesterol concentrations, and ages.

Fig 1
Fig 1

Kaplan-Meier cumulative survival curves for patients with stroke grouped according to their serum total cholesterol concentrations (demarcation of groups by quartiles of cholesterol concentration)

Table 1

Cox proportional hazard model for mortality after stroke in 997 patients, adjusted for stroke type and Oxford classification* (χ2=124.2, df=7, P<0.00001)

View this table:
Table 2

Relative hazard for mortality after stroke in 997 patients associated with serum cholesterol concentration within each stroke type or Oxford category

View this table:
Table 3

Cox proportional hazard model's prediction of survival after stroke for 997 patients by type of stroke and Oxford classification, age, and serum total cholesterol concentration

View this table:

Logistic regression showed that, when outcome at three months was coded as good (alive at home) or bad (alive in care or dead), increasing levels of cholesterol were still associated with better outcome after adjustment for age, plasma glucose concentration, stroke type, and Oxford classification.


Our data suggest that the concentration of cholesterol independently influences survival of patients with acute stroke. Higher cholesterol concentrations were associated with improved survival. The effect seemed to be greatest in elderly patients–for example, 1000 day survival would be 51% for a 75 year old patient with a serum cholesterol concentration of 5.2 mmol/l and 59% if the cholesterol concentration was 7.5 mmol/l, an absolute risk reduction of 8% and a relative risk reduction of 16%. Cholesterol concentration did not correlate with either blood sugar or age. The prognostic effect was robust after adjustment for known prognostic indicators such as stroke type, Oxford group, blood sugar concentration, and age. Further adjustment for factors that were not significant within our model–such as smoking, alcohol intake, packed cell volume, and ischaemic heart disease–did not alter the results.

Possible bias

This counter-intuitive effect of cholesterol cannot be taken at face value without considering possible sources of bias in our study. This was a retrospective study based on prospectively collected data. We examined a hospital population and, while referrals were admitted unselectively, even from nursing homes, some patients with minor symptoms would not have been referred as inpatients. A community study would be equally likely to miss patients with severe stroke who die within 24 hours of the onset of stroke.

Serum cholesterol concentrations may reflect the severity of stroke, rather than the premorbid concentration. Our results are consistent across a range of stroke severity, however. In addition, serum cholesterol concentration did not correlate with stroke severity, as measured by the National Institutes of Health stroke scale, in a subgroup of our population (Spearman correlation coefficient -0.081, n=323, P=0.15).

Our blood samples were collected on admission or within 24 hours. We have data on serial samples collected for up to three months in a subgroup of patients, and these show no downward trend as a result of stress or poor nutrition. A recent formal study has shown that serum cholesterol measurements within the first 48 hours are identical to those after three months, although between these times a fall in concentration does occur.15 Only a community based survey that measured premorbid cholesterol concentrations could conclusively establish the effect of cholesterol on stroke outcome, but we consider that obvious sources of bias have been excluded.

Explanation of results

There is no established biological mechanism that explains these results, but cholesterol is known to have effects on the vasculature and is essential for normal membrane fluidity. High blood cholesterol concentrations modify the action of platelets such that exposure to low density lipoprotein cholesterol enhances platelet aggregation by its action on platelet activating factor.16 Rabbits fed a high cholesterol diet have larger experimentally induced infarcts associated with an increase in platelet deposition in the thrombus at the infarct.17 Exposure to high levels of cholesterol also reduces the responsiveness of large blood vessels, though not small vessels, to vasodilatory stimuli.18 All of these effects suggest that a higher serum cholesterol concentration would predispose to a poorer outcome from a stroke event.

High cholesterol concentrations may exhibit a neuroprotective effect by modulating the action of the enzyme γ-glutamyltransferase and acetylcholinesterase: a high cholesterol diet results in increased γ-glutamyltransferase activity but reduced acetylcholinesterase activity.19 γ-Glutamyltransferase has a role in amino acid uptake and transport; thus, its increase in patients with higher serum cholesterol could reduce the neurotoxic effects of excitotoxic amino acids. Early survival does not, however, seem to be affected by cholesterol: rather it seems that the difference in the survival curves of stroke patients with high and low cholesterol concentrations gradually increases over time. This suggests that the protective mechanism may have a more prolonged effect.

Cholesterol therefore seems to be a marker for long term rather than short term survival. Why this should be so is unclear as all conventional rationales suggest that patients with higher cholesterol concentrations would have an increased risk of coincidental cardiac disease, of subsequent sudden cardiac death, and of larger cerebral infarctions. A low cholesterol concentration is, however, known to be associated with underlying serious illness, and it might be expected that such patients would have a poorer outcome than those with higher cholesterol concentrations. A retrospective analysis of our data dividing patients into five groups according to cholesterol concentration confirmed that both early and long term mortality were greater in the groups with lowest cholesterol concentrations. It is possible that lower cholesterol in these relatively elderly patients (mean age 69.9 (SD 12.4)) may simply have reflected poor nutritional status, which could predispose to a poor outcome after stroke.


Trials of lipid lowering drugs have concentrated on middle aged patients with a low risk of stroke, and as a result it is unknown whether lipid lowering treatment is beneficial in older patients with elevated cholesterol concentration and cerebrovascular disease. Epidemiological data suffer from failure to subclassify stroke into different types, which masks any positive relation between infarction and cholesterol and a negative relation with cerebral haemorrhage. Postmortem examination of stroke patients also suggests that high cholesterol concentrations may increase the risk of only certain types of infarct, such as large vessel atherothrombosis, but not lacunar events.

A prospective study of the incidence and outcome of stroke in relation to blood cholesterol in elderly patients (with appropriate computed tomography and clinical diagnosis) is required to define this relation accurately. A randomised controlled trial of lipid lowering drugs in elderly patients would evaluate the safety and cost effectiveness of lowering cholesterol and could reveal effectiveness in terms of reducing the incidence of both stroke and coronary events. Current practice recommends reducing blood cholesterol in patients aged over 55 with a high risk of vascular disease despite the lack of evidence for benefit in patients over 70 years old. As all trials of lipid lowering drugs have studied middle aged patients their results cannot be extrapolated to elderly people.

Our data suggest that lower serum cholesterol concentrations may have an independent adverse effect on survival after stroke. Further studies to investigate and confirm this relation are required. The efficacy of cholesterol lowering drugs as primary or secondary prevention of coronary and cerebrovascular events in elderly patients remains unproved.


We thank Dr G T McInnes, Professor J L Reid, and Dr P F Semple for permission to study their patients. We thank Chris Povey of Scottish Record Linkage, NHS Information and Statistics Division, Edinburgh, for record linkage to death and hospital discharge records.

Funding: None.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
View Abstract