OUP user menu

Effects of pravastatin on coronary events in 2073 patients with low levels of both low-density lipoprotein cholesterol and high-density lipoprotein cholesterol: results from the LIPID study

David Colquhoun , Anthony Keech , David Hunt , Ian Marschner , John Simes , Paul Glasziou , Harvey White , Philip Barter , Andrew Tonkin
DOI: http://dx.doi.org/10.1016/j.ehj.2004.03.013 771-777 First published online: 1 May 2004


Aims Fibrates or nicotinic acid are usually recommended for secondary prevention of coronary heart disease in patients with low plasma levels of both low-density lipoprotein cholesterol (LDL-C) ⩽140 mg/dL (⩽3.6 mmol/L) and high-density lipoprotein cholesterol (HDL-C) ⩽40 mg/dL (⩽1.03 mmol/L). The LIPID trial, a randomised, placebo-controlled trial in 9014 patients at 87 centres in Australia and New Zealand, provided an opportunity to investigate the effects of an HMG-CoA reductase inhibitor in patients with low LDL-C and low HDL-C.

Methods and results Participants in this post hoc substudy were 2073 patients aged 31–75 years with baseline LDL-C ⩽140 mg/dL (⩽3.6 mmol/L), HDL-C ⩽40 mg/dL (⩽1.03 mmol/L), and triglyceride ⩽300 mg/dL (⩽3.4 mmol/L). The relative risk reduction with pravastatin treatment was 27% for major coronary events (95% CI 8–42%), 27% for coronary heart disease mortality (95% CI 0–47%), 21% for all-cause mortality (95% CI 0–38%), and 51% for stroke (95% CI 24–69%). The number needed to treat to prevent a major coronary event over 6 years was 22.

Conclusions Treatment with pravastatin in patients with both low LDL-C and low HDL-C significantly reduced major coronary events, stroke, and all-cause mortality. The level of HDL-C is crucial to the risk of recurrent CHD events and, consequently, the benefit of lowering LDL-C.

  • Statin
  • Fibrate
  • Low-density lipoprotein cholesterol
  • High-density lipoprotein cholesterol
  • Clinical trial substudy

See page 716 for the editorial comment on this article1


The Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) study provided convincing evidence of the benefit of pravastatin in patients with stable coronary heart disease (CHD) after an acute coronary syndrome and initial plasma cholesterol levels ranging from 155 mg/dL (4.0 mmol/L) to 271 mg/dL (7.0 mmol/L).1,2 A similar relative benefit from treatment with pravastatin was seen across major subgroups, including tertile divisions of lipid levels and in patients with a qualifying diagnosis of acute myocardial infarction (MI) or previous hospitalisation for unstable angina (UA).3

About a third of patients with an acute coronary syndrome (MI or UA) have a low-density lipoprotein cholesterol (LDL-C) level of 140 mg/dL (3.6 mmol/L) or less and a high-density lipoprotein cholesterol (HDL-C) of 40 mg/dL (1.0 mmol/L) or less.4 The joint American Heart Association/American College of Cardiology guidelines for secondary prevention recommend fibrate or nicotinic acid therapy if LDL-C is less than 129 mg/dL, accompanied by low HDL-C or elevated triglycerides.5 An expert group's consensus on the management of patients with low HDL-C recommended fibrates as first-line management.6 This expert opinion differs from the Adult Treatment Panel III guidelines, which claim that the primary target of therapy in these patients is LDL-C and therefore recommend statins as first-line therapy.7 The panel considered as options fibrates and nicotinic acid, or even combined drug therapy (low-dose statin plus a fibrate or nicotinic acid). Consistent with the differing opinions, the most recently published European Third Joint Task Force on Cardiovascular Disease executive summary states that “values of HDL-C and triglyceride should be used to guide the choice of therapy” without explaining how.8

These recommendations are based largely on the benefits observed with gemfibrozil in the Veterans Affairs High-Density Lipoprotein Intervention Trial (VA-HIT)9 and nicotinic acid in the Coronary Drug Project.10 There have been no prospective trials of 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors in patients exclusively with low levels of both LDL-C and HDL-C.

The LIPID trial is one of the largest completed trials of lipid-modifying therapy for secondary prevention in CHD patients. It provided the opportunity to assess the benefits of lowering LDL-C with an HMG-CoA reductase inhibitor in patients with low levels of both LDL-C and HDL-C, the subgroup for which gemfibrozil and nicotinic acid are usually recommended. We therefore conducted a post-hoc analysis of a subgroup of LIPID patients with both low LDL-C and low HDL-C. We used the lipid entry criteria for the VA-HIT trial as these are standard cut-points and would facilitate comparison of results with this trial.


Study design

The LIPID study has been described in detail elsewhere.1 A total of 9014 patients (7498 men and 1516 women, aged 31–75 years) were recruited at 87 centres in Australia and New Zealand between June 1990 and December 1992. Patients had had a MI or a hospital discharge diagnosis of UA 3–36 months before randomisation.

During an 8-week, single-blind, placebo run-in phase, standard dietary advice was given. Patients were then eligible for randomization if they had a total plasma cholesterol concentration of 155–271 mg/dL (4.0–7.0 mmol/L) and fasting triglycerides lower than 445 mg/dL (5.0 mmol/L). There was no entry criterion relating to HDL-C. Major exclusion criteria included a significant medical or surgical event within 3 months before study entry, significant cardiac failure, renal or hepatic disease, and current use of any lipid-modifying agents.

Patients were randomly assigned 40 mg of pravastatin or matching placebo once daily. Total plasma cholesterol concentrations were measured by the core laboratory at baseline, 6 months later, yearly after randomization, and at the end of the study. Fasting HDL-C and triglyceride concentrations were measured at baseline, 1, 3, and 5 years after randomization and at study close. Concentrations of LDL-C were estimated indirectly by the Friedewald equation.

Subgroups and outcomes

Two subgroups were studied in this analysis, one primary and one exploratory. First, patients were selected for the primary subgroup if, at the screening visit and before treatment, their HDL-C was 40 mg/dL (1.0 mmol/L) or less, LDL-C was 140 mg/dL (3.6 mmol/L) or less, and triglyceride was 300 mg/dL (3.4 mmol/L) or less. These cut-points were chosen to be consistent with definitions of low HDL-C patients and were the entry criteria for VA-HIT.9 The second subgroup had the same HDL-C and triglyceride levels but very low LDL-C, defined as LDL-C at screening of 116 mg/dL (3.0 mmol/L) or less. This stricter criterion for LDL-C was used because the distribution of baseline LDL-C in this subgroup more closely approximated the distribution of baseline LDL-C levels in VA-HIT.

The primary endpoint of the LIPID study was death from CHD. The prespecified primary endpoint for subgroup comparisons was a major CHD event, defined as CHD death or nonfatal MI. Secondary outcomes included all-cause mortality, CHD mortality, and stroke.

Statistical analysis

All analyses were by intention-to-treat and based on two-sided significance tests. Results from analysis of the primary subgroup are presented for each of the four outcomes. In addition, results are presented for mortality and CHD events in the smaller exploratory subgroup but not for stroke rates because of the small numbers of patients and events.

To facilitate comparison with results from the VA-HIT study, the relative risk reduction due to treatment and the associated confidence intervals were calculated from Cox proportional hazards models, with treatment as the single covariate.11 The adequacy of the proportional hazards assumption was inspected visually from log–log plots of survival functions. Plots of the proportion remaining event-free used Kaplan–Meier estimates. Math-values for comparison of treatment arms were based on the log-rank test.11 The number needed to treat (NNT) was calculated as the reciprocal of the absolute risk reduction.12 Lipid levels were compared between treatment groups with the Wilcoxon test.


The distribution of patients is shown in Fig. 1.

Fig. 1

Numbers of patients in the LIPID study, by LDL-C status.

Baseline characteristics

The baseline characteristics of the LIPID subgroup (Math) who satisfied the low LDL-C and low HDL-C criteria are shown in Table 1. Of the 2073 patients, 1022 were randomly assigned to receive placebo and 1051 pravastatin. The two treatment groups were well-matched, with no significant difference in baseline characteristics. We also assessed 536 patients who had a very low LDL-C (Math116 mg/dL (Math3.0 mmol/L)), 264 of whom were randomised to placebo and 272 to pravastatin. In this group, the baseline characteristics were similar to the distribution in the primary subgroup except for a lower mean total cholesterol level of 179 mg/dL (4.6 mmol/L) and an LDL-C level of 113 mg/dL (2.9 mmol/L). The mean HDL-C was 32 mg/dL (0.8 mmol/L) and the mean triglyceride level was 172 mg/dL (1.9 mmol/L), which were similar to those in the primary low LDL-C subgroup (Table 1).

View this table:
Table 1

Baseline characteristics of 2073 patients with low LDL-C and low HDL-C in the LIPID study and the subgroup of 536 patients with very low LDL-C and low HDL-C

Effects on lipid parameters

In the primary subgroup assigned to pravastatin, mean total plasma cholesterol level one year after randomisation was 155 mg/dL (4.0 mmol/L), 20% lower than the mean of 193 mg/dL (5.0 mmol/L) in those assigned to placebo (Table 2). The mean LDL-C and mean triglyceride levels in the pravastatin group were, respectively, 29% and 9% lower at one year than in the placebo group; mean HDL-C was 6% higher than in the placebo group. Differences in mean lipid parameters at one year between patients assigned pravastatin and those assigned placebo were statistically significant (Math in each case). The mean lipid changes varied little throughout the trial period and were very similar to the changes in the full LIPID cohort and in the LIPID subgroup with very low LDL-C.

View this table:
Table 2

Mean lipid levels 1 year after randomization in the LIPID study subgroups

Effects on outcomes

The effects of treatment on cardiovascular outcomes in the primary subgroup are shown in Fig. 2. For major events, the risk curves separate early, with no loss of benefit during the study period, as shown in Fig. 3. Pravastatin treatment was associated with an absolute risk reduction for a major CHD event of 4% and a relative risk reduction of 27%.

Fig. 2

Comparison of change of relative risk and 95% confidence intervals by treatment group in LIPID and LIPID subgroups. Major CHD events comprise CHD death and nonfatal myocardial infarction. On the basis of the differences in the proportion of patients who had a major CHD event over 6.1 years, treatment with pravastatin in 25 patients prevented one event. CI, confidence interval; NNT, number needed to treat to prevent one event over a median 6.1 years.

Fig. 3

Kaplan–Meier estimates of major CHD events (CHD death or nonfatal infarction). The relative reduction in risk with pravastatin therapy was derived from the Cox proportional-hazards model. The relative risk reduction at 5 years (27%, Embedded Image) was the same as at the 6-year close-out (27%, Embedded Image).

The absolute risk reduction in total mortality with pravastatin was 2.9%, with a relative risk reduction of 21% (Fig. 2). There was also a significant reduction in stroke from 5.5% to 2.8%, representing a 2.7% absolute risk reduction and 51% relative risk reduction (95% CI 24–69%). The effects did not differ between this primary subgroup and the full LIPID cohort, except for the outcome of stroke (Math, unadjusted for multiple comparisons).

The LIPID subgroup with very low LDL-C had absolute and relative risk reductions with pravastatin treatment that were consistent with those of the primary subgroup, although the sample was too small for statistically significant differences to be reliably detectable.


This analysis clearly shows that treatment with pravastatin in patients in the LIPID study with both low LDL-C and low HDL-C reduced major CHD events, stroke and, probably, mortality. Over a median period of 6.1 years, pravastatin treatment reduced major CHD events in this subgroup by 27%, a 4% absolute risk reduction (Fig. 2). These benefits in absolute and relative terms were consistent with the benefits observed in the whole LIPID study population.

This subgroup has baseline characteristics similar to the entire LIPID cohort except for a greater proportion of men and high triglyceride levels (Fig. 3). In the entire LIPID cohort, risk varied according to baseline LDL-C levels. For each mmol/L increase in LDL-C, there was a 28% higher risk of a major CHD event (CHD death or nonfatal myocardial infarction). The absolute risk of a major CHD event in the placebo group with the lowest baseline LDL tertile (Math135 mg/dL, 3.5 mmol/L) was 14%, compared with 16% in the middle and 18% in the highest tertile (Math for trend Math0.01). In this substudy with low LDL-C (⩽140 mg/dL, 3.6 mmol/L), almost identical to the lowest tertile of LDL-C of the entire LIPID cohort, the addition of low HDL increased the risk of a major CHD event up to 15.9%, the same as in the middle tertile.

In the overall LIPID cohort, a benefit from pravastatin therapy was seen across all baseline tertiles of lipid levels. Consistent with the low background risk and probable lack of statistical power, the group with the lowest third of LDL-C levels in the cohort did not independently show a statistically significant benefit from pravastatin therapy; however, there was no evidence of heterogeneity compared to other tertiles of LDL-C.

In this group with both low LDL-C and low HDL-C, pravastatin therapy was associated with a 49% relative risk reduction and a 2.7% absolute risk reduction (Math) in stroke. This is greater than we previously reported for the whole LIPID cohort, which had a relative risk reduction in stroke of 14% (Math) and in nonhaemorrhagic stroke of 23% (Math) with pravastatin treatment.13 This greater effect may be a chance finding, but low HDL-C may modify stroke risk considerably. Low HDL-C has been associated with an increased risk of death from stroke and echolucent carotid plaques, which are prone to rupture in those with low baseline LDL-C level.14,15,27

The Heart Protection Study (HPS) confirmed the benefit of further reducing low pretreatment LDL-C in patients with or at high risk of vascular disease.16 The background risk was similar in the HPS and LIPID studies. In HPS over 5 years, 14.7% of the patients taking placebo died, compared with 12.9% assigned simvastatin 40 mg/day (Math). In the LIPID study, over 6.1 years 14.1% taking placebo died, compared with 11% assigned pravastatin (Math). A subset of 6793 patients in the HPS with baseline LDL-C of 3.0 mmol/L (116 mg/dL) or less was reported. Treatment with simvastatin 40 mg/day was associated with a 17.6% rate of any vascular event compared with 22.2% in placebo-treated patients (Math). HPS did not report major CHD events in the low LDL-C group nor events in patients with both low LDL-C and low HDL-C. However, as in the LIPID study, HDL-C influenced risk in HPS, with a baseline level of less than 0.9 mmol/L (40 mg/dL) being associated with 29.9% risk of a vascular event compared with 20.9% if HDL-C was over 1.1 over five years of follow-up.

The VA-HIT study confirmed the high risk of patients with both low LDL-C and low HDL-C and the benefit of lipid-modifying treatment with gemfibrozil 1200 mg/day.9 Compared with the patients in VA-HIT, patients in this lipid subgroup were three years younger and less likely to have diabetes or hypertension; 7% were women. The benefit was similar to that found in the LIPID substudy with different lipid changes, except for the same increase in HDL-C of 6% (32 mg/dL increase to 34 mg/dL) in both studies. In the LIPID subgroup with equivalent lipid entry criteria, after one year pravastatin 40 mg/day had reduced total cholesterol by 20%, LDL-C by 29%, and triglycerides by 9%, and had increased HDL-C by 6% compared with the levels in the placebo group. In contrast, in the VA-HIT trial, gemfibrozil 1200 mg/day lowered total cholesterol by 4%, had no effect on LDL-C, lowered triglycerides by 31%, and increased HDL-C by 6%.

Gemfibrozil treatment was associated with a 17.2% rate of major CHD events in VA-HIT compared with 21.7% in placebo-treated patients (Math). Total mortality was 17.4% in placebo-treated and 15.7% in gemfibrozil-treated patients (Math). These event rates were considerably higher than in this LIPID substudy (see Fig. 2). The still narrower LIPID subgroup of patients with very low LDL-C and low HDL-C at baseline more closely resembles VA-HIT patients. In this group major CHD events were reduced from 15.9% to 11.8% by pravastatin therapy, which is virtually identical to the benefit achieved in the full subgroup. However, this 4.1% absolute risk reduction did not reach statistical significance owing to the small number of events.

Another way to compare the efficacy of treatment between trials is to match CHD patient groups categorised by absolute risk. We have previously reported a very high-risk group with a 20.2% major CHD event rate over a five-year period on placebo, which is almost identical to the placebo group in VA-HIT.17 Pravastatin reduced major CHD events by 16% over this period. The efficacy of therapy, measured by the absolute risk reduction of 4.1%, is almost identical to that seen in both the low LDL-C and very low LDL-C subgroups, and very similar to that observed in VA-HIT.

Further analysis of VA-HIT reported no benefit of gemfibrozil if baseline LDL-C was 103.5 mg/dL (2.7 mmol/L) or lower.18 This is in contrast to the HPS, which showed benefit in this group with very low LDL-C group, irrespective of HDL-C levels.

The mechanism by which low HDL-C increases CHD risk is still not fully elucidated. Low HDL-C may act directly to increase cardiovascular risk by reducing reverse cholesterol transport.19,20 Low HDL-C may also be a marker of nonlipid risk factors associated with the metabolic syndrome or act as a surrogate marker of systemic inflammation.21 HDL-C inhibits oxidative modification of LDL-C,22 and inhibits expression of the endothelial cell adhesion molecules, vascular cellular adhesion molecule 1 (VCAM-1), intercellular adhesion molecule 1 (ICAM-1), and E-selectin.23

During an acute-phase reaction, HDL-C can become proinflammatory, and preliminary data suggest it may persist in this state.24,25 Six weeks of simvastatin treatment largely reversed the effects of inflammatory-type HDL-C.26

The extent of the benefit from increasing HDL-C is unclear. A recent combined analysis of LIPID and CARE showed the greater importance of background HDL-C in modifying CHD risk when LDL-C is low.27 In the group with LDL-C under 125 mg/dL (3.2 mmol/L), an increase of 10 mg/dL (0.25 mmol/L) in HDL-C was associated with a decrease in the CHD event rate of 29%, but only a 10% decrease for those with LDL-C of 125 mg/dL or higher. The relative importance of HDL-C when LDL-C was low was greater in people with diabetes above all.27

In the LIPID Research Clinics Trial, a 1% increase in HDL-C was associated with a 0.6% reduction in CHD events independent of changes of LDL.28 In the Helsinki Heart Study, a 1% increase in HDL-C was associated with a 2–3% reduction in CHD events, again independent of LDL-C changes.29 The benefit was greatest in those with a low HDL-C in combination with triglyceride levels over 2.3 mmol/L. Therefore, the 6% (2 mg/dL) increase in HDL-C in both this study and in VA-HIT would be expected to contribute, at most, 30% of the treatment benefit; in the Sacks et al. analysis of LIPID and CARE, it contributed less than 10% of the benefit.27

Indeed, the VA-HIT investigators noted that HDL-C at baseline was not related to the development of CHD events, but the level achieved with therapy was.18 The relative risk reduction for a 0.13-mmol/L (5-mg/dL) increase in HDL-C with gemfibrozil was 11%. In the trial, HDL-C increased by 2 mg/dL; therefore, the HDL-C increase was responsible for up to 5% of risk reduction.

Interestingly, in VA-HIT event rates did not change across the quintiles of baseline triglycerides and there was no independent benefit of triglyceride reduction. The authors estimated that lipid changes with gemfibrozil explained only 23% of treatment benefit.18 Presumably, the nonlipid (pleiotropic) effects were responsible for most of gemfibrozil's treatment effect.30

However, baseline plasma triglyceride levels were not elevated in VA-HIT, mean levels being 160 mg/dL (1.82 mmol/L). Similarly, in another fibrate study, bezafibrate treatment reduced CHD events only in patients with baseline triglycerides higher than 200 mg/dL (2.3 mmol/L).31 In the LIPID study, triglycerides, even on the unadjusted analysis, did not contribute to risk (hazard ratio 1.07, Math). However, baseline plasma triglycerides were not elevated, being 142 mg/dL (1.6 mmol/L) in the group randomised to pravastatin. In this subgroup, as expected, baseline triglyceride levels were a little higher than in the full LIPID cohort (161 mg/dL [1.8 mmol/L]), almost identical to levels in VA-HIT, and below the threshold in the Bezafibrate Infarction Prevention study for treatment benefit. In contrast, in LIPID, the adjusted relative risk reduction for LDL-C per mmol was 1.28 (Math) and 0.52 for HDL-C (Math).32

In summary, although these analyses cannot be conclusive, the HDL-C changes in both VA-HIT and LIPID probably account for a small proportion of the benefit of treatment. In VA-HIT, the HDL-C level was the only lipid change related to CHD event reduction, with the benefit mainly related to nonlipid effects. In contrast, in this substudy of LIPID, CHD event reduction was mainly related to lipid modification, especially LDL-C reduction. Pleiotropic properties of pravastatin (consistent with statins in other trials) appear to have played a minor role in improving clinical outcomes.33

The results of this LIPID subgroup analysis are statistically significant and clinically compelling. In patients with low LDL-C and low HDL-C, over a six-year period the relative risk reduction in CHD death and nonfatal MI was 27% and the absolute risk reduction was 4%. The NNT to prevent a major CHD event over six years was 25 patients. Likewise, one death was prevented over a six-year period by treating 34 patients. In this high-risk group, the risk of stroke was reduced by 51% (Math) and one stroke would have been prevented by treatment of 37 patients.

Despite the impressive reduction in risk, major CHD events occurred in 11.8% on pravastatin and in 17.5% on gemfibrozil during follow-up. The high residual risk in this large group of patients warrants further investigation. There is a growing case for considering the addition of a fibrate to statin therapy (perhaps fenofibrate because of the absence of serious adverse interactions with statins) in subjects at high risk, such as those with vascular disease, diabetes, or the metabolic syndrome. The different properties of these drugs and trial results suggest an additive benefit of combined therapy or from more vigorous lowering of LDL-C and triglycerides and raising HDL-C. These hypotheses require testing in future clinical trials, as bias may be introduced in post-hoc analyses.

In conclusion, the results of this analysis are consistent with observational studies, are clinically compelling, and are supported by plausible biochemical mechanisms. The results are consistent with ATP-III guidelines and help clarify treatment options for patients with low HDL-C noted in the recent European Third Joint Task Force recommendations. Pravastatin treatment should be considered as a first-line treatment to prevent future cardiovascular events in patients with both low LDL-C and low HDL-C who have recovered from an acute coronary syndrome.


  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]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
View Abstract