Author + information
- Received April 1, 1998
- Revision received September 18, 1998
- Accepted December 4, 1998
- Published online March 15, 1999.
- ↵*Reprint requests and correspondence: Dr. Stuart M. Cobbe, Department of Medical Cardiology, Royal Infirmary, 10 Alexandra Parade, Glasgow G31 2ER, United Kingdom.
The purpose of the study was to assess the effect of lipid reduction with pravastatin on hospital admissions in middle-aged men with hypercholesterolemia in the West of Scotland Coronary Prevention Study.
A prospective, randomized controlled trial was undertaken in primary care centers in the West of Scotland.
A total of 6,595 participants randomized to receive pravastatin 40 mg or placebo daily were followed up for a mean of 4.9 years (range 3.5 to 6.1 years). Analysis of hospital admissions was undertaken according to the “intention to treat” principle both for cardiovascular diseases and noncardiovascular diseases (including malignant neoplasms, psychiatric diagnoses, trauma and other causes). A secondary analysis of hospitalization in patients who were ≥75% compliant was performed.
During the trial, 2,198 (33%) of the 6,595 men were admitted to hospital on 4,333 occasions, of which 1,234 (28%) were for cardiovascular causes. Pravastatin reduced the number of subjects requiring hospital admission for cardiovascular causes by 21% (95% CI [confidence interval] 9 to 31, p = 0.0008) overall, and by 27% (95% CI 15 to 38) in compliant participants. The number of admissions per 1,000 subject-years for cardiovascular disease was reduced by 10.8 (95% CI 4 to 17.4, p = 0.0013) in all subjects, and by 15.6 (95% CI 8.3 to 23, p < 0.0001) in compliant participants. Pravastatin had no significant influence on hospital admission for any noncardiovascular diagnostic category. There were 13.4 fewer admissions per 1,000 subject-years for all causes in the pravastatin-treated group (95% CI −0.4 to 27.3, p = 0.076). No significant difference in duration of hospital stay was found between the pravastatin and placebo patients in any diagnostic group.
Pravastatin therapy reduced the burden of hospital admissions for cardiovascular disease, without any adverse effect on noncardiovascular hospitalization.
The West of Scotland Coronary Prevention Study (WOSCOPS) (1)demonstrated the benefit of pravastatin therapy in the prevention of coronary heart disease (CHD) events in middle-aged men with hypercholesterolemia within prior myocardial infarction (MI). This result has helped to end the long-running controversy over the safety and efficacy of reducing plasma cholesterol levels in the primary prevention of coronary events (2,3). Meta-analysis of earlier studies using cholesterol-lowering diets, fibrates or cholestyramine had raised the possibility that the reduction in CHD deaths achieved by cholesterol lowering was balanced by an increase in noncardiovascular deaths, particularly due to cancers, accidents, suicide or violence (3–5). In contrast, the West of Scotland study showed no evidence of excess noncardiovascular deaths either overall or from any specific cause. This observation is reinforced by the absence of any increase in noncardiovascular deaths in secondary prevention trials using pravastatin or simvastatin (6,7). In addition to assessment of mortality rates, analysis of nonfatal events is also necessary in determining the overall benefits and safety of lipid-lowering drug therapy. The objective of the current analysis was to study the effects of pravastatin therapy on morbidity as expressed by the requirement for and duration of hospital admission in the WOSCOPS.
The design, baseline characteristics and principal results of the WOSCOPS have been reported elsewhere (1,8,9). In brief, 6,595 men ages 45 to 64 years with low-density lipoprotein levels in the range 4 to 6 mmol/liter (155 to 232 mg/dl) despite dietary therapy were recruited by population screening. The principal exclusion criteria were previous MI or coronary revascularization, angina pectoris requiring hospitalization within the previous 12 months and life-threatening noncardiac illness. Participants with major electrocardiographic (ECG) abnormalities such as Q-waves, major ST-T abnormalities or left bundle branch block (Minnesota Codes 1-1 to 1-3, 4-1, 5-1 or 7-1-1) were excluded, but minor ST-T-wave changes (Codes 4-2, 4-3, 5-2, 5-3) were permitted. The presence of self-reported angina pectoris or intermittent claudication as determined by the Rose questionnaire was permitted, subject to the exclusions mentioned.
Participants were randomized to receive either pravastatin 40 mg or a placebo each evening. They were followed up every three months for a mean of 4.9 years (range 3.5 to 6.1 years), giving a total of 32,216 subject-years of observation. The dates of any hospital admission and discharge were obtained, with a copy of the hospital discharge summary, which documents the primary and secondary reasons for hospitalization.
Diagnoses were coded according to the International Classification of Diseases 8th Revision (ICD 8) and procedures according to the Office of Population Censuses and Surveys Revision 4 Coding (OPCS 4). In addition to direct reports of hospitalization from the trial participants, the occurrence of hospital admission was verified from central records using the Scottish Record Linkage System (10). We previously validated the accuracy of this system compared with hospital admission data obtained by direct patient enquiry (11). By means of direct contact and record linkage, we were able to obtain details of hospitalization both for subjects remaining in the trial and for those who had withdrawn from trial therapy and follow-up attendance. The study protocol was approved by the local research ethics committees in the West of Scotland.
The primary analysis of hospital admissions was undertaken according to the “intention to treat” principle. Among cardiovascular causes for hospitalization, we determined separately admissions for atherosclerotic CHD, nonatherosclerotic heart disease, cerebrovascular disease and other vascular disease. Noncardiovascular hospital admissions were analyzed under the trial categories of malignancy (including lymphoma, myeloma and malignant melanoma but excluding minor skin cancers), psychiatric diagnoses, trauma and other causes. An individual may have experienced one or more hospital admissions for diagnoses within a given category, as well as admissions for problems in one or more categories. We also performed a more detailed cause-specific analysis of noncardiovascular hospital admissions based on the ICD 8 body systems classification.
We previously reported the influence of compliance on the reduction in coronary risk in subjects receiving pravastatin (12). A secondary analysis of the hospitalization data was performed for compliant patients, defined as in the previous report as those attending and being issued with trial medication at ≥75% of potential trial visits.
The intervals from randomization to first hospital admission for both cardiovascular and noncardiovascular categories were compared among treatment groups using the log-rank test and displayed as Kaplan-Meier plots. Comparisons among treatment groups of the proportion of subjects hospitalized were made using chi-square statistics and corresponding confidence intervals (CI) for risk ratios. The comparisons involving number of admissions and total days hospitalized were made using permutation tests. Approximate 95% CI for the differences in rates of these outcomes were calculated from the mean differences ± 1.96 standard error (SE). For those participants who were hospitalized, the duration of stay was compared among the treatment groups using a Wilcoxon rank-sum test.
To compare admission rates with respect to underlying risk of coronary events, participants were categorized according to quartiles of baseline five-year untreated risk of definite CHD death or nonfatal MI. These were calculated using a Cox proportional hazards survival model incorporating smoking, diabetes mellitus, nitrate consumption, angina pectoris, family history of CHD, widowhood, age, diastolic blood pressure and total/HDL (high density lipoprotein) cholesterol ratio, as previously reported (13). The numbers of patients admitted within each quartile for each treatment group were compared using the Fisher exact test. Differences in the effect of pravastatin across the quartiles were tested using a logistic regression model incorporating treatment, risk quartiles and their interactions. Statistical calculations were performed using SAS 6.12 for Windows.
Results are presented (unless otherwise indicated) as numbers of participants, numbers of admissions or total days hospitalized per 1,000 subject-years. The observed number of subjects, admissions or total days hospitalized can be recovered by multiplying the rates presented by the observed duration of follow-up in subject years and dividing by 1,000.
Duration of follow-up
In the primary analysis, there were 15,766 treatment-years of observed follow up in 3,293 placebo-treated participants and 15,909 treatment-years in 3,302 pravastatin-treated patients.
Number of trial subjects hospitalized
During the trial, 2,198 (33%) of the 6,595 men were admitted to hospital on one or more occasions. Cardiovascular disease was the principal cause of admission in 736 individuals. The number of participants (per 1,000) requiring hospitalization according to trial category is listed in Table 1, and the Kaplan-Meier curves for time to first cardiovascular and noncardiovascular admissions are illustrated in Figure 1A and B, respectively. Pravastatin reduced the number of participants requiring hospital admission for CHD by 29% (95% CI 16 to 41, p < 0.0001), and for all cardiovascular causes by 21% (95% CI 9 to 31, p = 0.0008). There was no significant difference between treatment groups in the number of subjects or time to first hospitalization for noncardiovascular causes.
Number of hospital admissions
Among the 2,198 participants admitted to hospital in the study, there was a total of 4,333 hospital admissions, a median of one (interquartile range [IQR] 1 to 2, range 1 to 29) per participant. Of all hospital admissions, 1,234 (28%) were for cardiovascular causes. Pravastatin reduced the number of admissions per 1,000 subject-years for atherosclerotic CHD by 9.3 (95% CI 4 to 14.6, p = 0.0008), and for all cardiovascular causes by 10.8 (95% CI 4 to 17.4, p = 0.0013). There were no significant differences in the number of admissions due to malignant neoplasms, trauma or psychiatric causes, and no significant impact on the overall risk of hospital admission for noncardiovascular causes (Table 2and Fig. 2). In total, there were 13.4 fewer hospital admissions per 1,000 subject-years for all causes in the pravastatin-treated group (95% CI −0.4 to 27.3, p = 0.076).
The number of hospital admissions for noncardiovascular diagnoses was analyzed in more detail according to major body systems using the primary ICD 8 code (Fig. 2). The only significant differences in hospital admission rates between the pravastatin- and placebo-treated groups were small reductions in the rates of admission for hepatobiliary and dermatologic causes. The former may reflect in part the beneficial effect of statins on bile lithogenicity (14,15).
Number of hospital bed-days
Among participants who were hospitalized, no significant difference was seen in the median duration of hospital admission between those treated with pravastatin and the placebo in any diagnostic category (Table 1). The 736 patients hospitalized for cardiovascular causes occupied a total of 9,306 hospital bed days (median 7, IQR 3 to 13, range 1 to 520), which represented 31% of the 29,851 hospital-days for all causes. Pravastatin treatment reduced the number of hospital bed-days for CHD by 61.4 per 1,000 subject-years (95% CI 25.2 to 97.5, p = 0.001), but there was no influence on the number of bed-days for other cardiovascular diagnoses. The number of bed-days for noncardiovascular causes was reduced by 94 per 1,000 subject-years (95% CI −64.8 to 252.6, p = 0.28) in the pravastatin-treated group. This resulted in a nonsignificant overall reduction in hospital bed usage of 168.7 per 1,000 subject-years (95% CI −18.9 to 356.3, p = 0.086) in the pravastatin-treated group.
Effect of pravastatin therapy in compliant subjects
For the participants with ≥75% compliance, there were 11,697 treatment-years of follow-up on 2,452 placebo-treated subjects, and 11,765 treatment-years on 2,450 pravastatin-treated individuals. Pravastatin treatment reduced the risk of compliant participants requiring hospital admission for atherosclerotic CHD by 35% (95% CI 19 to 47, p < 0.0001) and for all cardiovascular causes by 27% (95% CI 15 to 38, p = 0.0001). The number of compliant patients requiring hospital admission, the total number of admissions, the total hospital bed-days and the median length of stay are given in parentheses in Table 1. Compliant participants receiving the placebo had rates of hospitalization for cardiovascular causes that were similar to the total cohort, but had markedly lower rates of admission for malignancy and other noncardiovascular diagnoses. This is likely to be explained by the tendency of participants to discontinue preventive therapy for CHD in the event of diagnosis of an alternative life-threatening illness.
Effect of baseline risk
The number of hospital admissions for cardiovascular and noncardiovascular causes according to quartiles of baseline risk of the primary trial end point (definite CHD death or nonfatal MI) is illustrated in Figure 3. Pravastatin therapy resulted in reductions in cardiovascular hospitalizations in all quartiles of risk, with no evidence of a differential effect of treatment across the quartiles. The risk of noncardiovascular hospital admission also increased progressively from quartile 1 to quartile 4, but there was no trend towards any difference in noncardiovascular events between treatment groups.
Effect of pravastatin on hospital admissions
The data on hospitalization extend the principal results of the WOSCOPS in indicating that pravastatin treatment in middle-aged men with hypercholesterolemia substantially reduces the probability of hospital admission for cardiovascular causes. The effects noted here are largely attributable to the effect of pravastatin in reducing the relative risk of fatal CHD or nonfatal MI by 31% and coronary revascularization by 37% (1). As expected, the compliant subgroup showed a stronger treatment effect on cardiovascular hospitalizations than that seen for all participants.
Unlike the report of the primary end points of the trial, based on a “time to first event” analysis, the current results permit assessment of the effect of pravastatin on the overall requirement for hospital admissions during the average five-year follow-up of the trial. By including the effects of second and subsequent admissions, the number of hospital admissions for analysis is nearly doubled. The effect of pravastatin in reducing total cardiovascular admissions is not offset by any increase in noncardiovascular hospitalizations. However, in this primary prevention population, the reduction in cardiovascular hospitalizations was diluted by the greater number of noncardiovascular events to the extent that the overall requirement for hospital admission was not significantly reduced, although a favorable trend was seen.
Duration of hospital stay
The results of our analysis differ from those presented by the Scandinavian Simvastatin Survival Study (SSSS) because we found no significant effect of therapy on duration of hospital admission for CHD (16). We also found the distribution of length of hospital stay to be skewed and therefore undertook a nonparametric analysis and expressed the duration of admission as the median (IQR) and range. In contrast, the SSSS investigators reported an “average” duration of hospital stay and used parametric statistics, although they did not report whether their data were normally distributed, nor how they dealt statistically with multiple hospitalizations in the same participant.
The analysis of hospital days takes into account the number of hospital admissions as well as the duration of admission. A powerful beneficial effect was noted in the pravastatin-treated group in respect to cardiovascular diseases. A nonsignificant reduction occurred in total hospital-days for noncardiovascular causes in the pravastatin group, resulting in a total reduction of 168.7 hospital-days per 1,000 subject-years for all diagnoses in the pravastatin-treated group. There is no clear explanation for the reduction in the number of hospitals-days for noncardiovascular causes in the pravastatin group. This analysis is influenced by a small number of patients who had very prolonged hospital admissions (>1 year) and should be viewed with caution. However, this reduction might be attributable to prevention of cardiovascular complications other than myocardial infarction in subjects initially hospitalized for noncardiovascular causes.
Effect of baseline risk
The finding that hospital admissions for cardiovascular causes were reduced by pravastatin therapy across all quartiles of risk is as expected. Our previous report had shown that the proportional effect of pravastatin in reducing risk was similar regardless of the baseline level of risk (13). The observation that the probability of noncardiovascular hospitalization also increased in parallel with the cardiovascular risk may appear surprising at first sight. However, it is explicable on the basis that the subjects in the higher quartiles of cardiovascular risk tended to be older and were more likely to be smokers, thus also increasing their risk of noncardiovascular diseases such as cancer and respiratory disease.
In summary, these results confirm the beneficial effects of pravastatin therapy on reducing health care costs attributable to hospital admission (17). They provide further reassurance that therapy did not produce any adverse effects on noncardiovascular events, particularly on cancers, trauma and psychiatric hospitalizations.
Members of the West of Scotland Coronary Prevention Study Group
James Shepherd, MD, PhD (Chairman, Co-Principal Investigator), Stuart M. Cobbe, MD (Co-Principal Investigator), A. Ross Lorimer, MD, James H. McKillop, MD, PhD, Ian Ford, PhD, Christopher J. Packard, PhD, Peter W. Macfarlane, PhD, Christopher Isles, MD. All at Glasgow Royal Infirmary and Glasgow University, with the exceptions of I.F. (Robertson Centre for Biostatistics, Glasgow University) and C.I. (Department of Medicine, Dumfries & Galloway District General Hospital). This committee also constitutes the Publication Committee for the study.
Data and Safety Monitoring Committee
Michael F. Oliver, MD (Chairman) (National Heart and Lung Institute, London, United Kingdom), Anthony F. Lever, MD (Dept. of Medicine and Therapeutics, Western Infirmary, Glasgow, United Kingdom), Byron W. Brown, PhD (Stanford University, Stanford, California), John G.G. Ledingham, MD (Nuffield Department of Clinical Medicine, John Radcliffe Hospital, Oxford, United Kingdom), Stuart J. Pocock, PhD (London School of Hygiene & Tropical Medicine, London, United Kingdom), Basil M. Rifkind, MD, PhD (National Institutes of Health, National Heart, Lung & Blood Institute, Bethesda, Maryland).
Cardiovascular Endpoints Committee
Stuart M. Cobbe, MD, Barry D. Vallance, MD (Department of Cardiology, Hairmyres Hospital, East Kilbride, United Kingdom), Peter W. Macfarlane, PhD.
Adverse Events Review Board
A. Ross Lorimer, MD, James H. McKillop, MD, PhD, David Ballantyne, MD (Department of Cardiology, Victoria Infirmary, Glasgow, United Kingdom).
Data Centre Staff
John Norrie, Liz Anderson, David Duncan, Sharon Kean, Audrey Lawrence, June McGrath, Vivette Montgomery, PhD (Robertson Centre for Biostatistics, Glasgow University).
Melvyn Percy (Minerva Medical plc, Glasgow, United Kingdom).
Clinical Coordination, Monitoring and Administration
Elspeth Pomphrey, MD, Andrew Whitehouse, MD, Patricia Cameron, Pamela Parker, Fiona Porteous, Leslie Fletcher, Christine Kilday (Glasgow Royal Infirmary).
Computerized ECG Analysis
David Shoat (deceased), Shahid Latif, Julie Kennedy (Glasgow Royal Infirmary).
Margaret Anne Bell, Robert Birrell (Glasgow Royal Infirmary).
Company Liaison and General Support
Margot Mellies, MD, Joseph Meyer, MD, Wendy Campbell (Bristol-Myers Squibb Company, Princeton, New Jersey).
- coronary heart disease
- confidence interval
- ICD 8
- International Classification of Diseases 8th Revision
- myocardial infarction
- OPCS 4
- Office of Population Censuses and Surveys Revision 4 coding
- West of Scotland Coronary Prevention Study
- Received April 1, 1998.
- Revision received September 18, 1998.
- Accepted December 4, 1998.
- American College of Cardiology
- Davey Smith G.,
- Pekkanen J.
- Hulley S.B.,
- Walsh J.M.B.,
- Newman T.B.
- West of Scotland Coronary Prevention Study Group
- Pedersen T.R.,
- Kjekshus J.,
- Berg K.,
- et al.
- Caro J.,
- Klittich W.,
- McGuire A.,
- et al.