Long-term Assessment of Psychological Well-being in a Randomized Placebo-Controlled Trial of Cholesterol Reduction With Pravastatin FREE

LARGE CLINICAL trials have confirmed that lowering serum cholesterol levels with statins decreases the risk of myocardial infarction and death from coronary artery disease.^1- 5 There has, however, been controversial evidence that reducing serum cholesterol levels may increase the risk of morbidity and mortality from other causes. Initial evidence from early nonstatin clinical trials suggested that patients randomly allocated to active treatment had a higher risk of death from trauma, suicide, and homicide.^6- 8 Subsequent research has examined many aspects of the hypothesized relation between cholesterol levels and psychological morbidity and mortality. Large cohort studies have shown that participants with very low serum cholesterol levels have a higher risk of suicide^9- 11 and mortality from injury.¹² Cross-sectional studies have found associations between low serum cholesterol levels and depression.¹³ In small studies of criminal populations or individuals with psychiatric illness, associations have been reported between low serum cholesterol levels and the risk of suicide,^14- 16 antisocial personality,¹⁷ aggressive conduct disorder,¹⁸ and violent behavior.¹⁹ In other studies, no evidence of an association between low cholesterol levels and psychological variables has been found.^20- 26 The most widely supported biological mechanism postulates a link between serum cholesterol and central nervous system serotonin.^27- 28

More recently published clinical trials of 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors have shown no increase in violent death among those receiving active treatment, although the number of suicide and trauma deaths has been small.^1- 5 Several studies have assessed psychological well-being, which is likely to be a more sensitive measure than death due to violence or suicide, and have also found no evidence for an adverse effect of statin therapy.^29- 31 However, because of the widespread and increasing use of cholesterol-lowering therapy, a possible association between low cholesterol level and psychiatric morbidity, and the potential public health importance of a small increase in psychiatric morbidity or violence, it is important to exclude with confidence any small or infrequent adverse effect of prolonged cholesterol reduction.^32- 34 The Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) trial was designed to investigate the effects of pravastatin therapy on mortality in patients with ischemic heart disease and serum cholesterol levels between 4.0 and 7.0 mmol/L (155-271 mg/dL).^5- 35 The present report arises from a substudy of the LIPID study. We assessed self-reported measures of anxiety and depression, anger expression, impulsiveness, excessive alcohol consumption, and adverse life events at baseline and during 4 years of follow-up.

A detailed description of the LIPID study has been published elsewhere.^5,35 Participants were aged 31 to 74 years, had a fasting serum cholesterol level of 4.0 to 7.0 mmol/L (155-271 mg/dL) before randomization, and had a history of acute myocardial infarction or hospitalization for unstable angina pectoris in the previous 3 to 36 months. After a run-in phase, patients meeting entry criteria were randomly assigned to pravastatin sodium (40 mg/d) or placebo. Patients were randomly selected for the psychological study by the National Health and Medical Research Council Clinical Trials Centre in Sydney, Australia. Thirty-six LIPID centers from Australia and New Zealand participated in this psychological substudy, and from these, a representative sample of 1222 LIPID study participants were invited to participate. The study was approved by the relevant ethics committee for each center. Psychological well-being was assessed by a standardized self-administered questionnaire before random allocation of patients to pravastatin or placebo, and again at 6 months and at 1, 2, and 4 years, during follow-up visits. Questionnaires were completed by participants in the clinic (22%), or were completed at home and returned to the clinic by post.

Sample size calculations based on differences in means indicated that a study with 1160 patients would have 90% power to detect differences of 0.2 SD at the .05 significance level. Of the 1222 LIPID study participants invited to take part, 1130 completed and returned the baseline questionnaire, for a response rate of 93%. The median number of participants per center was 20, with a range of 3 to 115. The patients who completed and returned a baseline questionnaire were given follow-up questionnaires. Among these, the response rate was 90% at 6 months, 90% at 1 year, 88% at 2 years, and 77% at 4 years. One center with 50 subjects (4.5%) did not administer the year 4 questionnaire.

The primary outcomes were chosen to cover the dimensions of psychological well-being that may be associated with suicide or violent behavior: anxiety, depression, anger expression, and impulsiveness.

The 30-question version of the General Health Questionnaire (GHQ) was chosen to assess anxiety and depression.³⁶ Severity was assessed from Likert scores, which sum the graded responses (0, 1, 2, 3) to each question. To define a "case" of depression using the GHQ, a different scoring system was used. Responses to each question were coded categorically (0, 0, 1, 1). A participant was classified as a "GHQ case" if the total score for all questions was greater than 10.³⁷ Subscales identifying the following factors were also examined: anxiety, worry, and tension; feelings of incompetence and low self-esteem; depression and hopelessness; difficulty coping; and social dysfunction.³⁸

The Spielberger Anger Expression Scale (STAXI) was used to assess anger.³⁹ The total anger expression score measures the frequency with which anger is expressed. The 3 subscales are (1) anger-out, which measures anger expressed toward other persons or objects; (2) anger-in, which measures the degree to which anger is suppressed or directed toward self; and (3) anger control, which measures the degree to which individuals control anger.

The Barratt Impulsiveness Scale (BIS-11) was used to measure the dimensions of impulsiveness.^40- 41 This scale has 3 subscales: (1) attentional impulsiveness; (2) motor impulsiveness, which involves acting without thinking; and (3) nonplanning impulsiveness, which involves a lack of concern for the future.

The following outcomes were measured to assess possible consequences or causes of psychological ill-health.

The occurrence of the following life events during the preceding year were recorded: personal serious illness, financial difficulty, divorce or separation, marital or family problems, change of job or residence, major or minor accident, and mugging or robbery.

Total alcohol consumption (the number of units consumed during the preceding week) was recorded. A score of 2 or more on the CAGE questionnaire⁴² was classified as evidence of alcohol dependence.

Consumption of sleeping pills, antianxiety medications, and antidepressants during the preceding 2 weeks was noted.

The estimated number of visits with friends or family during the preceding month was used as a measure of social integration or isolation.⁴³

Participants were asked to rate their overall health during the last 2 months as poor, average, good, or very good, and to state whether they had used painkillers during the preceding 2 weeks, and whether they had experienced serious illness, injury, or an operation in the past 12 months. Employment status was classified as: 1, employed, including keeping house and family; 2, retired; or 3, unemployed or out of work owing to sickness or disability. Socioeconomic status was classified using the Registrar General's Classification.⁴⁴ Information was available from the main study on lipid levels, baseline smoking status (never smoker, current smoker, ex-smoker), and body mass index.

The large number of items on the GHQ, the Spielberger Anger Expression Scale, and the Barratt Impulsiveness Scale meant that occasional items were missed. Where only 1 or 2 items on a scale were missing from a completed questionnaire, those item values were replaced by the individual's average score on the remaining items. Linear mixed effects models were used to compare psychological well-being in the treatment and control groups for the continuous outcome measures.⁴⁵ A random intercept was included, but all other effects were fixed. Each continuous outcome was also categorized according to a cutoff level corresponding to the top 5% at baseline. For binary outcomes, odds ratios were estimated, using generalized estimating equations with an independence working correlation.⁴⁶ All follow-up measures were included as outcome data in the model, with the baseline value of the measure and an effect for time included as covariates. For ease of presentation, the final models were built containing a fixed set of covariates across all models, including any baseline variables significantly associated with any outcome or found to be a confounder. Prespecified subgroup analyses evaluated treatment effects by sex, age (>70 years), baseline cholesterol level (<10th percentile), and baseline psychological status. For anger expression and alcohol consumption, the subgroup of individuals scoring above the 90th percentile was assessed. In the models presented, missing data categories were included for covariates. Analyses were carried out on an intention-to-treat basis, with the exception of the relation between a change in cholesterol level and psychological outcome. The latter analysis included all patients and the effect of a change in cholesterol level was examined ignoring treatment assignment.

The baseline characteristics of the study population, which are similar to those of the full sample in the main LIPID study, are presented in Table 1. Differences in the proportion of individuals with characteristics at baseline that may influence psychological well-being during follow-up, such as the occurrence of personal life events, accidents, alcohol consumption, health status, and psychological scores, were in all cases less than 5%. There was no association between baseline serum cholesterol levels and baseline GHQ, anger expression, or impulsiveness scores.

Serum cholesterol level decreased by an average of 1.3 mmol/L (50 mg/dL) in the pravastatin group over the first 6 months, and the decrease was maintained during follow-up (Figure 1). There was no change in the mean serum cholesterol level for participants randomly assigned to placebo. The primary outcome measures, the GHQ (Likert), anger expression, and impulsiveness scores at the baseline assessment and during follow-up, are presented in Table 2 and Figure 1. There were no statistically significant differences between the treatment and control groups. To show the sizes of the differences excluded by this study, the adjusted mean differences and 95% confidence intervals from Table 2 are presented in Figure 2 as effect sizes. The width of the 95% confidence intervals was less than ±0.2 SD for all primary outcome measures, indicating that a difference between treatments of more than 0.2 SD is unlikely.

The results for the secondary outcome measures are shown in Table 3. There were no differences in the odds of adverse life events between the pravastatin and placebo groups, as indicated by the odds ratio of 1.0 and the narrow 95% confidence interval. The numbers of some events, such as mugging and divorce or separation, were small, so a small difference in risk according to treatment allocation cannot be ruled out. The placebo group were more likely to use antidepressants, although this difference was not statistically significant. There was also no significant difference in problem drinking between the placebo and pravastatin groups or average alcohol consumption (data not shown). Analyses also compared the proportion in each group with psychological problems. There was no significant difference between the treatment and control groups in the proportions classified by GHQ score as a case (odds ratio, 0.8; 95% confidence interval, 0.6-1.2). Nor did the treatment group have significantly more people scoring above the 95th percentile on the anger expression or impulsiveness scales.

Analyses of prespecified subgroups were undertaken to assess whether an effect of cholesterol reduction could occur in individuals with some predisposition, such as preexisting psychiatric illness (Table 4). There was no evidence for a difference in the GHQ Likert score, anger expression, or impulsiveness during follow-up for men, those older than 70 years, and those with a serum cholesterol level in the lowest 10% at baseline. The statistically significant difference in impulsiveness in women should be interpreted with caution given the large number of statistical tests performed. There was also no evidence of a treatment effect for individuals with high scores for depression, anger expression, and alcohol consumption at baseline (Table 4).

Confounding by events occurring during follow-up was considered. Analyses were repeated to include life events, use of antidepressant or antianxiety medications, and general health status as time-dependent covariates. There was no evidence that these factors were influencing the treatment effect.

All analyses described were carried out on an intention-to-treat basis. Examination of the relationship between a change in serum cholesterol level and psychological well-being (Table 5) did not reveal any association. The regression coefficients were negative, but they were not significantly different from zero and the confidence intervals were narrow, indicating that a decrease in cholesterol level was not associated with an increase in psychological ill-health.

While the overall response rate was fairly high, the response rates for specific outcome measures were lower. There were 1 or more missing responses for the GHQ in 5% of questionnaires, 22% of the impulsiveness scales, and 23% of the anger expression scales. Response rates were lower at the 4-year follow-up, but were similar for the placebo (78.5%) and pravastatin (76.6%) groups. To determine whether missing data could cause bias in the treatment effect, analyses were repeated for the primary outcome measures with the 4-year follow-up data excluded, and a complete case analysis performed. The results were similar to those presented in Table 2. Sensitivity analyses, in which missing outcome data were assigned various values, indicated that significant treatment harm was possible only in the extreme case in which all nonrespondents randomly allocated to pravastatin were in poor psychological health and all nonrespondents receiving placebo were in good psychological health. Sensitivity analyses also indicated that missing covariate information was unlikely to influence the results.

This study examined the effects of cholesterol-lowering therapy with pravastatin on psychological well-being. The results provide evidence that lowering cholesterol levels with pravastatin does not increase self-reported levels of depression, anxiety, anger expression, impulsiveness, or alcohol consumption, nor increase the risk of adverse life events. The serum cholesterol level was reduced on average by 1.3 mmol/L (50 mg/dL) among patients randomly assigned to pravastatin, and this reduction was maintained over 4 years of follow-up. There was no average decrease in serum cholesterol levels among patients randomly assigned to placebo. The measures of psychological well-being were similar in the treatment and control groups throughout the follow-up. The 95% confidence intervals for all primary outcomes measured excluded differences by treatment allocation of more than 0.2 SD, indicating that clinically important effects were unlikely to be missed. While there were some small baseline differences between the 2 groups, possible confounding was controlled in the analysis. There was no evidence of a direct association between either the baseline serum cholesterol level or a change in serum cholesterol level and psychological scores.

There is evidence that some statins reduce cardiovascular events in individuals with preexisting disease and serum cholesterol levels in the low normal range.^3,5 This and the availability of more potent cholesterol-lowering drugs make it likely that low serum cholesterol levels will be the target for an increasing proportion of patients at high risk. Several large epidemiological studies have suggested an association, during long-term follow-up, between serum cholesterol levels below 4.16 mmol/L (161 mg/dL) and the risk of suicide or violent death.^9- 10 In most clinical trials, serum cholesterol has been in a higher range during follow-up, even for subjects receiving active treatment. In the present study there was no evidence for an effect of treatment on psychological well-being for the 10% of patients with the lowest serum cholesterol levels. In this group, the mean total cholesterol level fell to 3.7 mmol/L (143 mg/dL) with pravastatin therapy.

It has been suggested that low serum cholesterol may influence psychological well-being only in predisposed individuals, for example, those with preexisting psychiatric illness or alcoholism.⁹ Several studies have reported associations between cholesterol levels and violence and suicidal tendency in criminal and psychiatric populations, respectively.^33- 34,47 As in most clinical trials, patients with known mental disorders or recognized poor compliance were excluded from the LIPID study. However, in predefined subgroup analyses of individuals with the most extreme scores for anxiety and depression, anger expression, impulsiveness, and alcohol consumption, there was no clear evidence of an adverse effect of cholesterol reduction. In addition, there was no evidence of an effect of age or sex. These results suggest that pravastatin can be used safely by a broad range of patients at risk of vascular disease, regardless of the initial cholesterol level or the presence of psychological conditions such as depression. The study population did not, however, include young persons and those at highest risk of suicide or accidental death. In addition, some subgroups were small. Although unlikely, the present study does not reliably exclude a treatment effect for all potential at-risk groups.

Psychological well-being may be influenced by the occurrence of cardiovascular events during follow-up. A decrease in the rate of myocardial infarction or stroke for subjects randomly allocated to pravastatin could mask a small direct adverse effect of reducing serum cholesterol levels. However, psychological well-being remained stable during the first 6 months of follow-up when cardiovascular events were similar for subjects taking active treatment and placebo.⁵ In addition, adjustment for health status during follow-up did not alter the results of the analysis.

Complex concepts, such as psychological well-being, are difficult to measure, and the ability of self-reported responses in questionnaires to predict behavior is uncertain. Epidemiological studies suggest that victims of trauma, homicide, and suicide share a number of psychological characteristics, which include depression, hostility, and impulsiveness.^48- 50 The GHQ, the Spielberger Anger Expression Scale, and the Barratt Impulsiveness Scale were chosen to measure the major outcomes because they provide a standardized method for measuring psychological parameters potentially associated with risk of accidents, violence, or suicide. These scales are widely used and have been well validated for a single assessment, but are less well validated for detecting change over time.

The need for large randomized clinical trials to reliably detect or exclude possible small adverse effects of cholesterol reduction has been emphasized by several investigators.^32- 34 The results of cohort and cross-sectional studies have been difficult to interpret because of the possibility that depression may reduce serum cholesterol levels by decreasing appetite, and the likely common association between low serum cholesterol level, depression or risk of suicide, and poor general health.⁵¹ The results of the present study confirm the findings of earlier assessments by randomized trials of HMG-CoA reductase inhibitors. In 5 large randomized, placebo-controlled clinical trials, including the LIPID study, the total number of deaths from trauma or suicide was 26 for subjects randomly allocated to a statin, compared with 31 subjects randomly allocated to placebo.^1- 5 In addition, previous randomized studies have found no evidence for an adverse effect of statins on psychological measures. In the Expanded Clinical Evaluation of Lovastatin (EXCEL) study, the incidence of self-reported depression and other adverse psychiatric events did not increase.⁵² In a blinded analysis from the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS), there was no significant change in health perception or emotional well-being after 12 months of treatment with either lovastatin or placebo.²⁹ In the Oxford cholesterol study there was no evidence of a difference in mood or use of psychotropic medication (assessed by questionnaire) after a mean of 152 weeks of therapy in 414 subjects randomly allocated to simvastatin compared with 207 randomly allocated to placebo.⁵³ Small studies examining the short-term effects of statins on mental function have also found no evidence for an adverse effect.^31,54 The present study has the advantages of a large sample size, a more detailed and comprehensive assessment of psychological well-being, a baseline assessment before random allocation to pravastatin or placebo, and repeated assessments during long-term follow-up. It therefore has a greater ability than previous studies to confidently exclude any small adverse effect of long-term reduction of serum cholesterol levels on psychological well-being. An adverse effect of pravastatin itself on these aspects of psychological well-being is also excluded.

In conclusion, this study provides evidence that long-term reduction of serum cholesterol levels with pravastatin has no adverse effect on psychological well-being.

Accepted for publication January 11, 2000.

The LIPID study was supported by a grant from Bristol-Myers Squibb, Princeton, NJ, and by the National Heart Foundation of Australia.

The study design, management, analysis, and interpretation of results were conducted independently. We are indebted to Deborah Scott, Dunedin, New Zealand, and Sarah Mulray from the National Health and Medical Research Council Clinical Trials Centre, Sydney, for managing the study; the LIPID study steering committee for their support; and the LIPID investigators and participants for their contribution to the study. A complete list of LIPID study investigators is given in the appendix to reference 5.

Australia: Jan Garrett (Fremantle Hospital), Carol Singh and Lyndy O'Connell (Launceston Hospital), Barbara Wicks (Pindara Specialist Centre), Mary Threlfall (Royal Prince Alfred Hospital), Heather Briggs (Alfred Hospital), Denise Gow, Maura Hopkins, and Susan Keynes (Flinders Medical Centre), Helen Harrap (Gippsland Base Hospital), Leslie Ross-Lee (Princess Alexandra Hospital), Nerelyn Whitehouse and Debra d'Silva (North West Regional Hospital), Jackie Padley and Frances O'Rielly (Royal North Shore Hospital), Jan Crettenden (Repatriation General Hospital), Lucy Coupland (Royal Canberra South [Woden Valley] Hospital), Meg Hooten and Leonie Carrick (Austin and Repatriation Medical Centre), Yvonne Cavenett (Box Hill Hospital), Michelle Sallaberger and Nola Green (Royal Melbourne Hospital), Janice Crowe (Coffs Harbour Hospital), Marianne Martin (Dandenong Hospital). New Zealand: Mary Denton and Lynette Pearce (Green Lane Hospital), Liz Low (Waikato Hospital), Jane Kenyon and Kay Hall (Hastings Hospital), Deborah Scott, Lenore Armstrong, and Mary Blok (Dunedin Hospital), Mary Gluyas (Ashburton Hospital), Phillipa Wright and Janet Brown (Auckland Hospital), Doreen Clarke (Gisborne Hospital), Anne Doone (Hutt Hospital), Pauline Barclay (Nelson Hospital), Patricia Heuser (Wellington Hospital), Lyn Gibson (Napier Hospital), Pam Scobie (Southland Hospital), Jeannie Bruning (Tauranga Hospital), Heather Roy (Timaru Hospital), Trish Healy (Wairau Hospital), Julie Calton (Northland Hospital), Alison Clague (North Shore Hospital), Tracey Lawson (Princess Margaret Hospital), Rohini Anandaraja (Taranaki Hospital).

Reprints: Ralph A. Stewart, MD, FRACP, Cardiovascular Research Unit, Green Lane Hospital, Private Bag 92189, Auckland 1030, New Zealand (e-mail: ralph.stewart@ahsl.co.nz).