Low cholesterol concentrations and severe depressive symptoms in elderly peopleBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6940.1328 (Published 21 May 1994) Cite this as: BMJ 1994;308:1328
- S L Brown,
- M E Salive,
- T B Harris,
- E M Simonsick,
- J M Guarlnik,
- F J Kohout
- Epidemiology, Demography, and Biometry Program, National Institute on Aging, Bethesda, MD 20892, USA College of Dentistry, University of Iowa, Iowa City, IA 52242, USA
- Correspondence to: Dr S Lori Brown, Epidemiology, Demography, and Biometry Program, National Institute on Aging, 7201 Winsconsin Ave, Suite 3C309 Bethesda, MD 20892, USA.
- Accepted 10 March 1994
Objective : To investigate the reported association between low serum cholestrol concentration and severe depressive symptoms in an elderly population.
Design : Cross sectional analysis of pooled data from three communities of the established populations for epidemiologic studies of the elderly. Participants who completed their interview, including the Centers for Epidemiologic Studies' depression scale and consented to measurement of their cholesterol concentration were included in the study.
Subjects : 3939 men and women aged [elp]71.
Methods : X2 analysis, t tests, and multivariate regression analysis of the association between low cholesterol concentration and severe depressive symptoms. All analyses were stratified by sex, and multivariate analyses were adjusted for age, self reported health, physical function, number of drugs used, and weight loss. Main outcome measure - Score of depressive symptoms on the Centers for Epidemiologic Studies' depression scale.
Results : Depressive symptoms, cholesterol concentration, weight, and use of drugs were all associated with age in men and women. The relative odds of severe depressive symptoms (score >=16) for those with low cholesterol concentrations (<4.14 mmol/l) were 1.9 (95% confidence interval, 1.1 to 3.3) for the older group of men and 1.8 (1.1 to 2.9) for the older group of women. This association was also observed when depressive symptoms were analysed as a continuous rather than a categorical variable. In multivariate models that adjusted for age, self reported health, physical function, number of drugs used, and weight loss, the association was substantially weakened.
Conclusions : After several factors relating to health had been controlled for, no significant association between low cholesterol concentration and severe depressive symptoms was found.
High cholesterol concentration is a well established risk factor for heart disease
Low cholesterol concentration in elderly people is thought to cause severe depressive symptoms, which could increase mortality from external causes such as suicide, murder, and accidents
In this study men and women aged >=80 with a low cholesterol concentration were nearly twice as likely to have severe depressive symptoms as those with a normal or high concentration
After factors relating to health had been controlled for, however, low cholesterol concentration was not associated with severe depressive symptoms in this age group
>This finding is important for the management of elderly patients with high cholesterol concentrations
Although a high serum cholesterol concentration is a well established risk factor for heart disease,1 an analysis of results of cholesterol lowering clinical trials has suggested that the benefits of reducing heart disease by reducing cholesterol concentration are offset by an increase in deaths from external causes such as suicide, accidents, and murder.*RF 2-4* This increase in deaths from causes unrelated to illness occurred regardless of whether the cholesterol lowering regimen was dietary or pharmacological, prompting the authors to suggest that lowering cholesterol concentration might have neurochemical consequences. One explanation for this may be that lowering cholesterol concentration causes changes in the cholesterol content of the synaptosomal membrane and a decrease in the number of serotonin receptors.5 Because a low serotonin concentration has been associated with suicidal depression and impulsive behaviour6,7 some researchers have suggested that lowered serotonin concentration or perhaps fewer serotonin receptors may account for the increase in deaths from external causes in the treatment group in cholesterol lowering trials.5 Depression is a factor in a high proportion of suicides.8
On the basis of this argument, Morgan and colleagues investigated the association between low blood cholesterol concentration and an increase in depressive symptoms.9,10 They found an inverse association in men aged 70-89 but not in younger men, and they attributed this association to an affect of low cholesterol concentration. Evidence exists, however, for an alternative explanation. Poor or declining health has been associated with weight loss,11,12 and weight loss has been associated with lower serum cholesterol concentration.13 Poor or declining health is also associated with severe depressive symptoms or an increase in such symptoms among elderly people.14 Thus the inverse association between blood cholesterol concentration and depressive symptoms may result from their association with poor health.
Although Morgan et al considered health status to some degree in their analyses,9 their approach to adjusting for this factor may not have been adequate.15 Our study used data from the established populations for epidemiologic studies of the elderly to evaluate the association between low cholesterol concentration and severe depressive symptoms and adjusted for current illness, physical function, and weight loss as important potential confounders. We tried to replicate as closely as possible the variable and the types of analysis used by Morgan et al9 so that we could compare the main results.
The established populations for epidemiologic studies of the elderly were initiated by the National Institute on Aging in the United States to gather information on the health of people aged 65 and over.16,17 Baseline interviews were conducted between 1981 and 1983 in three communities - namely, east Boston, Massachusetts; Iowa and Washington counties, Iowa; and New Haven, Connecticut. During the participants' sixth annual follow ups, between 1987 and 1989, blood cholesterol concentrations were determined in those who consented. Participants, by then aged 71 or over, answered questions about their health, including questions about depressive symptoms as measured by the Centers for Epidemiologic Studies' depression scale.18 Participants were excluded from this analysis if they had not been able to answer the questions for themselves, if they had not been assessed with the depression scale, or if they had not had their cholesterol concentration tested. After exclusions 3939 (55%) participants remained in the study.
Measures of health, chronic illness, and physical function
At each annual interview participants were asked whether certain serious medical conditions had been diagnosed by a doctor since their last interview - namely, diabetes; heart attack, stroke, or brain haemorrhage; cancer, malignancy, or tumour; or broken hip. Participants scored one for each condition, resulting in a score of 0-5. A summary of which of these conditions had been diagnosed between the first and sixth annual follow ups provided a history of chronic or serious illness. The number of drugs (both those prescribed and those bought over the counter) that the participant had taken in the previous two weeks was used to measure participants' health status at the time of interview.
In addition to answering questions about serious medical conditions and use of drugs the participants answered questions about their ability to perform activities of daily living without the help of another person or equipment.19 Disability in any of six activities (walking across a small room, taking a bath, dressing, eating, transferring from a bed to a chair, and using the toilet, was summarised to give a physical function score of 0-6. Respondents self reported their health as excellent, good, fair, or poor.
Self reported weight was recorded at the sixth follow up, and the percentage change from the baseline weight was calculated. Participants were categorised according to whetherthey had lost more than 10% of their baseline weight over the six years of the study. More data were missing (6.6%) for this variable than for any other because either the baseline weight or the weight at the sixth follow up was missing. For all other variables <1% of values were missing.
Non-fasting blood specimens were collected by a phlebotomist in the home of consenting participants and flown to Nichols Institute in San Juan Capistrano, California. The serum cholesterol concentrations were measured by the enzymatic method with an instrument (Technicon SMAC, Tarrytown, New York) calibrated according to the manufacturer's specifications. Cholesterol concentrations were categorised as low (<4.14 mmol/l), normal (4.14-5.16 mmol/l), borderline (5.17-6.20 mmol/l), or high ([elp]6.21 mmol/l.9
Measurement of depressive symptoms
The Centers for Epidemiologic Studies' despression scale18 or a modification of it was used at each of the sites participating in the established populations for epidemiologic studies of the elderly. To minimise the burden on the participants a shorter version of the scale was used in east Boston and Iowa. The scores of these shorter versions were standardised against the first national health and nutrition examination survey, in 1974-5, by means of the equation xt=(sdt/sds (xs-ms) + mt, where xt is the transformed score, xs is the participant's score on the original scale, ms and sds are the sex specific mean and standard deviation of scores from participants at the sixth follow up, and mt and sdt are the sex specific mean and standard deviation from the first national health and nutrition examination survey. An adjusted score of >=16 indicated severe depression.18,20 Full details on the use of the modified scales, the methods for transforming the results to accord with the full scale, and the validity of these scales have been published. 20
All analyses were performed with SAS software (SAS Institute, Cary, North Carolina). Analyses were stratified by sex. For comparison of variables by age group Student's t testwas used for continuous data and the X2 test was used for categorical data. A Wilcoxon non-parametric test on ranks was used to assess the association between scores on the Centers for Epidemiologic Studies' depression scale and low cholesterol concentration. Pearson's correlation analysis of variables and depression scores was also performed.Multiple linear regression models with depression scores as the outcome were used to assess the relation between cholesterol concentration and depression, with potential confounders controlled for.
The mean scores for depressive symptoms and proportion of participants categorised as being severely depressed were both significantly higher in the older age group (>=80) than the younger age group (70-79) in both men and women (table I). The mean cholesterol concentration was significantly lower in the older age group. Low cholesterol concentrations were found in a higher proportion of the older participants (table I). A significantly lower mean weight and a significantly greater mean change from baseline weight were found in the older age group, and a significantly higher proportion of these participants had lost more than 10% of their baseline weight (table I).
Because Morgan et al9 had observed a significant association between low cholesterol concentration and severe depressive symptoms classified as a categorical variable, we analysed the data to determine if the proportion of severely depressed participants was higher in the group with low cholesterol concentration than in the other groups in the established populations for epidemiologic studies of the elderly (table II). For both men and women aged >=80 a significant association was found between low cholesterol concentration and severe depression. No significant association between cholesterol concentration and severe depression was found in the younger age group.
The Wilcoxon rank test of the depression scores of the participants with low cholesterol concentration compared with all other participants indicated that in both age groups of men low cholesterol concentration was associated with severe depressive symptoms (table III). This was also true in the older group of women.
Pearson's correlation coefficients indicated that the log depression score was inversely associated with cholesterol concentration (table IV). Weight loss, number of drugs taken in the previous two weeks, and other factors related to health were associated with both the log depression score and cholesterol concentration in both men and women (table IV).
Because the relation between low cholesterol concentration and depression might be confounded by the effects of advancing age and subsequent illness and weight loss, we analysed multiple regression models with log depression scores as the dependent variable. In the simplest model age, self reported health, physical function, and a low cholesterol variable (low upisilon all others) were analysed for both men and women. The results of this analysis indicated that the association between low cholesterol concentration and higher depression scores was not significant in either men or women (table V, model A).
The model that included weight loss and number of drugs used diminished the association between low cholesterol concentration and severe depressive symptoms (table V, model B). This model only slightly improved the explanatory power in the models for both men and women. For women, however, the coefficient for cholesterol concentration was halved when weight loss and number of drugs used were included in the model. The effect on the model for men was less drastic. Neither a six year history of chronic conditions nor the development of new chronic conditions was significant in these regression models. A logistic model that used severe depressive symptoms as the dependent variable yielded similar results (data not shown).
An association between low cholesterol concentration and severe depressive symptoms was observed in elderly men and women in the established populations for epidemiologic studies of the elderly. This confirms the findings of a report by Morgan et al on men from Rancho Bernardo, California.9 In contrast to their study, however, no significant association was found after we had controlled for self reported health and physical functioning. In a multivariate model that adjusted for weight loss and the number of drugs used as an indication of current illness no significant association was found between low cholesterol concentration and severe depressive symptoms.
Several possible explanations exist for the differences between our findings and those of Morgan et al.9 In both studies respondents were asked to rate their own health, including their psychological well-being.21 This party accounts for the association between self reported health and depressive symptoms measured with the Beck depression inventory or the Centers for Epidemiologic Studies' depression scale. The other variable in the regression model was the level of physical function. In the study of men in Rancho Bernardo physical function was assessed by participants being asked to rate their current physical function against that 10 years earlier. In the established populations for epidemiologic studies of the elderly it was assessed by participants being asked to identify difficulties in performing activities of daily living. This is a more direct and stronger measure of restrictions in physical function. The regression model used in the study of men in Rancho Bernardo explained 10% of the variance in depressive symptoms whereas the model in this study explained 13-15% of the variance (table V).
The difference may be further explained by the fact that the populations studied were not alike with respect to age and health: the age of the cohort in Rancho Bernardo ranged from 50 to 89 while in this study the age of male participants ranged from 71 to 100. Also, participants with severe depressive symptoms differed in profile: whereas in Rancho Bernardo the highest rate of severe depression was 5.4%, in the 80-89 year old age group, in the established populations for epidemiologic studies of the elderly the lowest rate was 9.1%, in the youngest age group of men. These distinctions could be due to differences in the two models but are probably due to characteristics of these populations, such as health status. In Morgan et al's study the number of drugs used was not entered into the final regression model because it did not reach the level of P<=0.05 in the regression model; this may have been because the younger participants were taking fewer drugs or because the older participants were heathier than those in the established populations for epidemiologic studies of the elderly. Use of drugs was positively correlated with the score on the Centers for Epidemiologic Studies' depression scale in this study. These differences in the study populations and in the variables examined may explain why the conclusions differ.
Our findings do not support the hypothesis that increased mortality from external causes in people with a lower cholesterol concentration is due to their having more severe depressive symptoms, which might lead to suicide or violence. Cholesterol concentration is influenced by dietary and genetic factors and by health.*RF 11-13* This study showed that after the confounding effects of poor health had been controlled for by adjustment for self reported health, physical function, recent weight loss, and the number of drugs used in the previous two weeks, no association between low cholesterol concentration and severe depressive symptoms in either elderly men or elderly women was observed. Doctors should not be deterred, therefore, from seeking to lower cholesterol concentration in elderly patients whose concentrations are high.