A model has been proposed in which otherwise clinically occult cerebrovascular disease contributes to the pathogenesis of depression in later life (1, 2). The model implies that risk factors for cerebrovascular disease are associated with depression in older persons. While we did not find a significant independent association between cerebrovascular risk factors and depression in primary care elderly patients (3), a finding that confirmed the results of most prior studies of other populations (see reference 3 for review), our work was limited by its cross-sectional design. We therefore tested the hypothesis that cerebrovascular risk factors at intake would be independently associated with depression at 1-year follow-up in a group of primary care elders.

As described in detail elsewhere (4), the subjects were recruited from internal medicine private offices and a family medicine clinic. All patients aged 60 years or older who gave informed consent (according to approved formal oral or written consent procedures) were eligible to participate. Stratified sampling was used to oversample patients with depressive symptoms, but the final study group included patients scoring across the full range of symptoms.

Subject assessments were based on the Structured Clinical Interview for DSM-III-R (SCID) (5) administered by trained raters at study intake and at 1-year follow-up. Medical measures were completed by a physician investigator (J.M.L.) on the basis of the interview and chart review. Cumulative severity of cerebrovascular risk factors was rated by using the cerebrovascular risk factors score (6) a weighted sum of the severities of seven specific risk factors (systolic blood pressure, antihypertensive treatment, cardiovascular disease, diabetes mellitus, atrial fibrillation, left ventricular hypertrophy, and cigarette smoking) based on the American Heart Association criteria for risk factors for stroke. Overall medical illness burden was rated on the Cumulative Illness Rating Scale (7). Psychiatric diagnoses were assigned by a consensus conference of raters and investigators, on the basis of the SCID and record review. As recommended by our group and others (8), the determination of depressive symptoms used an "inclusive" approach, and no subjects were excluded from the study on the basis of medical comorbidity.

The severity of depressive symptoms was measured by the 24-item Hamilton Rating Scale for Depression (9). Depression diagnosis groups were defined as follows: 1) current major depression (N=13); 2) current minor depression, based on criteria from the appendix of DSM-IV (N=9); 3) subsyndromal depression, defined as a score higher than 10 on the Hamilton depression scale and not major or minor depression, a definition that includes additional patients suffering functional disability comparable to that of minor depression (see our discussion in reference 10) (N=35); and 4) nondepressed (N=190). We also conducted separate analyses comparing the variables of interest in the patients with later-onset major depression (onset at 60 years or later, N=30) and the nondepressed group.

Compared to the 44 eligible subjects who refused follow-up, the 247 subjects in the present study did not differ significantly in age, gender, cerebrovascular risk factor score, Cumulative Illness Rating Scale score, or depression diagnosis, but they did have a higher mean level of education—13.4 years (SD=2.8) versus 12.2 years (SD=3.1) (t=–2.5, df=55.8, p=0.02)—and a lower mean initial score on the Hamilton depression scale—8.1 (SD=6.5) versus 10.7 (SD=7.3) (t=2.2, df=55.7, p=0.03).

Statistical analyses used multiple regression techniques to determine the independent association of intake score for cerebrovascular risk factors (without and with covariance for intake depression and cumulative illness ratings) with depression outcome while controlling for age, gender, and education. We used Poisson (log linear) regression, with an adjustment for extra-Poisson variation, when the outcome was discrete with a skewed distribution (Hamilton depression scale), and polytomous logistic regression for the categorical outcome (depression diagnosis) (11). We report two-tailed p values and indicate parameter estimates (coefficients and their standard errors) for significant results (p<0.05).

The subjects in the study group had a mean age of 71.1 years (SD=7.5, range=60–94), a mean education level of 13.4 years (SD=2.8, range=1–17), a mean score for cerebrovascular risk factors of 7.6 (SD=3.8, range=0–21), and a mean score on the Cumulative Illness Rating Scale of 6.0 (SD=2.8, range=0–16). More than one-half of the subjects (59%, N=145) were women. The mean scores on the Hamilton depression scale at intake and 1 year, respectively, for the four diagnostic groups were as follows: major depression, 23.0 (SD=6.2) and 14.8 (SD=6.8); minor depression, 13.3 (SD=4.8) and 10.4 (SD=5.4); subsyndromal depression, 13.9 (SD=2.9) and 11.5 (SD=6.1); and nondepressed, 5.0 (SD=2.5) and 5.5 (SD=4.3).

A higher initial score for cerebrovascular risk factors had a significant independent association with a higher 1-year score on the Hamilton depression scale (coefficient=0.03, SE=0.01, F=7.4, df=1, 242, p=0.007) and with the 1-year depression diagnosis group (coefficient=–0.09, SE=0.04, χ²=5.2, df=1, p=0.02). The association between cerebrovascular risk factors and 1-year Hamilton depression score remained significant when also controlled for initial Hamilton depression score (coefficient=0.02, SE=0.01, F=4.0, df=1, 241, p=0.04) but not when additionally controlled for initial score on the Cumulative Illness Rating Scale (F=1.6, df=1, 240, p=0.20). The association of initial cerebrovascular risk factors with the 1-year depression diagnosis group did not remain significant when also controlled for initial depression diagnosis (χ²=2.8, df=1, p=0.10) or initial score on the Cumulative Illness Rating Scale (χ²=1.6, df=1, p=0.21). In these regressions, a higher initial cumulative illness rating was independently associated with both higher 1-year Hamilton depression score (coefficient=0.05, SE=0.01, F=10.4, df=1, 240, p=0.001) and the 1-year depression diagnosis group (coefficient=–0.17, SE=0.06, χ²=7.5, df=1, p=0.006). Also controlling for antidepressant treatment did not change any of these findings.

Initial score for cerebrovascular risk factors did not have a significant independent association with later-onset depression (χ²=0.6, df=1, p=0.45). No individual cerebrovascular risk factor had a significant independent association with either 1-year Hamilton depression score or depression diagnosis group.

Our hypothesis was largely confirmed: cumulative severity of cerebrovascular risk factors was associated with depressive symptoms and diagnoses at 1 year, supporting the applicability of the cerebrovascular model of depression to this older primary care group. The results are consistent with a recent report of a cross-sectional independent association of coronary heart disease (albeit not hypertension) with depressive symptoms in an older community sample (12).

Overall medical burden (Cumulative Illness Rating Scale score) was a powerful predictor of depression outcome, a finding consistent with those from prior studies (8, 13). The cerebrovascular risk factors themselves, as well as their sequelae and comorbidities, are embedded in the Cumulative Illness Rating Scale in a manner too complex to disentangle definitively. Thus, these results suggest that cerebrovascular risk factors may play a pathogenic role, but other disease processes cannot be discounted.

Our risk factor approach is complementary to, but not a substitute for, studies examining brain processes more directly (e.g., neuroimaging). Lengthier longitudinal studies of hypothesized pathobiological and psychological risk factors should be undertaken in the context of mechanism-specific neurobiological investigations to test the cerebrovascular model and other complementary theories in diverse groups of depressed subjects.

Presented in part at the 12th annual meeting of the American Association for Geriatric Psychiatry, New Orleans, March 14–17, 1999. Received Oct. 13, 1999; revision received Jan. 31, 2000; accepted March 9, 2000. From the Laboratory of Depression and Medical Comorbidity, Program in Geriatrics and Neuropsychiatry, Departments of Psychiatry and Biostatistics, University of Rochester Medical Center. Address reprint requests to Dr. Lyness, Department of Psychiatry, University of Rochester Medical Center, 300 Crittenden Blvd., Rochester, NY 14642; jeffrey_lyness@urmc.rochester.edu (e-mail). Supported by NIMH grants MH-01113 (Dr. Lyness) and MH-18911 (Dr. Caine). The authors thank T.K. Noel, C. Doane, A. Wisner, A. Gleason, G. Kiernan, H. Stiner, and Z. Yoediono, the patients and staff of Drs. J. Allen, M. Bergin, M. Berliant, C. Conners, R. Greene, M. Labanowski, R. Maggio, B. Peyser, and R. Pincus, and the Family Medicine Center at Highland Hospital.