Nonfasting Plasma Total Homocysteine Level and Mortality in Middle-Aged and Elderly Men and Women in Jerusalem

The metabolism of homocysteine, a sulfur amino acid, is at the intersection of two metabolic pathways: transsulfuration and remethylation (1). McCully (2) first proposed that severe hyperhomocysteinemia is related to both atherosclerosis and vascular thrombosis. Recent evidence (3-5) has shown an association between mildly to moderately elevated blood concentrations of total homocysteine and vascular disease (including its coronary, cerebral, and peripheral manifestations). Much of the supporting evidence for this association has been obtained from casecontrol studies; reports of prospective studies of cardiovascular disease, however, are inconsistent (6-12). It remains to be established whether this relation is causal and whether reduction of plasma homocysteine level will decrease risk. Most studies have been done in Europe and North America. Only two recent reportsone from the Framingham Study (13) and one on patients with coronary heart disease in Norway (14)have used total mortality as an end point with which to assess health outcomes associated with a modestly elevated homocysteine level. We addressed the question of this relation in a study of nonfasting plasma homocysteine levels and 9- to 11-year all-cause mortality in a cohort of Jewish men and women 50 years of age and older living in Jerusalem. The study sample is ethnically heterogeneous, consisting mainly of persons from central and eastern Europe, northern Africa, and the Middle East who immigrated in the 1950s and 1960s, as well as those born in Israel. Methods Study Sample The third round of examinations of the Kiryat Yovel Community Health Study took place from 1985 to 1987. A neighborhood sample in western Jerusalem was identified by conducting a household census of dwelling units. All identified residents 50 years of age or older were invited for an interview and examination and were asked to give informed consent to participate (15-17). Data Collection A structured interview, administered by trained interviewers during the afternoon, was followed by an examination that included standardized measurements of blood pressure, anthropometric indexes, 12-lead electrocardiography, and a nonfasting blood sample (16, 17). Biochemical Measurements Blood was drawn into plain Vacutainers (Becton Dickinson, Carlsbad, California) and Vacutainers that contained EDTA; 90% of the samples were taken between 1:00 p.m. and 6:00 p.m. The EDTA tubes were immediately refrigerated for up to 2 to 3 hours until centrifugation. Aliquots were stored at 20 C for 9 to 11 years until they were shipped on dry ice to Boston, Massachusetts, for analysis of homocysteine. Plasma total homocysteine, the sum of protein-bound and free homocysteine, was measured by using high-performance liquid chromatography with fluorometric detection, as described by Araki and Sako (18), except for isocratic column elution. Pooled plasma was used for quality control. The interassay and intra-assay coefficient of variation for this method is less than 5%. Serum glucose level, cholesterol level, thiocyanate level, creatinine concentration, blood urea nitrogen level, and albumin level were measured on a Technicon SMAC (Technicon Instruments Corp., Tarrytown, New York). High-density lipoprotein cholesterol level was measured enzymatically (Laboratoires Biotrol, Paris, France) on a Cobas Bio autoanalyzer (F. Hoffman-La Roche Ltd., Basel, Switzerland). Physical Measurements and Interview Data Blood pressure was measured with a mercury sphygmomanometer (16). Body mass index was computed (17). Participants were asked whether a physician had ever told them that they had diabetes. They were also asked, Is your general health at present very good, good, not so good, poor, or very poor? The first two categories were combined as good health, and the last two categories were combined as poor health. Simple self-appraisals, which are common indexes of general health, are correlated with health ratings on the basis of objective measures and are predictors of subsequent death (19-23). Prevalence of cardiovascular disease was defined as typical angina (24) confirmed by a physician, reported history of heart attack, or reported history of stroke. A food-frequency questionnaire included assessment of the usual intake of fruit, fresh vegetables, and cooked vegetables. Follow-up and Causes of Death Deaths that occurred before April 1996 and the underlying cause of death as coded by the Israel Central Bureau of Statistics (International Classification of Diseases, Ninth Revision [ICD-9] codes) were identified by linkage with the national population registry. Statistical Analysis Skewed distribution of plasma homocysteine level was corrected by natural logarithmic transformation. Mean homocysteine levels, when presented, are geometric unless otherwise specified. Associations of homocysteine level with covariates were assessed by age-adjusted partial Pearson correlations and analysis of variance. Kaplan-Meier survival curves were computed for quintiles of homocysteine level after age adjustment by regression. The main analyses used Cox proportional-hazards regression to model survival according to sex-pooled quintiles of homocysteine level, adjusting for possible confounders. In these models, the outcome variable was time to event. Tests for trend were assessed with the logarithm of homocysteine level introduced as a continuous variable. Uniformity in the association of homocysteine level with survival over time was tested by introduction of time-dependent terms. Differences in hazard ratios between the sexes, age groups (<65 years of age and 65 years of age), and ethnic groups were tested by using multiplication terms. Analyses were implemented by using SPSS (SPSS, Inc., Chicago, Illinois). The population attributable fractionthat is, the proportion of all deaths in the population associated with elevated homocysteine levels (cut-off points used elsewhere [25, 26], 13 and 14 mol/L)was computed as Pe (HR 1)/1 + [Pe (HR 1)], where Pe is the proportion of patients with a plasma homocysteine level of at least 13 mol/L or at least 14 mol/L and HR is the respective hazard ratio computed from Cox models. Results Participant Characteristics All 2303 persons identified in 3434 dwelling units in the Kiryat Yovel community who were at least 50 years of age were invited to participate in our study. The response rate for the original census was 96%. A total of 1948 men and women gave informed consent and agreed to participate (85% response rate). Nonrespondents were similar to respondents with regard to age and sex. Plasma total homocysteine was measured in blood samples obtained from 1788 participants (92% of the total); the 160 participants with missing measurements did not differ significantly from those with no missing measurements with respect to age, reported diabetes, or self-appraised health. Covariate data were incomplete for 77 of the 1788 participants in the multivariate-adjusted models. The age distribution of the study sample is shown in Table 1. The mean age was 64.6 years for men and 64.5 years for women (range for both sexes, 50 to 92 years). Table 2 shows the heterogeneity of the sample with respect to place of birth and level of education. Of the 1788 participants, 13% (n=239) reported having received a physician's diagnosis of diabetes, 16% (n=283) had a history of cardiovascular disease, and 30% had hypertension (defined as systolic blood pressure 160 mm Hg, diastolic pressure 95 mm Hg, or current treatment for hypertension). Distributions of smoking and self-appraised health are shown in Table 2. Nineteen percent of men and 41% of women had a total serum cholesterol level of 6.5 mmol/L or greater ( 251 mg/dL) . The prevalence of obesity (body mass index 30.0 kg/m2) was high (32% in women and 16% in men). Table 1. Plasma Total Homocysteine Levels and Number of Deaths during 9- to 11-Year Follow-up, according to Age at Initial Examination (1985-1987) Table 2. Predictors of All-Cause Death during 9- to 11-Year Follow-up Correlates of Total Plasma Homocysteine Level We examined the relation of plasma homocysteine level to the time that had elapsed since food or drink was last consumed. A weak age-adjusted association (r=0.06; P=0.049) was restricted to women. Homocysteine levels increased with age, were higher in men than in women (Table 1), and were not significantly associated with place of birth or level of education (not shown). The strongest age-adjusted correlations were with serum creatinine concentration (r=0.21 in men and r=0.29 in women; P<0.001). The inverse relation with serum glucose level (r= 0.09 in men [P=0.011] and r= 0.15 in women [P<0.001]) was not affected by the amount of time that had passed since the participant's last meal. Age- and sex-adjusted mean homocysteine levels were lower among persons who reported that they had diabetes than among those who did not (1.1 mol/L; P<0.001). Homocysteine levels were 1.1 mol/L higher in men who smoked (P=0.004) than in men who did not smoke and were positively correlated with serum thiocyanate level (r=0.12; P<0.001); these associations were weaker and nonsignificant in women. No significant associations were seen with blood pressure, body mass index, or serum lipid levels. Correlations with intake of fruit, fresh vegetables, cooked vegetables, and total vegetables were weak (r= 0.08 [P=0.045], r= 0.10 [P=0.014], r= 0.04 [P>0.2], and r= 0.10 [P=0.015], respectively, in men; r= 0.06, r= 0.04, r= 0.05, and r= 0.04, respectively, in women [for all comparisons P>0.2]). Participants with poor self-appraised health had higher age-adjusted homocysteine levels than participants in fair health or good health (1.8 mol/L and 1.5 mol/L, respectively, for men [P=0.05] and 1.0 mol/L and 0.9 mol/L, respectively, for women [P=0.2]; P=0.012 in the total sample [sex-adjusted]). Homocysteine Level and Death during 9 to 11 Years of Follow-up Among the 1788 participants,