End Points

The end points were incident CHD (including nonfatal myocardial infarction [MI] and fatal CHD) and stroke occurring between return of the baseline questionnaires and January 31, 1994 (men), or June 1, 1994 (women). We inquired about occurrence of cardiovascular end points on each biennial questionnaire. Participants reporting an incident MI or stroke were asked for permission to review medical records. Nonfatal MI was confirmed by symptoms plus either typical electrocardiographic changes or increased activities in cardiac enzymes (World Health Organization criteria).30 Infarctions that required hospital admission and for which confirmatory information was obtained by interview or letter, but for which no medical records were available, were designated as probable. We included all confirmed and probable cases in the analyses because results were the same after excluding probable cases. Strokes were confirmed if characterized by a typical neurological defect of sudden or rapid onset, lasting at least 24 hours, and attributable to a cerebrovascular event. Strokes caused by infection or neoplasia were excluded. Strokes were subclassified according to the criteria of the National Survey of Stroke as due to ischemia (embolism or thrombosis), subarachnoid hemorrhage, intracerebral hemorrhage, or unknown cause.31 If no records could be obtained, strokes were considered probable if they required hospitalization and were corroborated by additional information provided by letter or interview.

Deaths were reported by next of kin, coworkers, postal authorities, or the National Death Index. Using all sources combined, we estimate that follow-up for the deaths was more than 98% complete.32 Fatal CHD was defined as fatal MI if this was confirmed by hospital records or autopsy, or if CHD was listed as the cause of death on the death certificate and this was the underlying and most plausible cause, and evidence of previous CHD was available. The statement of the cause of death on the death certificate was never relied on by itself as providing sufficient confirmation of death due to CHD. Sudden death within an hour of the onset of symptoms in subjects with no other plausible cause of death (other than coronary disease) was also included. Fatal stroke was also confirmed by medical records or autopsy reports, or considered probable if these were not obtainable but stroke was listed as the underlying cause on the death certificate.

Statistical Analysis

Participants contributed follow-up time from the return of the 1986 (men) or the 1980 (women) questionnaire up to the occurrence of a confirmed end point, death, or the end of follow-up (January 31, 1994, for men and June 1, 1994, for women). The relative risks (RRs) were calculated by dividing the incidence of CHD or stroke among men and women in various categories of egg consumption by the incidence among those in the lowest category of intake (<1 egg per week), adjusting for age (5-year categories). To adjust for other risk factors, we used pooled logistic regression,33 which is asymptotically equivalent to Cox regression for time-dependent covariates. Multivariate models included as covariates were (1) total energy intake (quintiles); (2) smoking (never, past, current 1-14, 15-24, and 25 cigarettes/d); (3) alcohol consumption (0-4, 5-9, 10-14, 15-29, and 30 g/d); (4) history of hypertension; (5) parental history of MI; (6) body mass index (calculated as weight in kilograms divided by the square of height in meters) (quintiles); and (7) current multivitamin use, and (8) vitamin E supplement use. In HPFS, we also adjusted for physical activity in metabolic equivalents per week (quintiles).34 In NHS, we adjusted for regular vigorous exercise (once or more per week) and menopausal status and postmenopausal hormone use. Tests of linear trend across increasing categories of egg consumption were conducted by treating the median values of consumption in categories (servings per day) as a continuous variable.
To reduce intra-individual variation and best represent long-term diet, we used repeated measures of diet in our primary analyses.35 In particular, CHD or stroke incidence was related to the cumulative average of egg consumption from all available questionnaires prior to the beginning of each 2-year follow-up interval. For example, in HPFS, CHD or stroke incidence during the 1986-1990 period was related to egg consumption assessed in 1986, while the incidence during 1990-1994 period was related to the average intake assessed in 1986 and 1990. For those who failed to complete the 1990 diet questionnaire (24%), we used the 1986 values to replace the missing values and a missing data indicator was included in the model to correct for potential bias. Average egg consumption at baseline was the same between the respondents and nonrespondents. In alternative analyses, we analyzed the incidence of CHD in relationship to egg consumption at baseline only. We also related the incidence to the most recent diet by using the baseline diet to predict incidence during 1986-1990, and the 1990 diet to predict incidence during 1990-1994. Because the diagnosis of hypercholesterolemia, diabetes, or hypertension may lead to changes in egg consumption and therefore confound the associations between egg consumption and risk of cardiovascular disease, we did not update diets for those who developed these conditions.35

RESULTS


In men, the average egg consumption declined from 2.3 eggs per week in 1986 to 1.6 eggs per week in 1990. In women, the average egg consumption declined from 2.8 eggs per week in 1980 to 1.4 eggs per week in 1990. At baseline, egg consumption was positively associated with smoking in men but inversely associated with smoking in women (Table 1 <../fig_tab/joc81683_t1.html>). Those with higher egg intake consumed more dietary cholesterol and protein but less carbohydrates. Egg consumption was positively associated with bacon intake among both men (r=0.35) and women (r=0.21). Men with higher egg consumption were more likely to consume whole milk, red meat, and bread, and less likely to consume skim milk, chicken, vegetables, and fruits. These relationships were less clear in women.

Association With CHD

We documented 866 incident cases of CHD in men during 8 years of follow-up (1986-1994) and 939 cases of CHD in women during 14 years of follow-up (1980-1994). The age-adjusted RR of CHD comparing more than 1 egg per day with less than 1 egg per week was 1.15 (95% confidence interval [CI], 0.85-1.55; P for trend=.37) for men and 0.85 (95% CI, 0.62-1.16; P for trend=.50) for women (Table 2 <../fig_tab/joc81683_t2.html>). After adjustment for smoking and other covariates, the corresponding RRs were 1.08 (95% CI, 0.79-1.48; P for trend=.75) for men and 0.82 (95% CI, 0.60-1.13; P for trend=.95) for women. Additional adjustment for dietary fiber intake had little impact on the RRs for women but it further attenuated the association for men (RR for >1 egg per day, 1.01; 95% CI, 0.74-1.38). Because of the relatively strong correlation between consumption of eggs and bacon, we further adjusted for bacon intake. The adjusted RRs across categories of egg consumption are less than 1 per week (1.0), 1 per week (1.00), 2 to 4 per week (1.04), 5 to 6 per week (0.78), and 1 or more per day (0.93) (P for trend=.36) for men; and less than 1 per week (1.0), 1 per week (0.81), 2 to 4 per week (0.96), 5 to 6 per week (0.91), and 1 or more per day (0.78) (P for trend=.73) for women. Additional adjustment for other foods including whole milk, fish, beef as main dish, chicken, or cereal had little impact on the results.

In further analyses egg consumption had no significant relationship with either fatal CHD or nonfatal MI (data not shown). After excluding cases with events occurring during the first 2 years of follow-up to reduce the effect of change in diet due to preclinical conditions, the results remained unchanged. In analyses excluding participants who reported changes in egg consumption in the previous decade from baseline, we observed a significant inverse association for men but a nonsignificant positive association for women. The results from analyses including eggs estimated from other foods were similar to those from the main analyses. We also observed no significant increase in risk when either only baseline or only recent egg consumption was analyzed. Only 4.8% women reported almost never consuming eggs and 1.6% women reported consuming 2 eggs or more per day. When these 2 extreme groups were compared (using never consumers as the reference group), the multivariate RR was 0.76 (95% CI, 0.43-1.35). The corresponding RR was 1.10 (95% CI, 0.67-1.79) for men.

In all above analyses, participants with diabetes or hypercholesterolemia at baseline were excluded. When these subjects were included in the analyses, the results did not appreciably change. The multivariate RRs as in Table 2 <../fig_tab/joc81683_t2.html> across categories of egg intake were less than 1 per week (1.0), 1 per week (1.05), 2 to 4 per week (1.04), 5 to 6 per week (0.96), and 1 or more per day (1.10) (95% CI, 0.83-1.45; P for trend=.78) for men; and less than 1 per week (1.0), 1 per week (0.87), 2 to 4 per week (0.98), 5 to 6 per week (1.03), and 1 or more per day (0.95) (95% CI, 0.74-1.22; P for trend=.54) for women.