13 18. Miller G, Shope T, Lisco H, Stitt D, Lupman M Epstein-Barr virus: Transformation, cytopathic changes, and viral antigens in squirrel monkey and marmoset leucocytes. Proc Nat Acad Sci USA 1972, 69: 383-87 19 Henle G, Henle W Immunofluorescence in cells derived from Burkitt’s lymphoma. J Bact 1966; 91: 1248-56 20 Reedman BM, Klein G Cellular localisation of an Epstein-Barr virus (EBV)-associated complement-fixing antigen in producer and non-producer lymphoblastoid cell lines. Int J Cancer 1973; 11: 499-520 21 Rickinson AB, Wallace LE, Epstein MA HLA-restricted T cell recognition of Epstein- Barr virus-infected B cells. Nature 1980, 283: 865-67. 22. Rickinson AB, Moss DJ, Pope JH Long-term T cell-mediated immunity to Epstein Barr virus in man. II. Components necessary for regression in virus-infected leucocyte cultures Int J Cancer 1979, 23: 610-17. 23. Bird AG, McLachlan SM Cyclosporin A and Epstein-Barr virus. Lancet 1980, ii: 418. 24 Svedmyr E, Jondal M. Cytotoxic effector cells specific for B cell lines transformed by Epstein-Barr virus are present in patients with infectious mononucleosis. Proc Nat Acad Sci USA 1975, 72: 1622-26. 25 Epstein MA, Achong BG Recent progress in Epstein-Barr virus research Ann Rev Microbiol 1977, 31: 421-45 VASECTOMY AND NON-FATAL MYOCARDIAL INFARCTION ALEXANDER M. WALKER JUDITH R. HUNTER RICHARD N. WATKINS HERSHEL JICK ANNE DANFORD LESLIE ALHADEFF KENNETH J. ROTHMAN Boston Collaborative Drug Surveillance Program, Boston University Medical Center, Waltham, Massachusetts; Group Health Cooperative of Puget Sound, Seattle, Washington; Department of Biostatistics and Epidemiology, Sidney Farber Cancer Institute, Boston, Massachusetts; and-Department of Epidemiology, Harvard School of Public Health, Boston, U.S.A. Summary The incidence of non-fatal myocardial in- farction among 4830 vasectomised men was 0·9 cases per 1000 man-years during 24 420 man-years of observation. This was slightly lower than the rate in 24 150 non-vasectomised men, matched with a vasectomised man for calendar year of birth and duration of observation. Review of medical records for a matched sample of study subjects in- dicated no measurable confounding by important cardiac risk factors. Introduction VASECTOMY is an effective, convenient method of con- traception which is popular in the U.S.A. and elsewhere. Although the surgical procedure seems to be safe, and serious short-term complications are uncommon, there is little de- tailed information on long-term medical consequences of vasectomy. Vasectomisd macaques had accelerated rates of atherogenesis both when fed high-fat diets’ and when fed regular monkey chow.2 There is as yet no evidence of any in- crease in rates of atherosclerotic disease among vasectomised men3 (Goldacre M. J., Clarke J. A., Heasman M. A., Vessey M. P. Unpublished) but experience is still very limited. There has been no attempt to control for confounding by in- dividual characteristics which might be associated with both vasectomy and atherosclerotic disease. The present study, part of an examination of all hospital admissions in men with vasectomy, was undertaken to investigate the risk of non-fatal myocardial infarction among vasectomised men. Subjects and Methods The men whose medical histories were analysed for this study were members of Group Hospital Cooperative of Puget Sound (GHC) based in Seattle, Washington. GHC is a consumer-owned health maintenance organisation which has been in operation since 1947 and currently has approximately 280 000 members. In- dividuals in the Seattle area can join GHC through contracts provid- ed by local employers and unions, through individual contracts, or through Medicare and Medicaid. The age and sex distribution of plan members reflects that of the U.S.A. population, but the pro- portion of younger and middle-aged adults is somewhat larger. GHC provides comprehensive medical care to its members, who receive outpatient services in scattered clinics and hospital treat- ment in two hospitals owned by the cooperative. Since 1972 records of hospital discharges have been computerised through the Com- mission on Professional and Hospital Activities-Professional Activi- ty Study in Ann Arbor, Michigan. All medical records for current GHC members are readily accessible; records of non-current members cannot be reliably obtained. Segments of the vas deferens are routinely sent to the pathology department when a vasectomy is performed. This enabled us to identify a group of vasectomised men by undertaking a hand review of all pathology department records, noting members for whom a vas specimen was submitted after routine vasectomy between Jan. 1, 1963, and Dec. 31, 1978. The record numbers of men so identified were then checked in central membership files, and classed as "cur- rent" or "non-current" members. We identified 4830 current members and 1339 non-current members who had had a vasectomy. The vasectomised current members identified represented 7 - 6% of the total GHC male population aged 25-64 in 1979. Of these men, 514 had had vasectomy before 1971, 2009 in 1971-74, and the re- maining 2307 in 1975-78. For each of the 4830 identified current members with a vasec- tomy, we selected five current members matched to the vasectomis- ed man on sex, year of birth, and membership in the plan at the time the vasectomy had been performed. We realised at the time of selec- tion that there would be some vasectomised men in the "non- vasectomised" group. The national prevalence of vasectomy in 1976 was 10 - 57o among husbands of women in the age range 15-44 years,.t and this was regarded as an acceptable level of misclassifi- cation in our non-vasectomised group. Data analysis proceeded in two steps: 1. We calculated the number of man-years at risk after vasectomy in the vasectomised group from 1972 through 1979, the years covered by computerised data on hospital admission. Similar calculations were applied to data on non-vasectomised men, coun- ting as the beginning of observation the date of vasectomy for the vasectomised man to whom each non-vasectomised man was match- ed. Because of the matching procedure, the distribution in the vasectomised and non-vasectomised groups of years of observation by age and by elapsed time since vasectomy (or corresponding matched date for the non-vasectomised) was identical. Discharge diagnoses of myocardial infarction (codes 410-0-410-9 9 in the In- ternational Classifications of Diseases, eighth revision) in the vasec- tomised and non-vasectomised cohorts were then noted and rates of acute myocardial infarction calculated on the basis of only the first such diagnosis recorded for each individual. Rates were cross- tabulated by age and years elapsed since vasectomy. "Exact" con- fidence intervals for rate ratios were calculated by a binomial model for the distribution of cases between exposure groups.5 2. We evaluated the extent of confounding in the myocardial in- farction rates obtained in step 1. To do this we derived a measure of the effect of vasectomy, corrected for confounding characteristics. The matched nature of the study design permitted an efficient system of obtaining this correction,b by means of a Cox proportional hazards model with stratification by matching group. Proportional hazards analysis is a flexible, multivariate life-table procedure which yields statistics which are estimates of the factors by which a baseline incidence rate is multiplied in the presence of various predictive characteristics. A consequence of the equations defining these estimates is that matched sets in which no event (in this study, myocardial infarction) occurs drop out of the analysis.6 Thus, the presence or absence of various confounding characteristics needed to be determined only for those matched sets in which at least one man had a myocardial infarction. To reduce the numbers of records to be abstracted without introducing systematic bias into the data,