© F e r r a t a S t o r t i F o u n d a t i o nveterans who underwent splenectomy with a follow-up of up to 27 years. In this cohort, we assessed patterns of hospitalization for infections, thromboembolism, and malignancies following splenectomy.
Methods
Study populationBased on USA census data, an estimated 30 million veterans were entitled to admission to VA hospitals during the study period. 33 The VA database has been previously described. 34,35 In the present study, splenectomized patients were identified from the hospital discharge summary records [coded in the 8 th and 9 th revisions of the USA version of the International Classification of Diseases (ICD): ICD8 45.1, ICD9 41.2, 41.43, 41.5] and were included in the study (n=8,149). To minimize the influence of reverse causality (i.e., undetected cancer requiring the splenectomy), all analyses were restricted to individuals whose first VA discharge with a splenectomy occurred at least 1 year prior to the first hospitalization listing a diagnosis of cancer. Thus, patients were followed from 1 year after their initial hospital discharge until the first discharge diagnosis of infection, thromboembolism, malignancy, death, or end of study, whichever occurred first. The time to develop infection, thromboembolism, or malignancy (i.e., latency) was estimated by subtracting the date of discharge from the first hospitalization listing a splenectomy from the date of the first hospitalization listing a diagnosis of infection, thromboembolism, or malignancy.Dates of death were ascertained from record linkage to Social Security Administration mortality files. With such linkage, death reporting is believed to be 96% complete. 36 Among the 8,149 splenectomized men who were selected for the study, 6,731 were eligible for matching to the National Death Index (alive as of January 1, 1979). In addition, a random sample (n=6,731) of the eligible VA cohort was selected to match these patients on the basis of race and year of birth (1:1 sample). The National Death Index provided death certificate matching for the men who had undergone splenectomy and for the matched controls (n=13,462).
Statistical analysisRelative risk (RR) estimates and 95% confidence intervals (CI) were calculated using time-dependent Poisson regression methods for cohort data. 37 Calculations were performed using the AMFIT program (Epicure Version 2.0; HiroSoft International Corporation, Seattle, Washington, USA). All risk estimates were adjusted for attained age (<40, 40-49, 50-59, 60-69, 70-79, 80 or more years) and calendar year (1969-1974, 1975-1979, 1980-1984, 1985-1989, 1990-1996), race (African-American or white), number of hospital visits (1-2, 3-4, 5 or more), and time between study entry and exit (2-3, 4-5, 6-9, 10-14, 15 or more years). Risk estimates for buccal, esophageal, and liver cancers were also adjusted for a hospital discharge diagnosis of alcohol-related disorders because inclusion of this variable in the regression models resulted in a >10% change in the risk estimates. Adj...