Serum Cholesterol Levels and In-Hospital Mortality in the Elderly Graziano Onder, MD, PhD, Francesco Landi, MD, PhD, Stefano Volpato, MD, MPH, Renato Fellin, MD, Pierugo Carbonin, MD, Giovanni Gambassi, MD, Roberto Bernabei, MD PURPOSE: Although total cholesterol levels among middle- aged persons correlate with long-term mortality from all causes, this association remains controversial in older persons. We ex- plored whether total cholesterol levels were independently as- sociated with in-hospital mortality among elderly patients. METHODS: We analyzed data from a large collaborative ob- servational study, the Italian Group of Pharmacoepidemiology in the Elderly (GIFA), which collected data on hospitalized pa- tients. A total of 6984 patients aged 65 years or older who had been admitted to 81 participating medical centers during four survey periods (from 1993 to 1998) were enrolled. Patients were divided into four groups based on total cholesterol levels at hospital admission: 160 mg/dL (n = 2115), 160 to 199 mg/dL (n = 2210), 200 to 239 mg/dL (n = 1719), and 240 mg/dL (n = 940). RESULTS: Patients (mean [ SD] age, 78  7 years) were hos- pitalized for an average of 15  10 days. The mean total choles- terol level was 186  49 mg/dL. A total of 202 patients died during hospitalization. Mortality was inversely related to cho- lesterol levels (160 mg/dL: 5.2% [110/2115]; 160 –199 mg/dL: 2.2% [49/2210]; 200 –239 mg/dL: 1.6% [27/1719]; and 240 mg/dL: 1.7% [16/940]; P for linear trend 0.001). After adjust- ment for potential confounders (demographic characteristics, smoking, alcohol use, indicators of nutritional status, markers of frailty, and comorbid conditions), low cholesterol levels con- tinued to be associated with in-hospital mortality. Compared with patients who had cholesterol levels 160 mg/dL, the odds ratios for in-hospital mortality were 0.49 (95% confidence in- terval [CI]: 0.34 to 0.70) for participants with cholesterol levels of 160 to 199 mg/dL, 0.41 (95% CI: 0.26 to 0.65) for those with cholesterol levels of 200 to 239 mg/dL, and 0.56 (95% CI: 0.32 to 0.98) for those with cholesterol levels 240 mg/dL. These esti- mates were similar after further adjustment for inflammatory markers and after excluding patients with liver disease. CONCLUSION: Among older hospitalized adults, low serum cholesterol levels appear to be an independent predictor of short-term mortality. Am J Med. 2003;115:265–271. ©2003 by Excerpta Medica Inc. T otal cholesterol is an independent predictor of long-term mortality from all causes in middle- aged persons, but this association is controversial among the elderly (1–7). Several studies have suggested that high cholesterol levels provide a mortality advantage over lower levels among persons aged 65 years or older (2,4 – 6). In contrast, others have reported a direct (7) or a U-shaped relation (3) between total cholesterol levels and mortality. A major difference between middle-aged and older adults is the greater prevalence of coexisting disease and poor health among the elderly (8). Older frail persons with a high burden of disease may have lower cholesterol levels and greater mortality than those with few (or no) medical problems who have low-to-normal cholesterol levels (9,10). Indeed, total cholesterol levels tend to de- crease among older men and women, and this trend may correlate with the effects of poor health (11). Thus, frailty may confound the relation between cholesterol levels and long-term mortality in older patients. In previous studies, clinical parameters of malnutrition (low levels of serum albumin [12], lymphocytopenia [13], and low body mass index [14]) have independently predicted mortality among hospitalized older patients. In another study, markers of inflammation—which were associated with low levels of total cholesterol—predicted in-hospital mortality in the elderly (15). We studied whether total cholesterol levels predict all- cause in-hospital mortality, independent of frailty and comorbidity, or whether it serves only as a marker of health status, among hospitalized elderly patients. METHODS The Gruppo Italiano di Farmacoepidemiologia nell’Anziano (GIFA), a group of investigators who are based in community and university hospitals throughout Italy, periodically surveys drug use, occurrence of adverse drug reactions, and quality of hospital care (16,17). Briefly, all patients admitted to 81 clinical centers in Italy were enrolled and followed until discharge. The study periods were May 1 to June 30 and September 1 to From the Department of Geriatrics (GO, FL, PC, GG, RB), Catholic University of the Sacred Heart, Rome, Italy; Department of Geriatrics (GO), J. Paul Sticht Center on Aging, Wake Forest University School of Medicine, Winston-Salem, North Carolina; Second Department of Clinical and Experimental Medicine (SV, RF), University of Ferrara, Italy; and Center for Gerontology and Health Care Center (GG), De- partment of Community Health, Brown University Medical School, Providence, Rhode Island. Requests for reprints should be addressed to Graziano Onder, MD, PhD, Centro Medicina dell’Invecchiamento, Universita ` Cattolica del Sacro Cuore, Policlinico A. Gemelli, L.go F. Vito 1, 00168 Rome, Italy, or graziano_onder@rm.unicatt.it. Manuscript submitted December 11, 2002, and accepted in revised form May 5, 2003. © 2003 by Excerpta Medica Inc. 0002-9343/03/$–see front matter 265 All rights reserved. doi:10.1016/S0002-9343(03)00354-1