Survival time from diagnosis of candidemia: An application of survival methods for epidemiology to the Mycoses Study Group multi-center observational study of hospitalized patients with candidemia

Candidemia is recognized as an important and potentially lethal infectious disease. Patients with these infections are reported to have increased lengths of hospital stay, high morbidity and mortality, and increased health care costs compared to patients without such infections. Major changes in the epidemiology of candidemia have occurred during the last twenty years. Infection incidence has increased by as much as 500%, and there has also been a major shift in the distribution of the Candida species causing candidemia infections. The traditionally predominant species, C. albicans, is now being replaced by the more obscure non- albicans species. The impact of this changing epidemiology of candidemia on patient morbidity and mortality is not well understood.; The present study analyzes the relationship between species-specific candidemia and the occurrence of death during three months of post-diagnosis follow-up while controlling for the effect of confounding variables. Data used for this study come from the National Institutes for Health, Infectious Disease National Institute for Allergies and Infectious Disease, Mycoses Study Group study number 34---a large multi-center, prospective, observational study of hospitalized patients with candidemia conducted in the United States between February 1995 and November 1997. A total of 1,570 cases were identified with one or more positive blood cultures for Candida, 657 (42%) deaths occurred during the three month follow-up period. Proportional hazards modeling was performed for each species group of Candida identified. Differences in the risk factor variables identified within each group and estimates of the reported hazard ratios were compared. Species-specific rates of death at the end of the three-month follow-up period were compared by calculating Kaplan-Meier estimates of survival time.; Descriptive statistics revealed that non-albicans species were predominantly (64%) the cause of candidemia. The most frequently isolated non-albicans species was C. glabrata (25%) followed by C. tropicalis (14%), C. parapsilosis (14%) and C. krusei (3%).; Cox proportional hazards regression identified unique species interactions with the predictor variables. Of particular interest was the elevated hazard ratio for HIV infection among patients with C. albicans (HR = 1.9, 95% CI: 1.5--4.3), and cancer (HR = 1.7, 95% CI: 1.0--3.0) and for prophylactic systemic antifungal therapy among patients with C. tropicalis. After adjusting for all the identified significant causes of death Kaplan-Meier estimates of survival during the 3-month observation period for untreated candidemia were lowest for patients with C. glabrata (65%) and C. tropicalis (77%), and highest for patients with C. albicans (88%) and C. parapsilosis (94%) group. The descriptive results confirm the epidemiologic shift towards non-albicans species. They also suggest that species-specific risk factors for mortality exist, that candidemia is an independent cause of mortality, and that survival rates are different among the Candida species evaluated. Not all candidemia infections are the same.; Treatment and intervention strategies that aim to reduce mortality should consider not only the underlying disease status of the patient but also the species of Candida causing infection, as well as potential interactions of these two factors.