Survival analyses of adult AIDS cases in Riverside and San Bernardino counties, California, between 1993 and 1999

摘要

In recent years, a longer survival after AIDS diagnosis and a sharp decline in AIDS incidence and deaths have been observed in the United States. This study is a follow-up study of 3,578 reported AIDS cases to the AIDS surveillance system in Riverside and San Bernardino Counties, California between January 1993 and December 1999. We compared our results using 5 analysis methods: Kaplan-Meier survival, time trend analysis, Cox regression, Spline survival regression, and multivariate multiple decrement life-table analysis program (MMDLAP).;Our study's results using Cox regression show that cases with a history of antiretroviral therapy (ART), younger age, and higher CD4+ T lymphocyte at AIDS diagnosis are significant predictors of AIDS survival. Survival improves with calendar time in all groups. Cases with a history of ART showed a hazard ratio (HR) of 0.62 for cases with CD4+ T lymphocyte count of 50 cells/mul (95% CI: 0.48, 0.79), HR of 0.57 for cases with 100 cells/mul (95% Cl: 0.45, 0.72), and HR of 0.49 for cases with 200 cells/mul (95% CI: 0.31, 0.77) at the age of 40 years after controlling for significant confounders. Survival appears to improve with secular time. For every 10 years increase in age at diagnosis, HR increased by 26 percent (HR 1.26, 95% CI: 1.02, 1.56) and for every 50 cells/mul increase in CD4+ T lymphocyte count, HR decreased by 31 percent (HR 0.69, 95% CI: 0.59, 0.82) in the group that did receive ART treatment. Compared to AIDS cases diagnosed in 1993, cases diagnosed in 1994-95 had an HR of 0.77 (95% CI: 0.62, 0.94); cases diagnosed in 1996--97 had an HR of 0.40 (95% CI: 0.31, 0.52); cases diagnosed in 1998--99 had an HR of 0.48 (95% CI: 0.35, 0.66).;Traditional Cox regression gives estimates of survival as a monotonic exponential function in a continuous Cox regression, and a step function in categorical Cox regression while Spline regression offers a smooth representation of a complex pattern of changes of hazard ratio (HR). Moreover, categorical Cox regression overestimates HR in the beginning and underestimates it in the end compared to Spline regression. MMDLAP is intuitive in estimating risk of death and mean survival time in defined time periods and also estimates risk ratio (RR) as HR with values similar to Cox regression.;Use of potent ART in a community setting is as effective as it has been in clinical trials. The effectiveness of antiretroviral therapy on survival can be observed from surveillance data collected at the community level.