DIMACS/BIOMAPS Seminar Series on Quantitative Biology and Epidemiology.

Title: Estimating the Impact of P. Carinii Prophylaxis on HIV Disease

Speaker: Donald Hoover, Rutgers University

Date: Wednesday, March 17, 2004, 1:00 pm

Location: Hill Center, Room 260, Rutgers University, Busch Campus, Piscataway, NJ


Without intervention, about 50% of AIDS patients develop P. carinii pneumonia (PCP), an AIDS defining condition that is often fatal. In 1989, the Food and Drug Administration approved use of prophylactic antibiotics by immunosuppressed HIV patients to prevent PCP. Short term clinical trials had shown that P. carinii prophylaxis prevented and delayed PCP, but did not quantify the full impact of P. carinii prophylaxis on HIV disease outcomes. We quantified this impact on a cohort of ~2500 HIV infected which had been followed starting from 1984 through 1995.

This talk describes: 1. Biomedical details of P. carinii prophylaxis. 2. The analytical formulation of costs/benefits of P. carinii prophylaxis we used. 3. The statistical approach taken to estimate P. carinii costs and benefits within this formulation, and 4. The findings, implications and limitations of the data analysis.

Benefits of P. carinii prophylaxis are a function of: (i) The number of PCP cases it prevents. (ii) The amount of post-PCP morbidity it reduces, and (iii) The amount of time it extends survival. The costs of P. carinii proportional to: (iv) The length of time it must be delivered. The quantities described by (i), (ii), (iii) and (iv) were estimated by comparing men who became immunosuppressed (CD4 < 100/?L) from 1989 - 1995 and used P. carinii prophylaxis to those who reached this level 1988 or earlier (before P. carinii prophylaxis was available). We truncated the analysis in 1996 since new antiretroviral therapies known as HAART which greatly extended lifespan became available that year. The quantities (i) ? (iv) are functions of transiting probabilities through four stages: A ? Alive without PCP, B ? Alive developed PCP, C ? Dead never developed PCP, and D ? Dead after PCP. Two sandwiching three stage multiple decrement models were used to estimate transition probabilities and times through these stages. Computational limitations and fallacies prevented use of Markovian models.

The analysis suggested that prior to the advent of HAART P. carinii prophylaxis would be used on average 2.35 years per individual. This use (prior to the advent of HAART) prevented PCP in 21.7% of users, reduced post PCP morbidity an average of 0.32 years per user and extended life span on average 0.37 years per user. Epidemiological limitations to the findings include use of historical controls and potential non-generalizability to other HIV populations.

Seminar sponsored by DIMACS/BIOMAPS Seminar Series on Quantitative Biology and Epidemiology.