Objective:To provide background on five HIV-infected cohorts with documented seroconversion times and serum immunoglobulin (Ig) A and β2-microglobulin (β2M), CD4+ cell count and haemoglobin levels. To give a relative risks (RR) regression summary of the prognostic value of serial CD4+ cell count, IgA, β2M and haemoglobin measurements for clinical AIDS, and to examine whether cofactors such as current age, sex and exposure category affect these RR.Design:The Multicohort Analysis Project (MAP) workshop was an international collaboration which brought statisticians, immunologists and clinicians from the five cohorts to work together for 10 days. A predefined restricted database was made available by each cohort for the workshop.Setting:The Medical Research Council (MRC) Biostatistics Unit, Cambridge, UK hosted the MAP workshop from 19 to 30 April 1993.Subjects:MAP workshop database comprised 1744 patients with documented HIV seroconversion times, with 407 women, over 900 injecting drug users (IDU) and over 500 homosexual men; 363 patients had AIDS and there were 308 deaths.Main outcome measures:Descriptive statistics on survival and progression to clinical AIDS by cohort and exposure category, CD4+ cell count at AIDS diagnosis and pre-AIDS zidovudine therapy. RR summarizing the joint prognostic significance of serial markers and cofactors such as age, sex and exposure category for progression to clinical AIDS.Results:Slower progression to AIDS for IDU [95% confidence interval (CD, 0.35–0.71 ] and heterosexuals (95% Cl, 0.19–0.98) compared with homosexual men was confirmed after adjusting for current age-group and serial CD4+ cell counts. CD4+ cell counts at AIDS diagnosis were much higher among homosexual men before than after 1988 (median, 150 and 90×106/l, respectively). Little zidovudine use was observed among AIDS cases diagnosed before 1988 (2%) but increased use was recorded after 1988 and 1989 (24%) and even greater use after 1990 (59%). Low serial CD4+ cell count, haemoglobin levels and high serum IgA and β2M levels were associated with an increased risk of progression to AIDS. CD4+ cell count always provided prognostic information in addition to other markers; IgA and P2M (95% Cl, 1.23–1.50 and 105–1.51, respectively) were jointly prognostic. P2M did not provide significant extra information (95% Cl, 0.91–1.47) to the combination of serial CD4+ cell count and IgA, although haemoglobin did (95% Cl: 0.74–0.91 for 10g/l increase in haemoglobin). Interactions between cofactors, particularly exposure category and serial markers, were used to test for modifications in RR. The association between AIDS risk and serial CD4+ cell count was weaker, and with elevated IgA stronger, for homosexual men; RR associated with high β2M values were.lower for IDU, in whom β2M may be elevated for reasons other than HIV disease.Conclusions:IgA and β2M, which can be measured in small volumes of stored blood, are jointly predictive of progression to AIDS. Results were broadly consistent between cohorts representing different age-groups, seroconversion periods and exposure categories. Some regression effect modifications by exposure category were noted, however, which merit further independent study.