Cross-sectional studies in AIDS pathogenesis: how far can they mislead us?

When investigating variables thought to be of potential importance to AIDS pathogenesis, it is common practice to initially carry out a cross-sectional study comparing those with symptomatic disease/low CD4 counts with those with no symptoms/high CD4 counts. While it is widely appreciated that such studies have weaknesses compared with those involving patient follow-up, the particular biases likely to arise in the context of HIV infection have not been formally evaluated. In an attempt to do this, data on HIV progression for 50,000 notional patients who became infected in the years between 1976 (assumed to be the start of the epidemic) and 2010 were simulated under various scenarios. In Scenario 1, we considered a variable called X, which was specified to be associated with more rapid progression of HIV infection. In Scenario 2, we considered a variable Y, which was specified not to be associated with more rapid progression but which tended to change in value as a result of HIV-induced immunosuppression. Variable Z, in Scenario 3, was specified to be associated with more rapid progression and to change in value when severe immunosuppression developed. Cross-sectional analyses evaluating the association between variables X, Y, and Z and the CD4 count/presence of AIDS were performed for data as of 1980, 1985, 1995, 2000, 2005, and 2010 to assess the extent to which the true role of the three variables could be ascertained. For Scenario 1, cross-sectional analyses performed on these data later in the epidemic falsely suggested that variable X was not related to HIV progression. Variable Y in Scenario 2 falsely appeared to be related to HIV progression, due to the inability of the cross-sectional design to distinguish the fact that changes in variable Y were as a result of immunosuppression and not a possible cause of it. In Scenario 3, variable Z showed a relationship with the presence of AIDS for which the typical interpretation would lead to the suggestion that its effect is in the reverse direction to that which was, in fact, the case. In conclusion, as the HIV epidemic progresses, cross-sectional studies will increasingly tend to give misleading results. Most importantly, it is incorrect to conclude that variables found to have no association with HIV disease stage in such studies are not of potential importance to pathogenesis.