Contrastación de la Hipótesis Especifica 2

In this chapter I have studied three markers which are time-dependent (in that they change over time) and which can be analysed as such because repeated measurements are available on patients over time. Two further markers are fixed in that patients only have measurements at certain, chosen time points.

Of the time-updated measurements, both the IgA level and the development of p24 antigenaemia appear to offer additional prognostic information to that provided by the CD4 count. The CDS count, on the other hand, simply identifies those with the lowest and most rapidly declining CD4 counts and as such does not provide any prognostic information once the patient’s CD4 count is known.

Of the fixed covariates studied, the B2M level appears to be a useful marker of future disease progression at high CD4 counts, suggesting that this marker may be useful for establishing long-term prognosis early in infection. The lo s s of antibodies to p24, however, simply identified patients with the lowest CD4 counts soon after seroconversion and as such did not provide any additional prognostic information.

When time-updated covariates were adjusted for each other in a multivariate model which also included the patients' age at seroconversion and the development of bacterial infections, the CD4 count appeared to be the most consistent prognostic marker, being significantly associated with all three clinical endpoints. Both the IgA level and the development of p24 antigenaemia were associated with early clinical events, being associated with the development of the first HIV-related event and AIDS, but not with death. The patient’s age, in contrast, was associated with only the more major events of AIDS and death.

The development of a bacterial condition prior to the development of AIDS continued to suggest that a patient’s condition was likely to deteriorate, irrespective of their immune status. The rate of CD4 decline continued to add some additional prognostic information, although this became marginally non-significant for progression to AIDS. 6.7.2 CD8 lymphocyte counts

The finding that CD8 counts are raised during HIV infection has often been reported®^'®°’^^’^®’^®®’^®^'^^^. A drop at late stages of infection has been suggested by both Raska at and Brinchmann at who found that CDS counts were lower in

AIDS patients than in HIV-negative controls. Buchbinder et found that healthy long­ term survivors had higher CD8 counts than rapid progressors, further confirming the possibility of a drop in CD8 counts at late stages of disease. However, findings from two studies^^^’^^® suggest that any decline, if it exists, is likely to be small. Through the most part of infection, however, there appears to be no large change in CD8 counts^^’^^®. I have found that there is a general trend for counts to increase over the first few years after infection, before dropping to low levels. Whilst multi-level models are an improvement on population-based methods for studying the pattern of CD8 change, there is still the possibility for the selective drop out of patients who die to have an effect. For example, if the CD8 counts of patients who die drop rapidly in the year prior to death, then the overall estimate of the CD8 count will be ‘pulled down’ as patients die. As no outcome is attached to patients in the multi-level modelling it has not been possible to consider whether there is a different pattern in those who die than in those who remain alive, de Stavola^®^ has suggested a method for considering this by censoring patients at different time points and then considering the changing patterns in those alive and those dead at each time point. However, with the relatively small number of deaths in this study, it is unlikely that this would reveal any major differences. The role of the CD8 count in assessing patient prognosis remains unclear. Some authors have suggested that it has additional prognostic value to the CD4 count®^’®^, but others find that it has no independent prognostic effect^®’®^’^^®’^^"^ or that it may only be predictive of AIDS close to the time of diagnosis®®. After adjustment for the CD4 count, the relative hazard associated with a 100 cell/mm® drop in the CD8 count has been reported to range between 0.93 and 0.96 when treated as fixed at some baseline®®®’®^® and 0.99 as a time-updated covariate (Ms A Mocroft, personal communication). Chevret®®® reported a non-significant increase in the progression rate to a CD4 count of 200 cells/mm® when the CD8 count rose above 1500 cells/mm®.

In a recent paper from this cohort®®®, the CD8 count was shown to be a useful marker of long-term disease progression and CD4 decline when measured soon after seroconversion, indicating that the immune activation seen very soon after the time of infection may be linked with the process of CD4 decline. This finding is of limited use for patients who do not usually present at the time of infection. In these patients, once their most recent CD4 count is known then it would appear that knowledge of their CD8 count does not add any further prognostic information in the short-term.

6.7.3 B2M

It has been suggested that B2M levels are generally higher in HIV positives than in HIV negatives^^'^^^'^^^.342,344,351,358.372 paper in particular has suggested that B2M

increases at a fairly slow rate until a year before AIDS diagnosis when it then increases rapidly^®'^. In contrast, a recent paper analysing routine B2M levels from five cohorts, including this cohort, suggested that B2M rises over the first 6 years of infection before dropping over later years^®®. However, this may be an artefact of the inclusion of data from cohorts who each measured B2M over different time periods. Unfortunately, routine follow-up data on B2M is not available for many patients in this cohort, so I am not able to consider this.

As a result of including only men with measurements taken within six months of their CD4 count falling to each level, the samples of men included in this analysis are small but homogeneous in terms of their CD4 count. At high CD4 counts the B2M level provides additional prognostic information for the development of AIDS to that provided by the CD4 count. At lower CD4 counts, however, the B2M level simply identifies those whose CD4 counts will drop more rapidly in the future and are, therefore, more likely to develop clinical disease.

Figures 6.5 and 6.6 provide a convenient tool for clinicians who may use these figures to assess long-term prognosis and who may, for example, choose to delay treatment in patients with low CD4 counts and low B2M levels whilst considering the early introduction of treatment in patients with high CD4 counts but high B2M levels. Although differences in B2M levels between exposure groups exist^^^, this does not invalidate the use of B2M as a prognostic marker. However, the cut-offs quoted in this thesis may need to be validated in other exposure groups before use.

The reported value of B2M as a prognostic marker independently of the CD4 count varies in the literature. In two independent analyses of data from the Los Angeles centre of the Multicenter AIDS Cohort Study, both Fahey^^^ and Hofmann^^^ suggested that the predictive power of B2M and CD4 counts were of a similar size and were independent. This independence has also been shown by other a u t h o r s ^ a l t h o u g h some suggest that any predictive power disappears after adjustment for the CD4 count^®°. Interestingly, the former studies were carried out using fixed covariates only. Alcabes et on the other hand, used time-updated values, suggesting that the B2M level may indicate those whose CD4 counts will decline rapidly, a finding confirmed by Sheppard at aP V

Some authors suggest that neopterin, which is highly correlated with has a stronger predictive power than Neopterin is not measured routinely in this cohort and therefore it is not possible to assess which has the greater predictive power.

The introduction of zidovudine has been shown to have a beneficial effect on B2M levels in patients^®®. I have previously shown that these results are, however, essentially unchanged when patient follow-up was censored in 1987, prior to the availability of zidovudine in the (See Appendix III - Published Papers).