PROCEDIMIENTO RECOLECCIÓN DE DATOS

In this thesis I have presented the findings from an epidemiologic follow-up of a well- characterised cohort of 111 HIV-positive men with haemophilia registered at the Royal Free Hospital Haemophilia Centre. The men were infected with HIV between 1979 and 1985 after treatment with infected blood products and dates of seroconversion can be estimated for all of the men. As the men are followed closely for the development of HIV disease, they are an ideal group in which to study the progression of HIV disease and to identify factors associated with disease progression.

By 14 years after seroconversion to HIV, 47 of the men have developed AIDS and 45 have died, the majority of deaths (93%) being either directly or indirectly related to HIV infection. The median time to the development of AIDS in this cohort is now 13.1 years after seroconversion, somewhat longer than that reported in other cohorts of HIV- infected individuals. This may partly be due to the long follow-up time in these individuals but may also be a result of careful patient monitoring and the use of prophylaxis and zidovudine once the CD4 count falls to a low level. Prior to AIDS the vast majority of patients (73.9%) have developed some manifestation of their HIV infection, the development of the first HIV-related event occurring an average of only 6.8 years after seroconversion. These conditions may include the development of bacterial infections, skin complaints, thrombocytopenia, oral Candida or herpes zoster.

The CD4 lymphocyte count declines in a quadratic manner over time, with patients becoming increasingly likely to develop AIDS or die once their CD4 count has fallen to a very low level. The CD4 count has been confirmed as a good prognostic marker for the development of AIDS, death and minor clinical events in this cohort. The rate of CD4 decline provides some additional prognostic information, suggesting that individuals whose CD4 counts have dropped to low levels rapidly are more likely to develop disease than those whose counts have dropped to the same low levels less rapidly.

As patients included in the cohort are of a wide age range, this cohort offers a good opportunity to study the effect of age on disease progression. Disease progression, especially progression to death, is more rapid in individuals who are older at the time of seroconversion, and this age effect cannot totally be explained by more rapidly declining CD4 counts in these individuals. Patients who are thought to have seroconverted either

prior to 1981 or from 1983 onwards also have faster disease progression. Whilst improved clinical care for HIV-infected patients has undoubtedly resulted in a gradual lengthening of incubation periods overtime, increased awareness of AIDS and changes in the AIDS definition may also have artefactually shortened incubation periods in some patients as clinicians diagnose patients with AIDS sooner.

The wide variation in incubation periods amongst patients in the cohort, even amongst those with similar CD4 counts, means that the search for additional markers of disease progression continues. With this in mind, a number of other routinely measured laboratory markers have been studied in this thesis. A drop in the CD8 lymphocyte count at first appears to provide prognostic information for HIV disease progression. However, at late stages of disease, patients with the lowest CD4 counts also tend to have low CD8 counts. Consequently, the CD8 count simply identifies these individuals. After adjusting for the CD4 count, the CD8 count does not, therefore, provide any independent prognostic information. As IgA levels rise throughout infection it appears that they may offer some additional prognostic information. Despite the finding that IgA levels are correlated with the CD4 count, and as such may be expected not to provide any independent information on prognosis, the IgA level does appear to be a useful prognostic marker for the development of clinical disease. The B2M level may also offer some additional prognostic information for the development of clinical disease when patients have relatively high CD4 counts, although it is debatable whether any additional information is provided at lower CD4 counts. The development of p24 antigenaemia, traditionally one of the first indications of HIV infection, heralds a deterioration in the patient's clinical condition, even after considering all other markers of disease progression. In addition, the development of bacterial infections is also a bad sign, with patients progressing more rapidly after the development of such conditions.

In general CD4 counts are measured every three to six months on patients in the cohort. However, their measurement has become more frequent as the value of the CD4 count has become recognised, and also as patients become ill and attend the Centre more frequently, resulting in irregular time intervals between measurements. Despite improvements in laboratory methods, the CD4 count is still an imperfectly measured marker and this may have implications for the value attributed to it as a prognostic marker. A simple model for the measured CD4 count has been presented in this thesis. The use of this model to simulate CD4 data for a large number of individuals has provided a simple way of studying the effects of both missing values and of variability on the relative hazard associated with a drop in the CD4 count. The findings in this thesis

could be applied to any other condition where a laboratory marker changes almost deterministically over infection and where the prognosis of the patients depends fully on this marker. As the CD4 count becomes more variable, the relative hazard estimate drops towards one. In the presence of a very poorly measured CD4 count, the value of the CD4 count may therefore be vastly underestimated and this may lead to incorrect conclusions about the role of this marker in HIV pathogenesis.