Supervisors info:
Πανταζής Νικόλαος, Διδάκτορας, Ιατρική, ΕΚΠΑ
Constantin T Yiannoutsos, Professor, Biostatistics, IU
Kara K. Wools-Kaloustian, Professor, Medicine, IU
Summary:
Human immunodeficiency virus (HIV) has infected more than 70 million people since the outbreak
of the epidemic, and has caused 35 million deaths. On the other hand, alcohol is consumed throughout
the world and 3.3 million deaths as well as 5.1% of the world's burden of disease is associated with or
is attributed to the use of alcohol. According to studies so far, HIV and alcohol are correlated. The
course of patients enrolling in a HIV-care program could be significantly affected by the concurrent use
of ethyl alcohol, and the recording of such information could be relevant to predict the likely course of
patients within such a program. This study examines the potential effects of alcohol consumption on
the course of HIV patients in HIV care by looking at the difference on the incidence of death, treatment
(ART) initiation and treatment (ART) eligibility for different levels of alcohol consumption.
This observational study is conducted at five clinics in Africa (Kenya and Uganda), and involves
765 adult HIV-infected patients taking part in a care program. The patients were monitored for 433
days starting on the day of joining the program. Three levels of exposure to alcohol were established
based on the World Health Organization's AUDIT questionnaire, depending on the AUDIT score:
hazardous drinking for an AUDIT score of above 7, hyper drinking for a score of above 15, and binary
drinking for a score of above 0. In addition, characteristics such as the WHO stage of the infection
upon entry into the study and the number of CD4 T helper lymphocytes upon entry into the study were
recorded. The prevalence of drinkers by exposure category was calculated and simple analyzes were
made to investigate for possible correlations between exposure and age, gender, and CD4 count at
enrollment. To assess the impact of alcohol on the incidence of events of interest, survival analysis for
data with competitive risks was performed, for the events of death, treatment eligibility (CD4 <350 and
/ or WHO stage> 2), and treatment initiation (ARV).
The prevalence of alcohol use for every level of the exposure was found to be 26.67% for hazardous
drinkers, 15.95% for hyper drinkers and 41.57% for binary drinkers. Male sex is associated with a
higher likelihood of using alcohol for all exposure levels (ORhazardous = 4.34 p-valuehazardous: <0.001,
ORhyper = 4.14 p-valuehyper: <0.001, ORbinary = 3.33 p-valuebinary: <0.001). Older age is associated with
hazardous drinking (age: medianhazardous = 32.79 years versus 30.68 years for those not exposed, p-value
= 0.014), possibly associated with the level of exposure of hyper drinking (age: medianhyper = 32.72
years versus 30.78 for those non-exposed, p-value = 0.053), but not associated with the level of binary
drinking. No association of the number of CD4 at enrollment with any of the exposure levels was
found. There were observed 521 ART initiation events with a statistically significant association with
binary drinking (sdHR = 0.775 / p-value = 0.011), a possible association with hazardous drinking (sdHR
= 0.819 / p-value = 0.079) and a non-significant association with hyper drinking (sdHR = 0.819 / pvalue
= 0.154). There were observed 48 deaths with no statistically significant association of the subdistribution
hazard for the event with any of the exposure levels (hazardous drinking sdHR = 1.647 / pvalue
= 0,206, hyper drinking sdHR = 1.191 / p-value = 0.723, binary drinking sdHR = 1.111 / p-value =
0.669). Finally, there were observed 382 ART eligibility events without a statistically significant
association with the sub-distribution hazard for the event with any of the exposure levels (hazardous
drinking sdHR = 1.041 / p-value = 0.681, hyper drinking sdHR = 1.007 / p-value = 0.958, binary
drinking sdHR = 1.046 / p-value = 0.599).
The direction of both the effects of alcohol on the events of interest and the associations between the
exposure and characteristics as age, CD4 count at enrollment, and gender are in concordance with what
88
is already known from literature. Out of the events examined in this study, the effect of alcohol
consumption on the incidence of ART initiation is significant (depending on the level of exposure).
Therefore, investigation for alcohol-related risks about a patient, using instruments like the AUDIT
questionnaire, may be of value for such a HIV-care-program. On the other hand, the low number of
events, particularly with death as the event of interest, reduces the power of the study for this event,
and makes it more difficult for statistically significant correlations to be detected.
Keywords:
Prevalence, Alcohol use, HIV, Survival analysis, Fine & Gray
