How common is neurocognitive impairment in people with HIV? It depends who you compare them with

This article is more than 13 years old. Click here for more recent articles on this topic

How much more common is brain impairment and dementia in people with HIV than in the general population? A study presented at the 13th European AIDS Conference in Belgrade today showed that the answer could vary from 'no more common' to 'four times as common' according to which group you studied, which sample of the HIV-negative population you compared them with, and whether you averaged their neurocognitive (NC) performance over all domains or picked out specific areas of impairment.

The study, of 560 people with HIV undergoing a randomised controlled trial of protease inhibitor monotherapy, notably failed to find any association between patients' CD4 nadir (their lowest-ever CD4 count) and NC impairment, unlike other studies. It also failed to find any association of NC with being on an ARV regimen without good penetration into the central nervous system (with a low so-called CPE score).

Researcher Alan Winston of Imperial College in London commented that this was probably because this was a study of patients on long-term, stable antiretroviral therapy and thus excluded ones who might be suffering from acute HIV-related cerebral symptoms.


multivariate analysis

An extension of multivariable analysis that is used to model two or more outcomes at the same time.


Relating to the heart and blood vessels.


In a case-control study, a process to make the cases and the controls comparable with respect to extraneous factors. For example, each case is matched individually with a control subject on variables such as age, sex and HIV status. 


Lowest of a series of measurements. For example, an individual’s CD4 nadir is their lowest ever measured CD4 count.


Taking a drug on its own, rather than in combination with other drugs.

In multivariate analysis, and after accounting for other probable bias, the only factor out of those measured that was associated with worse NC impairment was, in white patients, hepatitis C infection.

This analysis did find that NC impairment was over twice as common in black patients as in white compared to the general population.

Neurocognitive performance in the 'general population', however, depends on studies of groups that may not reflect the HIV-positive population.

Accordingly the researchers did a re-analysis of NC impairment, comparing it to a 'general population' group that was adjusted to take account of studies of impairment in HIV-negative people of black ethnicity – though the few studies that have been done are in African-Americans, not people born in Africa.

By adjusting for the ethnic mix in the patients studied, the researchers found that studies of NC impairment in people with HIV probably, if anything, tended to underestimate its relative prevalence in HIV positive white people (compared to HIV-negative white people) and to overestimate its prevalence in HIV-positive, compared to HIV negative, black people. In an analysis in which NC impairment was defined as a low score in two or more of the five performance areas measured, NC impairment, relative to the adjusted general population, was no worse in black patients than white – but it was, in all people with HIV, three times more common than in the general population according to this measure, and twice as common according to the global score.

The specifics

The study was a substudy of the PIVOT trial, a randomised controlled trial of protease inhibitor monotherapy versus standard combination therapy taking place at 40 centres in the UK.

This substudy included 560 patients, 382 (68%) of white ethnicity and the rest African. The researchers, at the start of the study while they were all still on combination therapy, gave all patients five neurocognitive performance tests: the grooved pegboard test, a measure of fine motor skills, two memory tests (sort-term and delayed recall) and two 'trailmaking' tests that involve a combination of fast motor and executive (decision-making) skills.

Three-quarters of the white patients were men and 40% of the Africans. The average CD4 nadir was moderately low, at 177 cells/mm3 in white patients and 146 in black, but current CD4 counts were good at 554 and 511 cells/mm3 respectively, reflecting the fact that patients regardless of ethnicity had been on combination therapy for an average of five years and with a viral load uner 50 copies/ml for four.

There were substantial differences in rates of smoking in the group, with 62% of white patients and 28% of black ever having smoked and of cardiovascular risk, with 48% of white patients at a more than 10% risk of a cardiovascular even in the next ten years compared with only 18% of black patients, which probably reflects smoking habits and gender in the two groups. The age of the two ethnic groups was the same and so was their psychological health as measured by anxiety scores.

In the general population, one would expect 16% of people given the five psychological tests to have a global performance score more than one standard deviation lower than the mean score. In the unadjusted figures, 17% of white patients fell into this category, thus having the same score as the general population, but far higher rates of black patients – 66%.

The general-population scores are in fact derived from studies of patients mainly in the US and Canada. The researchers adjusted these 'normative' comparison scores by using the quite limited data of NC performance in African-Americans to derive comparison scores more reflective of the ethnic balance of the UK patient group (though this assumes African-Americans and people born in Africa will tend to resemble each other in NC results). In these adjusted normative scores, 20% rather than 16% of HIV-negative people would, on average, have NC performance more than one standard deviation worse than mean.

They found that using these adjusted scores more than doubled the proportion of white patients with global NC impairment – from 17% to 38% - and reduced impairment in black patients, from 66% to 55%.

The researchers did the same analysis for impairment in at least two of the five tests measured. Unsurprisingly, if you defined NC impairment in this way, it was more common, seen in 38% of white patients and no less than 81% of black. If the adjusted normative scores were used, however, this increased to 64% in white patients and decreased to 69% in black, so the difference became statistically non-significant.

The researchers did a multivariate analysis of factors associated with NC impairment in the unadjusted normative figures. They found that white patients with hepatitis C were 33% more likely to have NC impairment. They also found that white patients on nevirapine were 34% more likely to have it too, but this is thought to be a 'channelling bias' whereby patients with pre-existing NC symptoms are more likely to be prescribed nevirapine than efavirenz. Oddly, black patients who smoked were 66% less likely to have NC than average, but this may reflect demographic differences not picked up in the study (e.g. black smokers might be more likely to be gay men, or born in the UK).

This study should not necessarily be taken to indicate more reliably than previous studies what the 'real' relative rates of NC impairment are in people with HIV in the UK compared to the general population, but it does illustrate that the results of all studies depend crucially on how we estimate neurocognitive impairment in general.

Researcher Alan Winston of Imperial College in London told the conference that the question of the relative rates of NC impairment in the HIV positive population would not be settled until more studies were done with matched control groups of HIV-negative peers, matched for more variables like sexuality, drug use, STI history and so on.


Winston A et al. Factors influencing neurocognitive function in a large cohort of HIV infected patients on effective antiretroviral therapy. Thirteenth European AIDS Conference, Belgrade. Abstract PS2/4. 2011.