Asymptomatic neurocognitive disorders in patients infected by HIV: fact or fiction?

From a statistical point of view, in addition to the results of the paper by Gisslèn et al. [13], it is interesting to note that in the case of independent tests exhibiting a Gaussian distribution the probability of having all scores of the five tests more than one standard deviation from their mean is only about 0.42 (as demonstrated both by the probability of only one test being abnormal, which is 0.84134, raised to five and by simulations that we have performed); so, the probability of having something not 'absolutely normal', according to the AAN proposed threshold (one standard deviation lower than the mean), is less favorable than tossing a coin. Of course, the very high probability value of making a diagnosis of ANI together with the above result immediately emphasizes the problem of the false positive proportion and, consequently, the need for a critical consideration of the validity of diagnostic criteria based on the 'z-score'. Indeed, it is common to obtain some z-score values associated with a low probability value, meaning that it is practically impossible from a Gaussian distribution [14]. Moreover, the Gaussian model could be limited since it may include impossible negative values; this happens, for instance, for values lower than two standard deviations less than the mean of the raw values (that is, the neuropsychological test results without adjusting for the normative values). According to these considerations, it would be more advisable not to rely upon some asymptotic results from the statistical theory (central limit theorem) leading to the Gaussian model, but to refer to skewed distributions and to calculate non-parametric thresholds (0.95 quantile, for example, or its lower 95% confidence limit). We actually think that, when an immediate medical treatment is not required as in the case of an ANI, knowing that a patient is under some relevant quantiles (0.5, 0.10 or 0.25) of the 'reference population' can be more informative for monitoring his/her follow-up than a result based on a parametric model (one standard deviation less than the mean), which may be too restrictive and not able to fit well with a set of real observations.

From a clinical point of view, it is important to highlight that, being asymptomatic, ANI is not an overt disease, so it would be clinically relevant only if it had been correlated with signs of definite pathologies or if it had been shown to be predictive of more severe neurocognitive impairments. Although a cross-sectional study demonstrated a correlation between pathological evidence of HIV encephalitis at autopsy and ANI diagnosed ante-mortem, this study is limited because it involved only the most compromised patients who died, so results are not transferable to the general population [11]. Moreover, it is a little disconcerting to find that, using the AAN criteria, even 16% to 19% HIV-negative young people about 35 years old who were used as controls had neurocognitive disorders in the study by Heaton et al. [3]. Lastly, in the five years after ANI was defined [9], no single study was performed to assess its predictive value for MND or HAD. Therefore, it remains plausible that patients with ANI have a greater risk of progression to MND/HAD than those without, but this is currently uncertain and may be difficult to demonstrate for several reasons:

Several studies have investigated biomarkers and found that signs of persistent immune-activation in the CSF (such as pleocytosis, increased albumin concentration, neopterin, neurofilaments, IL-6, cholesterol, myoinositol) [7, 27‐30] or chronic peripheral activation (indicated by CD14, CD16 or lipopolysaccharide) [31‐33] correlated with neurological impairment. As for neuroimaging, recently developed tools are promising. For instance, magnetic resonance spectroscopy is able to show a decreased level of N-acetyl aspartate, a sign of mature neurons and their axonal processes, in the basal ganglia of patients infected by HIV; moreover, decreased levels of N-acetyl aspartate were accompanied by increased choline and myoinositol as signs of cell turnover and inflammation [34]. Interestingly, these markers may be restored by HAART, and this may be correlated with improvement of neurocognition [35]. It has to be seen whether incorporation of these biomarkers or neuroimaging results into diagnostic algorithms may improve the clinical predictive value and relevance of subtle neurocognitive impairments such as ANI.

In the meantime, what should clinicians do when a diagnosis of ANI is made? Clearly, with the current diagnostic definition, ANI should not be assumed to be synonymous with an overt pathology or a major risk for HAD. Some experts recommend, as a cautious approach, to retest patients with ANI with the aim of identifying early on, any sign of progression of the neurocognitive disorder [12]. However, this practice should not excessively preoccupy our patients, leading to anxiety and depression for a condition that may not be significant. This is important in light of the fact that some patients, up to 53%, suffer from ANI (as currently defined) [5].