Background:  The efficacy of ART in HIV-1 infection is conditioned by the accumulation of multiple resistance-associated mutations (RAM). The specific the pattern of these RAM is determined by previous drug exposure and failures, and it their analysis is mandatory for choosing the appropriate salvage therapy. This analysis investigates the utility of a multivariate clustering statistical technique in the characterisation of homogeneous subgroups of multi-experienced patients with virological failure. This characterisation could identify homogeneous subgroups susceptible of further resistance analysis.

Methods:  The pattern of RAM of 195 genotypes from heavily multi-treated patients with virological failure was studied by means of a dendogram plot and k-means cluster analysis. The pharmacological history was evaluated for each of the mutation-clustered subgroups.

Results:  Analysis revealed 7 distinct clusters from a set of 195 genotypes belonging to patients with a median (IQR) CD4 cell count of 282 (252) cells/mm³, HIV-1 viral load of 4.2 (1.4) log₁₀ copies/mL, and 5 (2) lines of HAART regimens. The technique properly allocated together clusters of particular protease mutations, and the specific reverse transcriptase mutations of the TAM1 and TAM2 patterns. The different clusters were respectively defined by the presence of the mutations D67N and K70R; the protease mutations; the TAM 1 pattern together with the protease mutations and finally the TAM 2 pattern. According to the clustering the most frequent antiretroviral drugs received for patients allocated in each group were assessed. The cluster group presenting the M41L, M184V, T215Y, I54V, L63P, A71V, and V82A mutations allocated the patients that had received lopinavir or saquinavir with increased frequency. Another cluster identified the presence of D67N, K70R, L63P, V82A mutations, associated with the patients that had received AZT or D4T. A dendogram mutational pattern Figure is included.

Conclusions:  The dendrogram plot and k-means cluster analysis effectively identifies homogeneous subgroups of multi-treated HIV-1 patients with treatment failure according to the RAM harbored. This statistical technique is widely used to find homogeneous patterns in multi-dimensional heterogeneous datasets. Identifying groups with clustered mutations that emerge together may be useful to undertake further treatment or resistance analysis in this scenario.