Many challenges face us in our attempts to transform HIV genotypic resistance testing into a useful clinical tool that can directly improve patient care on a routine basis. As we know them now, genotypic assays generate data on mutations in the HIV genome, which are reported as a list of changes in the amino acid sequence of the reverse transcriptase (RT) and protease genes. Translating this list of numbers and letters (representing the codons or positions of mutations and specific amino acid substitutions) into a report that clinicians can understand and apply readily in the clinic is a key objective of resistance testing research, but still represents a major challenge.
Ideally, labs would be able to provide clinicians with reports that accurately explain the clinical corollaries of an individual's viral genotype in terms of 1) specific drug(s) to which the patient's virus is resistant, 2) specific drug(s) to which that patient's virus is likely to become resistant, 3) specific drug(s) to which the patient's virus is susceptible, and 4) recommended treatment strategies. But this is a difficult and complex task, as we do not currently have clear-cut data linking all possible mutational patterns with their effects upon the clinical potency of specific antiviral agents. In other words, a simple list in which the impact of a specific mutational pattern (often including many mutations) on the utility of a given drug or drugs does not exist. Therefore, our current understanding of the clinical significance of a given mutational pattern is based on an expert's best educated guess.
Laboratory studies correlating genotypic mutations or combinations of mutations with changes in phenotypic resistance to a specific drug or drugs are one source of information, as are clinical trials associating the impact of a mutational pattern had on the effectiveness of a specific drug or combination of drugs. Unfortunately, information from such sources is patchy, incomplete and of varying quality. Therefore, when faced with a specific pattern of mutations, the expert will attempt to put together all the data available linking the pattern, or key components of it, to its effects upon the activity of available drugs. The requirement of extensive knowledge of the literature and the ability to consider varying techniques and data quality place the interpretation of genotypic resistance assay results out of the reach of many clinicians, who spend the majority of their time treating patients.
The great complexity involved in interpreting genotypic results often compel clinicians to seek the advice of experts if they choose to incorporate these assays into their practice. Although simple lists relating individual mutations to individual drugs are available today, and may already be "better than nothing", they greatly oversimplify this very complex issue and leave out much of our understanding of interactions and additive effects of mutations.
One possible way to enable clinicians to understand the results of a genotypic assay without requiring the consultation of an expert is to build a computer program that interprets the results for them. Such a system would rely on an extensive data base combined with an array of rules and algorithms designed by experts in the field. A specific mutational pattern generated by genotpyic analysis would be fed into the program, which would then identify the proper set of rules that apply to that specific pattern. These rules would then translate mutational data into recommendations regarding the degree of resistance to, and the (probable) clinical utility of, specific drugs.
The degree to which such a system relates the severity or probability of resistance to specific drugs, or actually recommends which drugs to use, may vary between systems. The Figure below delineates some of the tradeoffs inherent in such systems. Going from a report showing raw data to one giving a direct clinical recommendation makes genotypic resistance test results much easier to understand, but at the risk of losing accuracy and flexibility in using these data in different patients and different clinical situations. Systems which list the probability of resistance to the individual drugs without attempting to give specific recommendations may represent an appropriate format for the initial application of these systems.
