Treatment exhaustion of highly active antiretroviral therapy (HAART) among individuals infected with HIV in the United Kingdom: multicentre cohort study

Selection of patients

The criteria for inclusion of an individual in the UK CHIC study were that the patient was HIV positive, older than 16 years, and had attended one of the centres for care at any time after 1 January 1996. To date, data from existing clinical databases from six centres (Chelsea and Westminster NHS Trust, King's College Hospital, Mortimer Market Centre, St Mary's Hospital, Royal Free Hospital, and Brighton and Sussex University Hospitals, see bmj.com) have been merged. Data on demographics, AIDS events and deaths, antiretroviral treatment, and laboratory tests, were provided in a standardised format, including all information from the time of diagnosis. The analyses presented here are based on data collected over the period 2003-4.

Clinics provided data in a pseudo-anonymised form, using Soundex codes derived from surnames. We identified patients who had transferred between the centres in the study by matching on the basis of Soundex code and date of birth, and their records were linked.

Statistical methods

Patients were included in the cohort from 1 January 1996, their first attendance at one of the centres, or their 16th birthday, whichever occurred latest. We classified patients as under follow up in each year if the dates when they were first and last seen at any of the centres indicated that they were under follow up at that centre in that year. We present data for 1996-2002.

We defined individuals as being exposed to an antiretroviral drug if their treatment history included any use of that drug. We reported use of the following antiretroviral drugs: nucleoside reverse transcriptase inhibitors (zidovudine, lamivudine, stavudine, didanosine, zalcitabine, lodenosine, and abacavir), protease inhibitors (ritonavir, saquinavir (soft gel or hard gel formulation), indinavir, nelfinavir, amprenavir, or lopinavir), non-nucleoside reverse transcriptase inhibitors (efavirenz, nevirapine, delavirdine, and loviride), and a fusion inhibitor (enfuvirtide (T-20)). Use of two nucleotide reverse transcriptase inhibitors, adefovir and tenofovir, was limited over the study period; these drugs are included as nucleoside reverse transcriptase inhibitors in this analysis.

We used each patient's last available CD4 count and plasma HIV RNA burden in a year in the analysis of trends over time. We included patients only if they were under follow up and had a CD4 count or HIV RNA burden measured in that year. We did not consider patients who were lost to follow up as treatment failures.

We classified patients as having experienced virological failure to a regimen containing a protease inhibitor if two consecutive HIV RNA measurements had been above 500 copies/ml after at least six months' exposure to protease inhibitors. If patients had discontinued this class of drugs before the second HIV RNA measurement above 500 copies/ml we did not classify them as experiencing virological failure. We defined virological failure on a regimen that contained non-nucleoside reverse transcriptase inhibitors in a similar manner: we then classified patients as experiencing three class failure if they had experienced failure with regimens containing both protease inhibitors and non-nucleoside reverse transcriptase inhibitors (we assumed that failure had occurred with nucleoside reverse transcriptase inhibitors, as protease inhibitors or non-nucleoside reverse transcriptase inhibitors were rarely used without nucleoside reverse transcriptase inhibitors); we took the date of experiencing three class failure as the later of the dates of failure on a protease inhibitor and non-nucleoside reverse transcriptase inhibitors containing regimen. We considered CD4 count and HIV RNA trends separately in three groups: patients who had ever received antiretroviral therapy, patients who had been exposed to three or more classes of drugs, and patients who had experienced three class failure. We used the statistical package SAS, version 8, for all analyses.

Potential limitations

We have considered the impact of antiretroviral therapy on HIV infected individuals in the United Kingdom at a population level. Such trends should always be interpreted cautiously. In particular, increased CD4 counts over time may result from an increase in the number of newly diagnosed individuals with high CD4 counts attending the centres in the collaboration. However, results from the CD4 surveillance scheme in the United Kingdom12 and from individual cohorts in the collaboration13 14 indicate that this is not the case. A second concern is that as this is a dynamic cohort, patients may leave the cohort at any time as they move to centres that are not part of the collaboration. CD4 counts and HIV RNA values cannot be included in our analyses once patients leave the cohort, which may be a source of bias, particularly if those who were starting to experience therapeutic failure were more likely to drop out of the cohort. Although it is certainly true that some patients will have dropped out with therapeutic failure, many will have dropped out for unrelated reasons. The median last available CD4 count and HIV RNA measurement for those who dropped out of the cohort was above 200 cells/mm³ in 41% and < 2.7 log₁₀ copies/ml in 32%, respectively, implying that many of these patients did not have therapeutic failure at their last visit. These proportions have remained relatively stable over time; we therefore do not believe that this will affect the trends seen.

One of the benefits of using data from a large multicentre study is that it is more representative of the UK epidemic than cohorts from single centres. However, although the dataset currently includes data on a third of all HIV infected individuals seen for care in the United Kingdom15 and patients in the cohort are broadly representative of HIV infected individuals in the UK,5 the cohort includes disproportionately more homosexual men and individuals of white ethnicity than are seen in the United Kingdom as a whole. Furthermore, as all centres actively participate in research studies (including clinical trials of new drugs), it is possible that exposure to novel treatments in these clinics may occur sooner than in other centres. Finally, discrepancies may still occur in histories of antiretroviral treatment if patients attend other centres either before or during the periods of follow up and treatment information from previous centres is incomplete. In this situation, our estimates of exposure to antiretroviral therapy may be underestimated.

Meaning of the study

We have identified two groups thought to be at high risk of treatment exhaustion: patients exposed to three or more drug classes and those who had experienced virological failure on regimens including both protease inhibitors and non-nucleoside reverse transcriptase inhibitors. Not all treatment switches are made as a result of virological failure; individuals may also change treatments for convenience or to reduce toxicity. Thus, patients who have been exposed to three classes of drugs may not have experienced virological failure while receiving these drugs and may not necessarily show signs of treatment exhaustion. Patients who are known to have experienced virological failure while receiving these drugs, however, would be expected to have developed some resistance to their failing regimens, possibly leading to cross resistance to other drugs in the same class16 and placing these individuals at high risk of therapeutic failure. Although this group had higher HIV RNA measurements and lower CD4 counts than other treated individuals, virological failure is an imperfect surrogate for the presence of resistance mutations, and some of these patients may not have developed resistance to both protease inhibitors and non-nucleoside reverse transcriptase inhibitors. Close links with the UK HIV Drug Resistance Database17 will allow us to deal with this question directly once the use of resistance testing has become routine in this group. Further follow up of clinical events in these patients will allow us to assess whether our definition of three class failure is a good indicator of subsequent poorer clinical outcome.

Our findings show that new drugs with low toxicity, which are not associated with cross resistance to existing drugs, will be needed for such patients.

The participating centres and collaborators are on bmj.com