How soon after HIV seroconversion is antiretroviral therapy initiated?

Objective: To estimate the time from HIV seroconversion to initiating antiretroviral therapy (ART) and whether this has changed over calendar time, and to estimate the CD4 cell count at which ART is initiated.

Design: Data from a cohort of HIV seroconverters in the UK were analysed with the initiation of ART as the outcome variable.

Method: The association of the time from seroconversion to initiating ART with age, sex, exposure category, calendar time, CD4 cell count and acute infection at diagnosis was examined using Kaplan-Meier plots and Cox proportional hazards models. The CD4 level at which therapy was initiated was examined.

Cited Here...: Of 1308 seroconverters, 710 (54%) had started ART by 30 September 1998. Median interval from seroconversion to initiating ART was estimated to be 59.5 months [95% confidence interval (CI), 54.7-63.6]. Compared with 1989-1994 (after adjusting for CD4 cell count), time to initiation of ART was significantly shorter in 1997-1998 with relative rates of initiating ART of 1.02 (95% CI, 0.67-1.55), 1.07 (95% CI, 0.88-1.31) and 3.16 (95% CI, 2.56-3.90) pre-1989, 1995-1996, and 1997-1998, respectively. Median CD4 cell count at initiation of treatment was 73 (95% CI, 30-109), 136 (95% CI, 118-161), 110 (95% CI, 84-140) and 221 (95% CI, 200-250) ¥10⁶ cells/l for the periods pre-1989, 1989-1994, 1995-1996 and 1997-1998, respectively. Of persons seroconverting in 1997-1998, 19.5% initiated ART within 6 months of seroconversion compared with 8.4% and 2.0% in 1995-1996 and 1989-1994, respectively.

Conclusions: ART is being initiated closer to HIV seroconversion than in previous years. Whether the improvements in survival reported from recent observational studies result solely from increased efficacy of the regimens or also from the timing of their initiation is difficult to determine. No clinical trial has yet addressed the optimum timing of initiating ART in the era of potent therapies.

Introduction

The question of when treatment for HIV infection should be initiated and with which drugs has become increasingly complex and remains unresolved. The decision to prescribe antiretroviral treatment (ART) for an HIV-infected person is generally guided by a number of factors: whether HIV-related symptoms have been diagnosed, the degree of immunosuppression as determined by CD4 level, the amount of circulating virus measured by plasma HIV RNA, the likelihood of compliance to the prescribed regimen, the views of the patient and any existing national and international treatment guidelines [^1-3].

In vitro studies and clinical trials [^4-10], mirrored in the findings of observational studies [^11-14], have demonstrated the effectiveness of treating HIV-infected individuals with combinations of nucleoside analogue reverse transcriptase inhibitors and protease inhibitors. This has been estimated by a decrease in viral load, often to levels below the detection limits of current assays (‚undetectable‚). The choice of a particular antiretroviral combination may be driven by published data; the toxicity, tolerance and ease of dosing schedule; and by personal clinical experience. Not only is the selected drug combination important but also the time in the course of HIV infection at which therapy is initiated may also be crucial. A drug regimen may have the greatest chance of controlling viral replication when the rate of replication is relatively low [^15,16]. It has been hypothesized that long-term success may be more likely with early initiation of therapy before the extent of immune damage is sufficiently great that reconstitution may not be possible. These are valid, though unproved, arguments for initiating therapy close to seroconversion. However, in the absence of ART, the median time from HIV infection to death is of the order of 11-12 years [^17-20] and, therefore, treatment may need to be taken for many years. Concerns over long-term compliance, toxicity and the limited duration of the effectiveness of the various treatment regimens argue for delayed initiation of ART. To date, two clinical trials have addressed the question of timing of HIV therapy, both relating to zidovudine monotherapy [^21,22]. However, no evidence exists, as yet, regarding the optimum time to initiate the new highly effective combination therapies.

In this paper, we describe the initial treatment regimen for HIV-infected individuals in the United Kingdom whose time of seroconversion to HIV antibody is known and who are enrolled in the UK Register of HIV Seroconverters. The time from seroconversion to the initiation of ART was estimated and factors that might be associated with this time were examined. We explore whether there have been changes over calendar time and the extent to which these are related to CD4 level at the time of initiation of treatment.

Methods

In October 1994, the UK Register of HIV Seroconverters was established and data collected on all eligible individuals in United Kingdom clinics. Eligibility criteria were HIV-positive persons, aged 16 years or over, with documentation of a negative HIV antibody test within 3 years of the positive test, or with laboratory evidence of seroconversion as defined by a consensus arrived at by the virologists on the Steering Committee. The Register was set up with the main objective of monitoring the AIDS incubation period distribution and of survival following HIV seroconversion. Full details and the first results from the Register have been published elsewhere [^20,23].

Baseline data are collected at registration and follow-up information is collected annually. Variables recorded include all CD4 cell counts and viral load measurements, all changes in antiretroviral drugs since last follow-up, whether AIDS had been diagnosed, new AIDS-defining events, and survival status. Data collected both prospectively and retrospectively are through review of case notes and computer records. Although follow-up is annual, cumulative records of changes in ART throughout the year are obtained.

Kaplan-Meier methods [²⁴] allowing for late entry [²⁵] (no-one is considered ‚at risk‚ prior to 1 January 1987, the earliest date that ART therapy was likely to have been prescribed in the United Kingdom) were used to estimate the time from seroconversion to the initiation of ART. Seroconversion was defined as the date of laboratory evidence of acute infection, where known, or otherwise estimated as the midpoint between the first positive and last negative antibody test dates. For this analysis, the initiation of ART was the outcome of interest and persons not known to have started ART were censored on the date they were last known to be ART naive. Persons who had initiated ART on the day of their HIV-positive test were nominally given a survival time of 1 day. The association of the following variables with time from seroconversion to the initiation of therapy was examined using the log rank test and Cox proportional hazard models [^26,27]: exposure category [sex between men, sex between men and women, injecting drug use (IDU), others], age at estimated seroconversion, sex, calendar period at risk as a time-dependent covariate (pre-1989, 1989-1994, 1995-1996, 1997-1998), HIV test interval (within 1 calendar month, 1-36 months) and CD4 cell count as a time-dependent covariate.

The HIV test interval is the time, in months, between the negative and positive antibody test dates. Persons testing positive within 1 calendar month of the negative test are likely to have been diagnosed during the acute infection phase and may, therefore, have different prognosis and clinical management from seroconverters with a longer test interval [^28-30].

The calendar period at risk is the time during which infected individuals who have not yet initiated therapy were observed and therapy may or may not be initiated during that time. For example, a person who seroconverted in 1989 and was last seen in 1993 without having initiated therapy is at risk from 1989 until 1993 at which time they are censored. A person who seroconverted in 1989 and initiated ART in June 1993 is at risk during each of the years 1989 through to June 1993, at which time the initiation of therapy is treated as an event. Persons who died without initiating ART are censored on the date of death. The four calendar periods were chosen to correspond with the uptake pattern of antiretroviral drugs in the United Kingdom: before antiretroviral therapy was available, monotherapy, dual combination therapy, and triple therapy including protease inhibitors.

In order to estimate the median CD4 cell count at which ART is initiated, we adapted the method proposed by Phillips et al. [³¹]. This used Kaplan-Meier estimations with the CD4 cell count in lieu of time and the initiation of ART as the outcome of interest. The ‚survival‚ interval was, in this manner, the minimum CD4 cell count prior to the initiation of ART (within a 6-month period) for those on ART, and the cell count when last seen (within a 6-month period) for persons with censored observations. The interval allowed of up to 6 months for the availability of a CD4 cell count is arbitrary but counts are unlikely to vary greatly within that period in the absence of ART. To estimate median cell count for each of the four periods at risk, we then considered only persons at risk during each period. For persons who remained at risk beyond that period we used their CD4 cell count at the end of it. For example, a person who seroconverted in 1990 and started ART in mid-1996 was at risk during the two periods 1989-1994 and 1995-1996. For estimates of median CD4 cell count in 1989-1994, we censored their observation at the end of December 1994 and used their CD4 cell count at that time as their ‚survival‚ interval. For the period 1995-1996, when they experienced the event, their cell count prior to the initiation of therapy was their ‚survival‚ interval.

Results

To the end of September 1998, data on 1451 documented seroconverters were reported to the UK Register. Details of exposure category, sex and age are given in Table 1. The median year of estimated seroconversion is 1991 with a range from 1982 to 1998. Median age at estimated seroconversion is 28 years (range 16-70 years) with IDU being younger at the time of seroconversion than persons exposed sexually. For 127 subjects, information on whether or not ART had ever been prescribed was not available; for seven persons who had been prescribed ART, the time of initiation was not recorded. A further eight individuals were not included in analyses because they died or were lost to follow-up before 1 January 1987. One person was excluded because the only ART prescribed pre-dated his antibody negative test to HIV. For the 1308 individuals included in the analyses, 46% (598) had not been prescribed ART, including 74 who had died. For the 710 individuals who had been prescribed ART, 418 (59%), 189 (27%) and 103 (15%) initiated therapy with regimens of one, two or three (or more) drugs, respectively. Of 103 persons who initiated with three or more drugs, for 74 this included a protease inhibitor with at least two nucleoside analogue reverse transcriptase inhibitors, and for 22 this included a non-nucleoside reverse transcriptase inhibitor with at least two nucleoside analogue reverse transcriptase inhibitors. Seven subjects initiated therapy with other combinations. As expected, the proportion of persons initiating ART with two and three (or more) drugs increased over calendar time while the proportion starting therapy with one drug decreased over time (Fig. 1). CD4 data were available for 1242 persons, of whom 687 had started ART.

Table 1

Fig. 1

Overall, the median interval from seroconversion to the initiation of ART was estimated to be 59.5 months [95% confidence interval (CI), 54.7-63.6]; that is, by 5 years from seroconversion half of those infected had started ART. However, this interval changed significantly over time. In Cox proportional hazard models we found that, at a given time from seroconversion, persons who seroconverted in 1997-1998 [when highly active antiretroviral therapy (HAART) was available] had a 437% higher rate of starting ART than persons who seroconverted in 1989-1994 (when monotherapy was likely to have been prescribed) (Table 2). Of those who seroconverted in the period 1997-1998, 19.5% had started on ART within 6 months of seroconversion. Given the margin of error in estimating the time of seroconversion using the midpoint between the negative and positive antibody test dates with a maximum period of 3 years between them, we repeated analyses restricting estimates to persons with a maximum 12-month HIV test interval. This gave an estimate of 24.2% of seroconverters in 1997-1998 starting therapy within 6 months of seroconversion. However, the time at risk rather than the time of seroconversion determines which, if any, ART is used; for example, a person seroconverting in 1989 may be prescribed drug regimens that only became available in 1997. For this reason time was modelled as calendar time at risk rather than calendar time of seroconversion.

Table 2

At the same CD4 cell count and the same time from seroconversion, the rate of starting ART was significantly higher in 1997-1998 than in previous calendar periods. For a given time from seroconversion (after adjusting for CD4 cell count) and comparing with the period 1989-1994, the relative rate of starting ART was 1.02 (95% CI, 0.67-1.55), 1.07 (95% CI, 0.88-1.31) and 3.16 (95% CI, 2.56-3.90) for persons observed pre-1989, 1995-1996, and 1997-1998, respectively.

Age was also found to be significantly associated with initiation of therapy, with the oldest individuals initiating therapy at an earlier time than younger individuals, with an estimated 2.0% (95% CI, 1.1-3.0) increase in the risk of starting ART for each 1 year increase in age. After adjusting for CD4 cell count, the effect of age became less, with an estimated 1.0% (95% CI, 0.04-2.1) increase in risk for a 1 year increase in age.

Injecting drug users were found to be more likely to start therapy than homosexual and bisexual men at the same time from seroconversion [relative rate (RR) 1.38; 95% CI, 1.07-1.78). However, no such difference was evident after adjusting for CD4 cell count (RR 1.16; 95% CI, 0.89-1.52). In univariate analysis, persons with an HIV test interval less than 1 month (likely to have been identified during acute infection) were significantly more likely to initiate therapy than persons with a longer interval (RR 1.52; 95% CI, 1.12-2.07). After controlling for the effect of age, however, the difference in the rate of initiating ART between the two groups was somewhat smaller and was not conventionally significant (RR 1.33; 95% CI, 0.97-1.82). This is likely to be because persons identified during acute infection (mean age 33.7 years; SD, 9.6) were slightly older than persons identified at other times (mean age 29.5 years; SD, 8.2). None of the other factors examined were found to be significantly associated with the time of initiating therapy. A summary of the effect of variables examined, controlling for all other variables, is given in Table 3.

Table 3

There is evidence that ART was initiated at a higher CD4 cell count in the most recent time period (Table 4). The estimated median CD4 cell count at the initiation of ART was 221¥10⁶cells/l in 1997-1998. Of the 524 persons who had not initiated therapy by the beginning of 1997, 353 did not initiate ART during 1997-1998. Of these, 33, 197, 94 and 29 had seroconverted in the calendar periods pre-1989, 1989-1994, 1995-1996 and 1997-1998, respectively. Their median CD4 lymphocyte counts when last known to be ART naive were 450, 442, 549 and 547¥10⁶cells/l, respectively. For 84 individuals, their CD4 cell count when last seen was below 350¥10⁶cells/l, 23 of whom had a count below 200¥10⁶cells/l.

Table 4

Discussion

There is strong evidence that ART is now being initiated earlier, in terms of time from seroconversion and CD4 level, than in previous years. This has been particularly notable since 1996 with the advent of HAART.

Seroconverters in the most recent time periods are much more likely to start treatment with regimens that contain at least three drugs. Until the second half of 1992, all those opting to initiate treatment in this study started on one drug, for the majority in our study (240 of 244) this was zidovudine. Thereafter, this proportion fell steadily over time, so that from 1997 less than 10% of seroconverters in our study started therapy on one drug. The proportion of seroconverters prescribed a two-drug combination as initial therapy reached a peak in the second half of 1996, from which time drug combinations of three or more drugs became increasingly common.

We estimate that approximately one in five persons who seroconverted in 1997-1998 is likely to initiate therapy within 6 months of seroconverting. At a given CD4 cell count, HIV-infected persons in 1997-1998 have up to three times the rate of starting therapy as persons in previous years who were at the same time from seroconversion. Current initiation of therapy at a median count of 221¥10⁶cells/l is evidence that initiation is at an earlier time in the course of HIV disease than in previous years; this appears to be in line with the recent revision of the British guidelines [²]. Not all persons for whom ART is recommended, however, have initiated therapy in our study, with 16% (84 of 524) of persons in 1997-1998 with a CD4 cell count below 350¥10⁶cells/l when last seen remaining ART naive.

As the UK Register was set up in 1994, data on subjects included in the analysis have largely been gathered retrospectively. It is unlikely that the changes observed in the calendar period 1997-1998 resulted from more careful data collection as data prior to 1994 and post-1994 were collected in a similar fashion. In addition, there was no evidence of a difference in the time to initiating therapy between the calendar periods pre-1994 and 1994-1996.

The finding that treatment is initiated earlier in older individuals was not unexpected. A number of studies have reported a worse prognosis with increasing age [^19,20,32-35], and the initiation of therapy is likely, therefore, to be related to this. The finding that this age effect is largely explained by CD4 cell count supports this and indicates a more rapid decline in CD4 cell counts in the older age groups. The results suggest, however, that at the same CD4 cell count ART is more likely to be initiated in an older individual.

In our study, IDU appeared more likely to initiate therapy than persons in other exposure groups. However, after adjusting for serial CD4 cell counts, no difference in the rate of starting ART remained. It is worth noting that within the UK Register data on treatment for 169 IDU were available for analysis and of these 74% (125) were from a single well-defined cohort in Edinburgh [³⁶]. Overall uptake of therapy among these IDU, who had seroconverted at a relatively earlier time of the HIV epidemic in the United Kingdom, appeared to be much higher (69% compared with 27% for other IDU in the UK Register). This closely knit group of IDU with low mobility and good follow-up may, therefore, be atypical of IDU in general.

Similar to a number of other studies, the Register has reported an improvement in survival times for the most recent calendar periods [³⁷]. While it may be tempting to infer that improved prognosis is related to the earlier initiation of ART reported here, extreme caution must be exercised. It is not known if survival improvements are related only to the drug regimens per se or are also a function of their time of initiation. Further, we referred throughout to the initiation of ART when in reality only prescription dates for these drugs are available, and compliance was, therefore, assumed. Any discrepancies between what is prescribed and what is taken must be borne in mind when attempts are made to examine the relationship between the initiation of therapy and subsequent survival, as well as any likely differences in compliance by age or exposure group.

While it is hoped that the improvements so far observed in survival and in delaying the diagnosis of AIDS will continue, these require long-term monitoring. An understanding of whether these relate only to the drugs themselves or are also a function of their timing will require data from clinical trials.

Acknowledgements

The authors are grateful to the work of many colleagues in clinics and laboratories throughout the United Kingdom who have provided information on subjects included in this analysis. The help of colleagues in the Clinical Trials Unit (Joanne Gillett, Patrick Kelleher, Debbie Johnson, Elizabeth Brodnicki, Steve Fleming, Robert Manning) is also gratefully acknowledged, in particular invaluable help and advice from Sarah Walker.