Safety of long-term interruption of successful antiretroviral therapy: the ATHENA cohort study

We studied the dynamics of CD4 cell counts after the interruption of virologically successful highly active antiretroviral therapy (HAART) in 139 patients. Changes in CD4 cell counts during HAART interruption followed a biphasic pattern: an initial rapid decline during the first month followed by a slow decrease. During 48 weeks of follow-up mean CD4 cell counts remained just above the mean pre-HAART level. This limits the feasibility of structured treatment interruptions for patients with low nadir CD4 cell counts.

Detailed information on the dynamics of CD4 cell counts during the long-term interruption of successful highly active antiretroviral therapy (HAART) is needed to determine the safety of treatment interruptions in this setting, and to estimate the time a particular patient can remain without HAART. We used the ATHENA cohort ^[1] to study the virological and immunological dynamics and safety of the interruption of successful HAART. In patients using HAART who had undetectable viral loads for at least 24 weeks, we studied the first treatment interruptions lasting at least one week. A CD4 cell count at the start of the treatment interruption had to be available. During treatment interruption at least one CD4 cell count had to be available. Changes in CD4 cell counts and viral loads were analysed by a 'repeated measurements' procedure. Parameters possibly associated with the changes in CD4 cell counts were included as covariates into the models: pre-HAART CD4 cell counts, pre-HAART viral loads, sex, age, duration of the use of HAART, duration of undetectable viral loads before treatment interruption, gain in CD4 cell counts during HAART, and previous diagnosis of an AIDS-defining illness.

Of 3691 patients available for analysis, 139 fulfilled all criteria. At the start of first-line HAART the median viral load was 4.5 [interquartile range (IQR) 4.0-5.1] log₁₀ copies/ml, the mean CD4 cell count was 287 cells/mm³ (SD 202), female sex 14%, median age 39.7 years, 75% were infected via homosexual contacts, 56% were pretreated with nucleoside analogues, and 34% had a previous AIDS diagnosis. The patients had been treated for a median duration of 111 weeks before they discontinued HAART. The median duration of having undetectable viral loads was 71 weeks (IQR 40-111). The most frequent reasons for stopping HAART were drug-related toxicities (49%) and personal reasons (34%, especially the wish for a drug-holiday). At the moment of treatment interruption the mean CD4 cell count had increased from 287 cells/mm³ at the start of HAART to 493 cells/mm³ (SD 283). The median duration of the treatment interruption was 17 weeks; 18 patients discontinued therapy for 48 weeks or longer.

Four weeks into the treatment interruption the mean plasma HIV-1-RNA levels had rebounded to pre-treatment levels, and remained stable during follow-up (Fig. 1a).

Fig. 1

The gain in CD4 cell count during HAART was the only parameter that was significantly associated with the decline in CD4 cell count during treatment interruption (P < 0.0001). Fig. 1b presents the decline in CD4 cell count in relation to the pre-HAART CD4 cell count of the individual patients. The decline of the CD4 cell count during treatment interruption follows a biphasic pattern. At the moment of treatment interruption the mean gain in CD4 cell count during HAART was 215 cells/mm³. Four weeks into the treatment interruption the mean CD4 cell count was only 96 cells/mm³ above pre-HAART levels. From week 4 onwards there was a further slow decrease to 54 cells/mm³ above pre-HAART levels at week 48.

To investigate whether this biphasic pattern of decline in the CD4 cell count was similar for patients with either high or low pre-HAART CD4 cell counts, we repeated the analysis stratified into pre-HAART CD4 cell counts above or below 200 cells/mm³. We found identical CD4 cell patterns for both groups (P = 0.79). The duration of pretreatment was equal in both groups (P = 0.41).

Four patients experienced a new AIDS-defining event during treatment interruption: two oesophageal candidiasis (CD4 cell count 0 and 130 cells/mm³, respectively), one wasting (CD4 cell count 50 cells/mm³) and one extrapulmonary tuberculosis (CD4 cell count 820 cells/mm³). Eight patients died; however, six of these occurred after HAART was restarted: one euthanasia, one suicide, one traffic accident, one myocardial infarction, one cytomegalovirus encephalitis (CD4 cell count 44 cells/mm³), and one unknown cause of death (CD4 cell count 130 cells/mm³). Two patients died during treatment interruption: end-stage AIDS (CD4 cell count 10 cells/mm³) and one patient was unexpectedly found dead (autopsy was not performed, CD4 cell count 203 cells/mm³).

Seventy-three per cent of patients restarted HAART. Within 24 weeks, 83% had a viral load below 400 copies/ml. Fig. 1c presents the changes in CD4 cell counts in relation to the pre-treatment interruption CD4 cell counts of the individual patients. Twenty-four weeks after restarting HAART the CD4 cell counts had increased again to pre-treatment interruption levels.

This retrospective observational study has several limitations. Patients were not evaluated at standardized timepoints. Only a limited number of patients discontinued HAART for more than 6 months. The lower number of available measurements at later timepoints resulted in larger standard errors for the estimates of the mean CD4 cell count at these timepoints. However, the observed trend for the decrease in the CD4 cell count remains clear during the entire period of follow-up. We have no data on the reasons for restarting HAART.

The observed rapid initial decline in CD4 cell count is strongly correlated with the previous gain in CD4 cell count during HAART. The changes in CD4 cell count during treatment interruption followed a biphasic pattern: an initial rapid decline to values just above the pre-HAART levels during the first 4 weeks, followed by a further slow decrease. In patients interrupting virologically unsuccessful HAART, a biphasic pattern of the decline of the CD4 cell count was previously observed by Mussini et al. ^[2]. The biphasic decline of CD4 cell counts resembles the inverse biphasic increase of CD4 cell counts after the initiation of first-line HAART ^[3]. The initial rapid decline during treatment interruption might be caused by a redistribution of CD4 cells from the peripheral blood into lymphoid tissues, followed by a slower decline typical of the natural history of untreated HIV disease.

The interruption of virologically successful HAART appears to be safe in patients with high nadir CD4 cell counts, just as deferring initial treatment appears to be safe in these patients ^[4]. In most patients, CD4 cell counts did not fall below pre-HAART levels. For patients with low pre-HAART CD4 cell counts, this limits the feasibility of so-called structured treatment interruptions as a tool for reducing drug exposure or decreasing pill fatigue. After restarting HAART, most patients resuppressed their viral loads, and the CD4 cells that were lost during the treatment interruption were quickly regained.

Acknowledgements

The authors would like to thank the members of the clinical and epidemiological working group of the ATHENA cohort study. The following clinical sites participate in the ATHENA cohort study: Medical Centre, Alkmaar; Onze Lieve Vrouwe Gasthuis, Amsterdam; Prinsengracht Hospital, Amsterdam; Slotervaart Hospital, Amsterdam; Medical Centre, Jan van Goyen Kliniek, Amsterdam; Medical Centre, Vrije Universiteit, Amsterdam; Academic Medical Centre, Amsterdam; Hospital Rijnstate, Arnhem; Medical Centre Haaglanden, the Hague; Hospital Leyenburg, the Hague; Catharina Hospital, Eindhoven; Medisch Spectrum Twente, Enschede; University Hospital, Groningen; Kennemer Gasthuis, Haarlem; Medical Centre, Leeuwarden; University Medical Centre, Leiden; University Hospital, Maastricht; University Medical Centre St Radboud, Nijmegen; Erasmus University Medical Centre, Rotterdam; St Elisabeth Hospital, Tilburg; University Medical Centre, Utrecht; Hospital Walcheren, Vlissingen.