Deferring ART by four weeks reduces mortality in patients diagnosed with cryptococcal meningitis
Nathan Geffen, CSSR
David Boulware from University of Minnesota, reported improved 26-week survival rates in patients diagnosed with cryptococcal meningitis (CM), by deferring ART for five weeks compared to initiating ART within a week. 
These results, from the Cryptococcal Optimal ART Timing (COAT) trial run counter to what has been found for trials that have examined immediate versus deferred strategies for other opportunistic infections.
Based on cohort studies, CM is estimated to be responsible for up to a quarter of AIDS deaths in sub-Saharan Africa. The disease has a high mortality rate, so optimisation of treatment has substantial benefit. CM is the highest cause of meningitis-related deaths in adults in Africa.
The study was designed to evaluate the risks of early or deferred ART including IRIS, tolerability of ART during CM illness and unknown drug-drug interactions between antiretrovirals and amphotericin B and fluconazole compared to benefits of reduced time with low CD4 count and high viral load.
As background, the ACTG A5164 trial published in 2009 compared disease progression and death in patients who initiated ART within 14 days of opportunistic infection diagnosis versus those who deferred until the end of acute treatment. Its primary endpoint had a non-significant trend favouring earlier ART, but its secondary endpoints provided compelling evidence for earlier ART: early treatment had less AIDS progression and death (OR = 0.51; 95%CI: 0.27–0.94). Interestingly, the trial also showed significantly less progression to AIDS or death for patients with CD4 counts <50 cells/mm3 who had fungal diseases (cryptococcal infections and histoplasmosis) and a trend to less disease progression and death when cryptococcal was considered alone. 
However, a Zimbabwe study by Azure Makadzange and colleagues found that patients with CM had greater mortality at three years when treated immediately with ART (54% vs 88%; p < 0.006), although CM treatment was fluconazole monotherapy. 
In a third small trial of only 27 patients with CM in Botswana by Gregory Bisson and colleagues, those initiated on ART within seven days had no difference in outcomes. Two of 13 (15%) and five of 14 (36%) patients died in the intervention and control arms, respectively (p = 0.39), but seven of 13 (54%) people in the intervention arm versus none in the control arm experienced CM-IRIS (p = 0.002). Amphotericin B was used with fluconazole in this trial. 
The COAT trial planned to randomise 500 ART-naive patients with CM who had been on anti-fungal treatment for 7 to 11 days to start ART within 48 hours or defer ART for more than four weeks. Randomisation was stratified by site and altered mental status. The primary endpoint was 26-week survival. Secondary endpoints included survival at 46 weeks. The trial was stopped early by the DSMB the trial before these numbers were reached. It was started in November 2010 and halted in April 2012.
The CM treatment regimen was amphotericin B and fluconazole 800 mg for the first two weeks, with fluconazole at this dose continuing until cerebrospinal fluid (CSF) was clear, with subsequent dose reductions of fluconazole.
A total of 389 patients were screened, of whom 237 had confirmed CM. A further 60 were not randomised, mostly because of death (n=33). In all, 177 patients were randomised, 115 in Kampala, 35 in Mbarara and 27 in Cape Town, of whom 88 were assigned to early ART and 89 to the deferred arm. The median time to starting ART on the immediate arm was eight days versus 35 on the deferred arm.
Baseline characteristics included median age 35, just over 50% male, 65% with no prior AIDS or TB and 23% had prior or current TB. About 14% had altered mental status. Median CD4 count was 19 (IQR: 9 to 69) in the immediate arm versus 28 on the deferred arm (IQR: 11-76). CSF opening pressure was similarly elevated in both arms. Number of colony forming units were a median of 5.2 and 4.8 log/mL in the immediate and deferred arms respectively. Slightly more patients in the immediate arm had white blood cell counts in CSF below 5 cells/mm3 (45% vs 32%).
The primary outcome of survival at six months was 55% in the immediate versus 70% in the deferred arm (p=0.03), with 68 deaths (40 versus 28). This was 20% lower than is usually seen in CM studies. A Kaplan-Meier graph showed that the excess mortality occurred in the treatment induction period of the immediate arm. After this period, the mortality graphs of the two arms were nearly identical.
Two groups drove mortality. Patients with altered mental status (less than 15 on the Glasgow Coma Scale – see Wikipedia for an excellent explanation of this scale) had a hazard ratio of 3.0 in the early ART group. Patients with low white blood cell counts in their CSF (<5 cells/mm3) were at even greater risk in the immediate arm, being 3.3 times more likely to die (p=0.01). By contrast the 60% of patients with more than 5 cells/mm3 had a hazard ratio of death of 0.8 on the immediate arm (i.e. non-significantly less likely to die than in the deferred arm, p=0.73).
Several other pre-defined sub-group analyses did not show significant differences between the two arms. There were more CM-related IRIS events in the early ART group but this was not statistically significant (16% vs 10%; OR: 1.7; 95%CI: p=0.35). A Kaplan-Meier graph showed time to IRIS was quicker in the immediate group.
Cause of death data was presented for 51 cases. CM was the cause of 21 and 10 deaths and septicemia was responsible for eight and five in the immediate and deferred arms respectively. There were two TB deaths on each arm as well as one IRIS death on the immediate arm and two on the deferred. Excess mortality was therefore primarily CM-related. Perhaps relevant is that a poster at CROI by the same team showed that early ART was associated with an increase in CSF cellular markers for macrophage activation among the Ugandan patients. At day 14, patients on the immediate ART (in both countries) were more likely to have an elevated CSF white blood cell count (58% vs 40%, p = 0.047). 
A Kaplan-Meier graph showed that more than 50% of patients had grade 4 or 5 serious adverse events, which Boulware explained was primarily due to amphotericin B toxicity. There were more adverse events in the immediate arm, but this was not statistically significant.
This trial showed that there is no benefit to starting ART during CM induction therapy. Deferred ART resulted in significantly fewer deaths and a trend for less IRIS events.
The conclusion proposed several recommendations:
- Treat CM first and optimally.
- Counsel patients for HIV and ART while they are in hospital.
- Verify that CSF culture is sterile at two weeks before starting ART or dropping fluconazole to 400mg.
- Aim to start ART at about four weeks, perhaps delaying ART initiation to five or six weeks in patients with low CSF white blood cell counts or altered mental status.
Boulware noted several limitations of the study including the early DSMB stop and the difficulty of discerning causes of death. For example, some of the deaths attributed to CM might have been IRIS-related. Also, the high level of care in the study cannot be easily generalised to operational settings: there was no loss to follow-up, 99% linkage to care, HIV counselling for patients while they were hospitalised and the full gamut of treatment as per the WHO guidelines.
During questions, Boulware explained that patients that are treated using fluconazole monotherapy clear cryptococcal from CSF about 30% slower than if amphotericin B is also used. He therefore suggested that patients in these settings should probably be initiated after five or six weeks.
WHO recommends that ART only be initiated when there is evidence of a sustained clinical response to anti-fungal therapy. It recommends against immediate treatment. These guidelines can now be changed from low- to high-quality evidence recommendations. 
This excellent study, with its relatively low mortality compared to other CM studies, once more shows the importance of using amphotericin B for the treatment of CM. Barriers to its wider adoption in many poor settings are price and the complexity of administering it.
- Boulware D et al. 2013. ART initiation within the first 2 weeks of cryptococcal meningitis is associated with higher mortality: a multisite randomized trial. 20th Conference on Retroviruses and Opportunistic Infections, 3-6 March 2013, Atlanta. Oral abstract 144.
- Zolopa AR et al. 2009. Early antiretroviral therapy reduces AIDS progression/death in Individuals with acute opportunistic infections: a multicenter randomized strategy trial. PLoS ONE 4(5): e5575. doi:10.1371/journal.pone.0005575.
- Makadzange AT et al. 2010. Early versus delayed initiation of antiretroviral therapy for concurrent HIV infection and cryptococcal meningitis in Sub-Saharan Africa Clin Infect Dis. (2010) 50(11): 1532-1538 doi:10.1086/652652
- Bisson GP et al. 2013. Early versus delayed antiretroviral therapy and cerebrospinal fluid fungal flearance in adults with HIV and cryptococcal meningitis. Clin Infect Dis. first published online January 29, 2013 doi:10.1093/cid/cit019.
- Scriven J et al. 2013. Early ART after cryptococcal meningitis increases cerebrospinal fluid macrophage activation and aberrant Th2 responses in a multisite randomized trial. 20th Conference on Retroviruses and Opportunistic Infections, 3-6 March 2013, Atlanta. Poster abstract 859.
- WHO. 2011. Diagnosis, prevention and management of cryptococcal disease in HIV-infected adults, adolescents and children.