Introduction

Improvements in the management of HIV including early initiation of antiretroviral therapy have reduced the incidence of opportunistic infections such as cryptococcal meningitis (CCM) [1]. In 2009, there were an estimated 975,000 cases of CCM; by 2014, this had fallen to approximately 225,000. Similarly, the number of estimated deaths fell from 625,000 in 2009 to 180,000 in 2014 [2, 3]. Unfortunately, as these figures illustrate, the risk of death once CCM develops changed little over those five years. CCM remains one of the leading causes of morbidity and mortality among HIV positive patients, coming second to tuberculosis [2–4], and is responsible for 15% of all AIDS-related deaths globally[3].

The highest burden of morbidity and mortality from CCM occurs in sub-Saharan Africa where there were an estimated 162,500 cases (73% of the global figure) with 135,900 deaths (75% of the global figure) in 2014 [3]. Many of the studies evaluating mortality published so far have included patients treated with best available, rather than guideline-recommended therapy, and few studies report mortality rates beyond the period of consolidation therapy [5–9]. This is relevant, because in many places the only available treatment has been fluconazole, although gold standard treatment has been based around intravenous amphotericin. As such, these studies may miss longer-term mortality effects that could be consequences of initial treatment. For example, different initial treatments may have different effects on the risk of disease relapse and disability, both of which may influence long-term survival [10].

Even so, the mortality on amphotericin-based therapy is high, reported to vary between 20% and 50% by 10 weeks from diagnosis. In addition to treatment, the risk of death has been associated with factors such as elevated intracranial pressure, slower fungal clearance from cerebrospinal fluid, and drug toxicities, amongst others [11–14]. Among the few studies that determined longer-term mortality, Butler K et al found a mortality of 66% at two years from treatment initiation among patients treated for CCM using recommended therapy in Mulago hospital in Uganda. This study explored the effect of only a few variables on mortality and the model included patients with cryptococcal antigenaemia but without CCM [15]. The majority of studies of treatment for CCM follow patients for between 3 and 6 months[14]. Up to this point, determinants of outcome include measures of disease severity at baseline, such as fungal burden (the amount of culturable yeast in cerebrospinal fluid), and treatment choice [14]. However, survival curves suggest that deaths may continue beyond this point for patients who do not receive initial treatment with Flucytosine [14]. We wanted to better understand the long term outcomes of cryptococcal meningitis in our country, and how this is influenced by treatment and non-treatment factors among CCM patients, including those previously associated with short term outcome. Identifying patients at increased risk of late death, while they are still engaged in care, may enable the development of strategies to improve their prognoses. We determined two-year mortality and associated factors (including the effect of dexamethasone adjunctive therapy) among patients previously enrolled in the CryptoDex trial (Clinical trial registration number ISRCTN59144167, PMID: 25391338) [16] in Uganda. The CryptoDex trial was stopped early due to safety concerns [16].

Methodology

Results

*All patients were treated with 2 weeks of amphotericin B 1mg/kg daily and 2 weeks of fluconazole, 800mg/day followed by 8 weeks of fluconazole 800mg daily, then fluconazole 200mg/day.

Baseline characteristics

Fig 1 is a flow diagram showing participants enrolled in the CryptoDex trial in Uganda and mortality through two years from treatment initiation of therapy. Out of 323 patients screened at the two sites in Uganda, 212 were found eligible for participation in the CryptoDex trial, one died before randomisation, and 211 were randomised. Data for all randomised participants was used in this analysis (Fig 1). Of these, 127 participants (60.2%) were male. The mean age (SD) was 36.4 years (8) and one hundred and five participants (49.8%) were randomised to receive dexamethasone adjunctive therapy. Sixteen patients (7.6%) were diagnosed with HIV on admission with CCM, while 195 (92.4%) were already known to be HIV positive with 116 (59.5%) of these already on ART. The median last CD4 (IQR) done prior to enrolment was 29 (7–71) cells/uL. Of all participants, 131 (65.5%) had CSF opening pressure ≥18cm of CSF, and 118 (55.9%) had more than 1000 yeast cells per ml of CSF. Twenty-five participants (11.8%) reported history of convulsions and fifty (23.7%) had haemoglobin counts below 10mg/dl. The baseline characteristics are summarised in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study flow diagram showing participants enrolled in the CryptoDex trial in Uganda and mortality through two years after initiation of therapy.

https://doi.org/10.1371/journal.pntd.0008823.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of study participants at randomisation in the CryptoDex trial.

https://doi.org/10.1371/journal.pntd.0008823.t001

Mortality by two years

All models assessed passed the test of proportional hazards assumption. At two years from initiation of treatment for CCM, 127 patients had died (proportionate mortality of 60%), giving a mortality rate of 67 deaths per 100 person-years after 190.7 person-years of follow up. The mortality rate was 57 and 79 deaths per 100 person-years among patients that received placebo and dexamethasone adjunctive therapy respectively. The median survival time was approximately 3 months. Basing on the Kaplan-Meier plot (Fig 2), mortality was highest during the first 6 months. Fig 3 shows mortality by intervention.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Kaplan-Meier plot showing survival of participants in the first 2 years of enrolment.

https://doi.org/10.1371/journal.pntd.0008823.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Kaplan-Meier plot showing survival of participants in the first 2 years of enrolment categorised per intervention.

https://doi.org/10.1371/journal.pntd.0008823.g003

Factors associated with mortality

Results from the univariate and multivariate analysis are summarised in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Mortality rates and effect of different factors on mortality at two years.

https://doi.org/10.1371/journal.pntd.0008823.t002

Discussion

By the end of two years of follow-up, 127 of the 211 study participants had died yielding 60% proportionate mortality and an overall mortality rate of 67 deaths per 100 person-years. This mortality rate contrasts markedly with that previously reported for Ugandan HIV patients with CD4 counts similar to those of our patients, i.e., below 50cells/ml, of 6.73 deaths per 100 person-years [17]. However, we believe our finding is plausible since it is consistent with the proportionate mortality of 66% reported by Butler et al from their study of patients with cryptococcal meningitis from Mulago Hospital, Uganda [15].

Apart from cryptococcal meningitis itself, other potential contributors to mortality during this period include drug-related toxicities, immune reconstitution syndrome (IRIS), morbidity associated with residual disability, and other infections, since patients are still severely immune-suppressed [18]. Like others, we found that most deaths did occur within the first 6 months following diagnosis [14, 19]. This suggests that efforts to improve early treatment of cryptococcal disease are likely to deliver the best long-term gains.

Unfortunately, there is a dearth of therapeutic options for cryptococcal meningitis. Cryptococcal meningitis continues to be managed using drugs that were discovered more than 60 years ago, whose use is also associated with multiple toxicities [20], while 30–40% of patients fail to attain CSF sterilization [14, 19, 21]. There has been little real impact on mortality in the last 20 years. Recent efforts examining drug repurposing and adjunctive therapy have been disappointing [18, 22]. We therefore agree with previous researchers that alternative and complementary approaches for the management of CCM still need to be investigated to curb mortality [11, 18]. Flucytosine remains expensive (>$120 per day) and unlicensed in most countries in sub-Saharan Africa including Uganda. The most practical intervention remains early HIV testing and intensive screening for CCM using the available Cryptococcal antigen tests which enable identification of asymptomatic antigenaemia a median of 22 days before meningitis sets in [23, 24]. It has been estimated that screening and pre-emptively treating people with positive antigenaemia in Uganda would cost only 15% of the cost of meningitis treatment, with more than 40% better long-term survival [15].

Mortality was not significantly different between patients treated with adjunctive dexamethasone therapy and those on placebo. This result is similar to findings from the CryptoDex study where no difference in mortality was observed at 10 weeks and 6 months [18]. This is perhaps not surprising given that the CryptoDex study was stopped after the enrolment of 451 patients following an interim safety analysis because of evidence of accumulating harm in patients receiving dexamethasone, in the form of increased rates of adverse events and slower fungal clearance. However, it is reassuring that deaths did not continue to accumulate as a consequence of the receipt of dexamethasone. It is however clear that dexamethasone neither improves short term or longer-term survival from CCM. The 2018 WHO guidelines on cryptococcal treatment recommend against use of corticosteroids during the induction treatment phase for CCM [25].

It had been hypothesised in the CryptoDex study that dexamethasone would improve outcomes by reducing both intracranial pressure and inflammatory complications, and decreasing the incidence of IRIS. Analysis done on CSF from a subset of patients in the CryptoDex study revealed that there was a downregulation of pro-inflammatory cytokines among dexamethasone recipients [26]. This could explain the absence of benefit.

Recently data has emerged from the ACTA trial of combination antifungal therapy for CCM in Africa, suggesting that despite good sterilising power, longer courses of amphotericin (2 versus 1 weeks) may be harmful to patients [27]. In this study, the best outcomes were seen in patients receiving amphotericin for one week, in combination with Flucytosine. Patients receiving two weeks of amphotericin had more adverse events, and worse survival (although the latter did not reach statistical significance). While the sterilising effect of one and two-week amphotericin regimes over the first two weeks seemed similar, it will be important to see whether there is any effect on long term survival. Differences in induction therapy appear to be able to influence long term outcomes, certainly up to 6 months after diagnosis [14].

We found that higher weight was associated with a reduced risk of mortality. Other studies among CCM patients have reported a similar finding, though this was after much shorter periods of follow up [19, 28]. An association between greater weight and improved survival is plausible since it is likely that the lighter patients had more advanced HIV disease and less physiological reserve. Scherzer R et al previously reported that lower muscle mass was associated with increased risk of mortality among HIV patients in general [29].

Presence of convulsions was also associated with increased risk of mortality. A similar finding has previously been reported in previous studies [30, 31]. Convulsions may be associated with more severe disease and/or other effects in the brain that could increase the risk of mortality. The possibility of other disease manifestations such as cryptococcomas, or comorbidities that can cause convulsions cannot be entirely ruled out in our patients because, unfortunately, advanced investigatory modalities like CT scans are not readily available in settings such as our own.

Although not statistically significant, participants from Masaka were less likely to die than those from Entebbe. Masaka is a rural setting, while Entebbe and its catchment area is mainly semi-urban. Possibly, the social set up of people in rural settings in Masaka enabled closer social support by family members for those that were sick or recovering from disease compared to those from Entebbe. Presence of social support has previously been reported to be associated with better health outcomes [32, 33].

We did not find that factors previously reported by others to be associated with short-term mortality such as white blood cell count in CSF, haemoglobin count, age, and altered mental status, affected longer-term mortality in our study [19, 28]. Such associations may diminish in importance as patients survive for longer periods.

We were unable to measure other factors that could have a bearing on long-term mortality such as antifungal treatment resistance. Antifungal susceptibility is considered to be relevant to treatment response in relapsed disease, but susceptibility testing at the point of first diagnosis appears to have little utility; few data exist on which to make a judgement regarding its ability to determine long term outcomes [34]. It would have been good to explore the effect of CD4 counts on mortality. Unfortunately, CD4 counts were not measured at baseline in the CryptoDex study. The requirement was for participants to report their last CD4 counts at enrolment. Many had forgotten/had no record or presented those done a long time prior. It would have also been good to see the effect of baseline viral loads on mortality. These were however not tested in the CryptoDex trial.

Conclusion

This study shows that long-term mortality in CCM patients, even among those receiving recommended therapy remains high. Adjunctive therapy with dexamethasone does not have a significant effect on mortality at two years, while wasted patients, those who present with impaired consciousness as defined by Glasgow Coma score <15, and those with histories of convulsions, are more likely to die.

We recommend extensive implementation of Cryptococcal screening in low resource settings to prevent CCM whose management remains a daunting and challenging experience. There is a pressing need to develop better treatment for those patients who present with the disease.

Acknowledgments

We acknowledge the study participants and staff of the MRC/UVRI & LSHTM Uganda Research Unit who participated in data collection.