Our analysis suggests that a 7-d course of amphotericin (1 mg/kg/d) with adjunctive high dose fluconazole (1,200 mg/d) for at least 2 wk is the most cost-effective cryptococcal induction treatment, with a cost-effectiveness ratio of US$20 per QALY gained. The short-course amphotericin regimen bridges the large cost disparity between the oral regimens and the 2-wk amphotericin regimens, and thus far seems to be similarly efficacious as the 2-wk amphotericin regimens in resource-constrained settings. Two-week amphotericin regimens are the most efficacious in high-income countries and are the first-line recommendation per WHO and US guidelines 
. In low-income countries, however, 2 wk of amphotericin, along with hospitalization and monitoring of side effects, is unacceptably expensive and resource-intensive. Conversely, fluconazole therapy alone is a more common, accessible option given its affordability. Unfortunately, low cost does not make fluconazole monotherapy an optimal induction treatment regimen because other regimens, such as short-course amphotericin, appear to bridge the expanse between the efficacy extremes that currently exist. For example, fluconazole monotherapy has nearly 2-fold higher 10-wk mortality and 30% higher absolute mortality than short-course amphotericin. The number needed to treat is an incredibly low 3.5 persons treated using short-course amphotericin instead of fluconazole monotherapy to save one additional life. However, more evidence for the efficacy of short-course amphotericin with adjunctive fluconazole is needed. This is the only study to our knowledge that weighs the costs and benefits of care for CM in order to decipher which alternatives are most efficient from a policy standpoint.
While clearly requiring more infrastructure than an oral regimen alone, short-course amphotericin may be feasible for sites with limited amphotericin supplies, limited laboratory capacity, limited hospital bed space, and healthcare worker shortages. Short-course amphotericin with fluconazole (1,200 mg/d) can be safely given with potassium supplementation but without any routine laboratory monitoring 
, as the cumulative amphotericin-related nephrotoxicity with urinary electrolyte wasting does not typically begin until after 5 d of amphotericin 
. As amphotericin results in logarithmic CSF clearance of fungi 
, the maximal clinical and cost-effective benefit may be realized with short-course therapy, whereas beyond 1 wk, the further benefit of amphotericin may be countered by its toxicity, particularly in resource-limited settings. Head-to-head trials to better determine the optimal duration of amphotericin induction therapy are essential.
The limitations of this analysis stem mainly from the lack of high-quality evidence supporting the short-course amphotericin regimen. There are only four studies anywhere that have evaluated this strategy, with a total of 127 participants. Thus, due to lack of power, these studies were pooled for the purposes of this analysis. Each study used a slightly different amphotericin B deoxycholate regimen: Bicanic et al. 
used 1.0 mg/kg/d alone for 7 d, Tansuphaswadikul et al. 
used 0.7 mg/kg/d alone for 7 d, and Muzoora et al. 
and Jackson et al. 
, respectively, used 5 and 7 d of amphotericin at 1.0 mg/kg/d with fluconazole at 1,200 mg daily in divided doses for 14 d 
. The moderately wide CIs pertaining to 10-wk mortality for all regimens, but especially short-course amphotericin (95% CI: 19%–35%), highlight the need for more clinical trials to evaluate the efficacy of this regimen. From the above-mentioned studies, one cannot draw evidence-based conclusions on superior efficacy for short-course (7-d) amphotericin with fluconazole; however, this regimen does appear to be the most cost-effective and clearly better than fluconazole monotherapy. Furthermore, there may be subtle differences between sites that limit precise cross-comparisons, such as stage of CM at presentation and management of elevated intracranial pressure. However, one illuminating experience is a series of three prospective studies in Mbarara, Uganda, between 2005 and 2012, in which the 10-wk mortality was 54% (31/57) with fluconazole (800–1,200 mg/d) monotherapy 
, 28% (8/29) with 5-d amphotericin + fluconazole (1,200 mg/d) 
, and 40% (16/40) with 14-d amphotericin + fluconazole (800 mg/d) in 2011–2012 (D. R. B., unpublished data from NCT01075152).
In addition, we have compared our cost–benefit analysis to the WHO-defined cost-effectiveness thresholds, which may be insufficient markers of cost-effectiveness. All of the treatment regimens analyzed here met the WHO criteria for being “highly cost effective.” However, clearly they are not equally efficacious regimens. In fact, our analysis suggests that within those “highly cost effective” regimens, some options are more efficacious than others, and merely following these WHO thresholds may lead policy-makers to institute a “cost-effective” strategy with lower efficacy, thereby ignoring superior regimens available with minimal additional cost. Additionally, from a policy perspective, we considered the cost-effectiveness only of CM induction treatment. Preventative health measures such as early ART (before AIDS) and targeted screening of asymptomatic cryptococcal antigenemia in individuals with a CD4 count<100 cells/µl are certainly more cost-effective to prevent disease than any of the aforementioned treatment options after cryptococcal disease has occurred 
As with any cost-effective analysis, what may be cost-effective may not be affordable for an individual patient or for a specific healthcare center, especially given that the Ugandan per capita expenditure on health in 2009 was approximately US$115 
. The purpose of this analysis is to inform stakeholders regarding where to invest resources on a national level. This analysis suggests that investing in building capacity for short-course amphotericin may be a cost-effective long-term strategy. Inclusion into the analysis of free fluconazole via the Diflucan Partnership Program, which is not available in all countries, does not change the ICER between fluconazole (1,200 mg/d) and short-course amphotericin + fluconazole (1,200 mg/d), as both regimens utilize the same amount of fluconazole. For stakeholder investment, the first resource to invest in is diagnostic capacity in order to diagnose cryptococcosis, such as the new point-of-care CRAG lateral flow assay (US$2 per test, Immy) 
, which is less expensive than traditional CRAG latex agglutination (US$13.50–US$16.25 in Malawi and Uganda), as the latex agglutination assay requires cold-chain shipping, which dramatically increases the real-world CRAG cost in Africa. Yet, many additional implementation science questions remain regarding how best to build capacity at rural healthcare facilities to administer short-course amphotericin. It may be most efficient to invest in larger referral centers that already have the capacity for amphotericin. Perhaps switching from a 14-d course to a 7-d course would free up healthcare workers, reduce burden on the laboratory, extend the benefit of amphotericin to more patients, and make more bed space available for other hospital admissions. If only larger centers are able to administer amphotericin, resources would have to be invested into triage and transport of patients with suspected CM at smaller, remote sites. The alternative is to develop the capacity to administer short-course amphotericin at these smaller rural centers, which would require a reliable supply and distribution of amphotericin, and further investment in healthcare workers for inpatient care of hospitalized patients.
One caveat is that control of intracranial pressure is equally important as the pharmaceutical regimen chosen for CM treatment 
. Simply switching to 7 d of amphotericin will not mirror the reported outcomes without intracranial pressure control. The vast majority of studies included in this analysis controlled intracranial pressure through repeated therapeutic LPs, although the frequency of therapeutic LPs was not reported in each study. The assumption in this cost-effectiveness analysis was an average of three LPs (one diagnostic and two therapeutic) that would need to be performed to control intracranial pressure. The cost of manometers ranged from 5%–16% of the total cost of care. However, manometers, which support safe and accurate control of intracranial pressure, are generally unavailable in most healthcare centers and referral hospitals throughout Africa. Therefore, the scope of any capacity building must include supplies for safe LPs. In the absence of manometers, one approach could be to pre-screen with the point-of-care CRAG lateral flow assay (by fingerstick, plasma, serum, or urine) before LP 
, and then to either prioritize which patients to use a manometer on or empirically remove 20 ml of CSF with repeated therapeutic LP in 24–48 h in confirmed CM 
. More operational research on the management of intracranial pressure in resource-limited settings is also urgently needed, including, for example, the use of high-quality IV tubing as a possible improvised manometer or surrogate noninvasive measures such as intraocular pressure 
This analysis is certainly generalizable to sub-Saharan African countries with respect to estimates of 10-wk mortality and projected estimates of long-term survival, as all of the referenced papers were taken from similar resource-limited settings with a significant burden of CM. In sub-Saharan Africa, patients often present with more advanced cryptococcosis than in the US or Europe, as evidenced by longer duration of symptoms, higher CRAG titers, higher intracranial pressure, and greater proportion with altered mental status 
. Variation in the cost-effectiveness ratio within Africa is likely, as the actual costs of hospital personnel vary from country to country, and even within a single country, depending on the funding of the healthcare facility. Although astoundingly low, the personnel costs utilized in this analysis are correct and based on Ugandan national government public-sector salaries, where salaries are approximately 5-fold lower than in private-sector hospitals. When private-sector salaries are considered instead, the proportion of total costs accountable to personnel costs would increase from 8% to 32% for the short-course amphotericin regimen, with a resulting cost–benefit ratio of US$27/QALY and no change in the ICER versus fluconazole monotherapy. In South Africa, the total cost of caring for a patient with CM is estimated at US$2,883 
. While medications, supplies, and laboratory costs in South Africa are similar to our estimates, the specific regimen and cost components are unclear. Personnel costs are approximately 15-fold higher in South Africa and 4-fold higher in Rwanda than in the Uganda public sector, and healthcare worker to patient ratios are substantially better in South Africa, which may account for the difference in total cost of CM care. Still, there is no difference in the personnel necessary (and thereby the costs) between the fluconazole monotherapy and 7-d amphotericin + fluconazole regimens, thus the ICER remains unchanged even with varying personnel costs.
There are approximately 720,000 cases of CM annually in sub-Saharan Africa 
. The WHO guidelines recommend amphotericin-based regimens, as they are clearly most effective. Many stakeholders instead rely on fluconazole monotherapy because it is more accessible, has lower upfront costs, and lacks the lab monitoring needs of amphotericin treatment, despite 25%–30% absolute higher mortality. We believe this is ineffective public health policy. Amphotericin should be moved from the “complementary list” to the “core list” in the WHO Model List of Essential Medications
. Our analysis suggests that using 7-d amphotericin (1 mg/kg/d) coupled with fungicidal doses of fluconazole (1,200 mg/d) for 2 wk appears to offer similar survival benefit but without the toxicity of 14 d of amphotericin in resource-constrained settings. Large-scale studies in resource-limited areas must be prioritized to determine efficacy, side effects, and risk of relapse with short-course amphotericin compared to more traditional 14-d amphotericin regimens. If the efficacy of short-course amphotericin is better determined, and survival is in fact ~30% better than with fluconazole monotherapy, then moving to short-course amphotericin could save ~150,000 lives annually (95% CI: 85,000 to 225,000) in sub-Saharan Africa, at a cost of US$220 per life saved.