August 2022
Kaposi sarcoma is one of the most common cancers in sub-Saharan Africa and is associated with substantial morbidity and mortality.1 Kaposi sarcoma is caused by the Kaposi’s sarcoma herpesvirus. Although Kaposi sarcoma was recognised in endemic form before the HIV epidemic, HIV has driven an epidemic, AIDS-associated form of the disease.2
Given the occurrence of Kaposi sarcoma in marginalised communities, research has been chronically underfunded. However, the preferred first-line therapies for advanced stage Kaposi sarcoma in high-income countries, where costs of therapy are infrequently the limiting factor, are pegylated liposomal doxorubicin (PLD) or paclitaxel.3
In sub-Saharan Africa, long-term, prospective data on Kaposi sarcoma outcomes is scarce, although recent studies have improved our understanding of chemotherapy responses.4, 5 Specifically, a recent AIDS Clinical Trials Group and AIDS Malignancy Consortium trial,4 done across sites in Kenya, Malawi, Uganda, South Africa, Zimbabwe, and Brazil, randomly assigned participants with advanced Kaposi’s sarcoma to paclitaxel, bleomycin–vincristine, or etoposide. The primary outcome was progression-free survival (PFS) at 48 weeks. Paclitaxel was superior (48-week PFS 64%) to bleomycin–vincristine (44%) and etoposide (20%). To our knowledge, there is no randomised prospective data on PLD in sub-Saharan Africa,6 although a randomised controlled trial is planned by the AIDS Malignancy Consortium. Despite the clinical superiority of paclitaxel, implementation of paclitaxel has been slow.
In this issue of The Lancet Global Health, Esther Freeman and colleagues7 employ published clinical data from sub-Saharan Africa and costing data from Kenya to model the cost-effectiveness of four chemotherapy regimens (ie, PLD, etoposide, bleomycin–vincristine, and paclitaxel). The authors showed that paclitaxel would be more effective than bleomycin–vincristine or etoposide and would increase life expectancy by 4·2 years per person. Furthermore, paclitaxel would be the most cost-effective strategy (incremental cost-effectiveness ratio of US$380 per year-of-life-saved compared with bleomycin-vincristine) and would remain cost-effective across a range of scenarios. PLD would further increase life expectancy by 0·6 years per person but did not meet their definition of cost-effectiveness. Implementing paclitaxel instead of bleomycin–vincristine would save approximately 6400 life-years and would increase overall 5-year Kenyan health-care costs by $3·7 million. The authors should be commended for a massive undertaking of costing, modelling, and sensitivity analyses of their study.
The study has several important strengths. First, the data collection is transparent and well reported. Second, extensive sensitivity analyses to account for uncertainty are well described. Finally, the use of an HIV-specific microsimulation model provides validity to modelling assumptions of long-term outcomes and competing risks.
The largest weakness of the model, however, is the assumption that patients who are progression free at 48 weeks will remain progression free. This is based on data from high-income countries where outcomes could differ due to differences in Kaposi’s sarcoma herpesvirus or HIV genetics, level of immunosuppression (eg, CD4 count), antiretroviral therapy use at the time of the study, or supportive care. Given this weakness, the use of subsequent chemotherapy and effects on life-years saved might all be substantially different than reported.
Cost-utility analyses are expected to adhere to recommendations from the Second Panel on Cost-Effectiveness Analysis.8 The study by Freeman and colleagues largely adheres to the recommendations; however, it does diverge in important ways. First, the authors report their primary outcome in incremental cost-effectiveness ratio of cost per life-year saved as opposed to cost per quality-adjusted or disability-adjusted life-year. On one hand, given the lifetime time horizon and similar toxicity profiles, the quality of life decrease of chemotherapy treatment is likely to be negligible, but on the other hand, those that relapse earlier will have decreased quality of life.9 The overall effect is hard to anticipate without modelling. Second, they use a definition of cost-effectiveness of 0·5 times the gross domestic product (GDP) per capita, which is a more stringent standard than the WHO recommendation of 3 times the GDP per capita. By the recommended definition, in fact, LD would be cost-effective compared with paclitaxel and is on the borderline of being extremely cost-effective by WHO’s definition of 1 times the GDP per capita.
This publication provides compelling evidence to support the implementation of paclitaxel across sub-Saharan Africa and the authors should be commended for the inclusion of a budget-impact analysis. At the same time, ultimately, the data reported here show the shortcomings shared by many cost-utility analyses. The measure of cost-utility can only be as good and as precise as the data underlying it. Investment in meticulous, prospective data collection is needed to refine our understanding of Kaposi sarcoma in sub-Saharan Africa.
As a takeaway, the results support a principle for cancer therapies in resource-limited settings: improvements in quality or quantity of life can be both the financial and humanitarian best buy. Patients who live longer and live better will benefit society through work, taxes, raising children,10 and improving communities. Currently, the burden of proof lies with researchers to show cost-effectiveness. However, it is time to shift the burden of proof; for cancer interventions that improve the quality or quantity of life, the onus should be on payers to show when the cost is overwhelming and why they have chosen not to pay for it. And when therapies are too expensive, it is the responsibility of us all to apply pressure on those who set the price, so the best therapies are available everywhere in the world.
MSP receives grant funding, including salary support, from the National Institutes of Health (U54CA254564 and K01TW011470), which supported this work.