Approximately 90% of children with cancer reside in low-income and middle-income countries (LMIC) where healthcare resources are scarce and allocation decisions difficult. The cost effectiveness of treating childhood cancers in these settings is unknown. The objective of the present work was to determine cost-effectiveness thresholds for common paediatric cancers using acute lymphoblastic leukaemia (ALL) in Brazil and Burkitt lymphoma (BL) in Malawi as examples. Disability-adjusted life years (DALYs) prevented by treatment were compared to the gross domestic product (GDP) per capita of each country to define cost-effectiveness thresholds using WHO-CHOICE (‘CHOosing Interventions that are Cost-Effective’) guidelines. The case examples were selected due to the data available and because ALL and BL both have the potential to yield significant health gains at a low cost per patient treated. The key findings were as follows: the 3:1 cost/DALY prevented to GDP/capita ratio for ALL in Brazil was US$771 225; expenditures below this threshold were cost effective. Costs below US$257 075 (1:1 ratio) were considered very cost effective. Analogous thresholds for BL in Malawi were US$42 729 and US$14 243. Actual costs were far less. In Brazil, US$16 700 was spent to treat each patient while in Malawi total drug costs were less than US$50 per child. In summary, treatment of certain paediatric cancers in LMIC is very cost effective. Future research should evaluate actual treatment and infrastructure expenditures to help guide policymakers.
- Pediatric Oncology
- Global Health
- Cost-Effectiveness Analysis
- Burden of Disease
Statistics from Altmetric.com
Worldwide, an estimated 175 000 children under the age of 15 are diagnosed as having cancer each year.1 In the USA and other high-income countries (HIC), about 90% of children with the most common types of malignancies such as acute lymphoblastic leukaemia (ALL), Burkitt lymphoma (BL) and Wilms tumour survive long term and with minimal disability.2–4 However, while overall childhood cancer cure rates in HIC approach 80%, event-free survival rates in low-income and middle-low-income countries (LMIC) range from 5% to 40%.5–7
Adult breast, cervical, lung and colorectal cancers in LMIC have recently received increased attention.8–12 Due to the recognition that lifestyle related risk factors and aging populations are fuelling an adult cancer epidemic globally, initial approaches for combating adult cancers in LMIC have included population-based prevention and screening interventions.13 Paediatric oncology poses a different set of challenges, especially among LMIC populations where 90% of the world's children live.14 Unlike adults, low-cost prevention strategies are neither economical nor logistically feasible for childhood cancers.15–17
Because resource-intensive medical services, such as chemotherapy and hospitalisation, are needed to cure childhood cancers, financial constraints are often cited as a barrier to their successful treatment in LMIC. Policymakers, clinicians and industry representatives use multiple inputs such as disease burden and societal preference in addition to cost estimates when defining medical priorities in LMIC.18 However, a critical evaluation of the costs required to treat these conditions is lacking and the case for their cost effectiveness has not been made. To guide prioritisation efforts in LMIC, we determined the thresholds of cost effectiveness for the treatment of paediatric cancer, using ALL in Brazil and BL in Malawi as examples, and compared these results to the actual costs of treatment in each setting.
Using standard methods from the WHO Global Burden of Disease working group, the total number of disability-adjusted life years (DALYs) due to ALL in Brazil and BL in Malawi were calculated.19 DALYs are a standard metric used to describe the number of years of life lost (YLL) due to ill health, disability or early death. This value is calculated by adding the YLL with the years lived with disability (YLD).20 The number of YLL per case without treatment was first calculated by subtracting the standard life expectancy of either Brazil (72.79 years) or Malawi (52.31 years) from the estimated age at death without treatment from either ALL (5.65 years) or BL (7 years), respectively.21 This value was then multiplied by the absolute annual number of cases of either ALL in Brazil (1344) or BL in Malawi (447) in children 15 years of age or younger to provide a national annual estimate of the number of YLL if no children were treated.1 The number of YLD per case was obtained by multiplying the estimated duration of illness with and without intervention by the disability weights for leukaemia (0.09) and non-Hodgkin's lymphoma (0.06) as set by the Global Burden of Disease working group.22 National estimates of YLD with and without treatment were then calculated by multiplying these values by the incident number of cases for ALL or BL.
To estimate the cost effectiveness of treating either ALL in Brazil or BL in Malawi, the WHO-CHOICE (‘CHOosing Interventions that are Cost-Effective’) framework for generalised cost-effectiveness analysis (CEA) was used.23 By this method, cost effectiveness is defined as the ratio of the monetary expense required to avert 1 DALY to the annual gross domestic product (GDP) per capita of a given country or region. A ratio of 1:1 is considered very cost effective and a ratio of 3:1 is cost effective.23 All GDP estimates and costs are presented in 2011 US dollars unless otherwise stated.
As suggested in a number of guidelines including the Global Burden of Disease Study and WHO-CHOICE, a continuous discount rate of 3% with non-uniform age weighting was also used for standard DALY calculations.20 ,23 Discounting is the process of converting future costs and benefits to their present value.23 To account for the uncertainty associated with discounting, a sensitivity analysis was conducted on all calculations with the discounting rate set at lower and upper bounds of 0% or 6%, respectively as suggested by the WHO-CHOICE guidelines.23 Uniform age weights were also applied when the discount rate of 0% was used in order to provide a scenario with no life value judgments (ie, the value of a year of life was considered to be the same, regardless of the age of the person). Costs were not discounted, however, because treatments for ALL and BL represent one-time investments.
All data were obtained from international databases and published studies (table 1). Country-specific data were used when available. However, for the proportion of leukaemia in Brazil that was ALL, data from the USA were used.24
Country and cancer case examples
The cases of ALL in Brazil and BL in Malawi were chosen as two examples since cost and outcome data were available in these two cases.25 ,26 Several factors make the decision to focus on these cancers a useful exercise. Since both have well above 80% event-free survival rates in HIC, each offers the potential for large health gains (63% event-free survival was used for ALL in Brazil and 48% for BL in Malawi).2 ,3 In addition, each causes a significant paediatric cancer burden in their respective countries. Each year, ALL comprises 32% (1344/4198) of all cancers presenting between 0–15 years of age in Brazil while BL makes up 46% (447/978) of cancers in the same age group in Malawi.1 ,21 Finally, since minimal to no surgical or radiation oncology services are needed to treat ALL and BL using modern protocols, they represent cancers where the required treatment can likely be successfully delivered in LMIC. It is important to note, however, only one study was used for each example and may not represent the overall outcomes for either disease at the national level.
In keeping with the natural history of ALL and BL, a death rate of 100% among all children not treated or who fail chemotherapeutic intervention was assumed. Patients not receiving treatment for ALL were assumed to survive for 3 months after diagnosis while those with untreated BL were assumed to live for 1 month in accordance with the expected clinical course of both diseases. When treatment was available, it was also assumed all incident cases in a country were treated on the same protocol at an equal cost regardless of location or type of healthcare setting. Since the interventions evaluated lead to relatively small changes in population-based death rates, a population model to obtain an accurate estimate of health gain was not used. To account for inter-cancer and outcome variability due to treatment, however, cure and lost-to-follow-up rates were included.
Risk stratification was not accounted for as patient outcomes were reported for the entire cohort in both studies, better reflecting the overall situation in Brazil and Malawi. Potential late effects of treatment such as second malignancies, cardiac and respiratory dysfunction, infertility and neurocognitive delays were also excluded from the model because they are uncommon in children with ALL and BL treated in HIC and their effect on the model would be negligible.21 ,27 ,28 It is worth noting that cranial irradiation was not used in the Brazilian study of children with ALL, thereby excluding a potentially significant cause of late effects.25 ,26
The calculations for both case-example models are presented in online supplementary appendices A and B. The primary outcomes reported were the upper limits of ‘cost effective’ and ‘very cost effective’ for the treatment of ALL in Brazil and BL in Malawi (3:1 and 1:1 ratio of cost per DALY averted to per capita GDP, respectively). YLL, YLD and the number of DALYs averted by treating all incident cases were also reported. The results of the Brazilian ALL model are summarised in table 2 and the BL model for Malawi in table 3.
The US dollar figures presented in both tables represent the amount of money a cancer programme could spend to treat a single child with ALL or BL in each respective setting yet still remain either cost effective or very cost effective as per WHO-CHOICE guidelines. Any amount spent lower than these thresholds would therefore meet established criteria for cost effectiveness. It is important to note that the incidence of disease does not impact cost thresholds because it is not a factor in the calculation of individual DALYs.
Further examining the contributing factors to the DALY calculation illustrates that the majority of health gains were due to reductions in the YLL. The contribution of YLD to the overall number of DALYs was less than 0.5% for both forms of cancer. Over the course of the 2 years a patient was expected to receive treatment for ALL in Brazil, 0.14 DALYs were attained. For the 2 months of treatment time factored into the Malawi model for BL yielded the equivalent of a half-week DALY. Unlike the cost thresholds, the total DALYs were directly affected by the incidence of disease.
The long-term benefits of treatment were most affected by discounting. The difference in YLL between 6% and 0% discounting was approximately 46.9 years for the Brazilian model. The same range of YLL due to discounting in the Malawian model was approximately 24.9 years. Discounting had minimal influence on YLD with no change in either model since these benefits were small and immediately obtained. A wide range in the sensitivity analysis was also noted for all cost-effectiveness thresholds. For example, there was a US$1 million difference between 0% and 6% discounting in the upper limit of cost effective in the Brazilian model.
Using the above methodology, the cost-effectiveness thresholds for ALL, BL and Wilms tumour were also calculated for sample study results from other countries. The results of this analysis are presented in table 4 and further detailed against actual costs in figure 1.
It is a common assumption that treating children with cancer in LMIC is not cost effective. The models presented in this study aimed to evaluate the balance between health gains and costs by providing cost-effectiveness guidelines for the treatment of ALL and BL in Brazil and Malawi, respectively. Through this generalised CEA approach, this study puts the cost effectiveness of childhood cancer treatment squarely into the prioritisation debate alongside the cost-effectiveness ratios for other childhood illnesses in LMIC.
We used the case examples of ALL in Brazil and BL in Malawi to examine how much the governments of these countries could theoretically spend and still maintain cost effectiveness. The results of our analysis challenge the assumption that it is too expensive to treat childhood cancers in LMIC. Patients in Recife, Brazil achieved a 63% 5-year event-free survival rate at a cost of US$16 700 per patient (based on 2004 US$ with inflation not reflected).1 ,25 This represents only 6% of the threshold for being very cost effective by the standard WHO definition and an incremental cost-effectiveness ratio, defined as the cost per DALY prevented by treatment, of US$477.1 ,23
BL in Malawi provides another example of how adapted treatment protocols may work in low-income countries. In Western Europe, BL carries a >90% event-free survival rate due to intensive treatment, extensive supportive care and multiple inpatient stays.29 These protocols may be prohibitively expensive in many LMIC settings. Using a short-course (30 days) regimen in Blantyre, Malawi, however, 48% of children with BL were cured.26 The cost of chemotherapeutic and supportive care drugs was reported as less than US$50 per child, representing less than 1% of the calculated US$14 243 threshold for very cost effective BL treatment in Malawi.25 ,30
Several limitations of this study deserve mention. First, the actual incidence of leukaemia and non-Hodgkin's lymphoma may be greater than that referenced due to underdiagnosis and under-reporting.1 ,6 This would consequently underestimate the true burden of disease as calculated in the models. Also, as described above in the Methods, late effects were not included. If such data were available, it would likely have decreased the number of DALYs saved and subsequently decreased the cost-effectiveness thresholds. However, since the true costs reported from Brazil were much lower than the calculated thresholds, the conclusions would likely not have changed. Additionally, the impact of discounting on the sensitivity analyses of YLL and the cost-effectiveness thresholds was significant. The ranges of the reported results highlight the anticipated uncertainties that come with future economic projections in CEA models.
Another major limitation of this study is that the model calculates per-incident fixed costs and does not include variable costs required to treat paediatric cancers. For example, infrastructure expenditures such as inpatient ward space, healthcare worker training, supply chains and radiology/laboratory equipment may be substantial. However, these costs will vary depending on the current state of the local health system, the type of cancer and the treatment protocol used. For instance, the services required to cure BL using the protocol from Hesseling et al are modest compared to the neurosurgical and hospital resources needed in HIC settings to treat many brain tumours. In addition, the extended benefits from improved laboratory capabilities may strengthen overall health systems and lead to additional health gains for other populations.
Paediatric oncology as a global priority?
The cost-effectiveness thresholds for treating a child with either ALL in Brazil or BL in Malawi as calculated in this paper highlight the potential role that paediatric oncology could play within the larger global health prioritisation framework. Over the past decade, the fight against adult cancer has expanded beyond the borders of HIC and into a global priority.8 Much of this effort has focused on the prevention of lung and cervical cancers and low-cost screening/prevention options for cervical, breast and colorectal cancers.9 Paediatric cancer, however, has remained isolated from the larger debate and understudied despite the potential for high cure rates and large numbers of DALYs saved per case. This study is an important first step towards placing paediatric oncology treatment within the wider debate about global health priority setting in LMIC.
Despite the cost effectiveness of treating certain paediatric cancers, the decision of whether to attempt to treat them in LMIC is complex and does not depend solely on costs. Much like tuberculosis and malaria, many childhood cancers are curable. However, unlike these diseases and HIV, the burden of disease measured in population based DALYs is comparatively small. This fact is compounded by the reality that many LMIC governments cannot implement every cost-effective intervention due to limited resources. Whether the funds used to cure one child with cancer could have saved more lives had they been spent on other interventions also remains an important question. It is important to note, however, that paediatric cancer treatment programmes in LMIC have already shown the ability to generate significant new philanthropic and grant revenue streams that supplement governmental support without detracting from other areas.7 ,31 These initial experiences suggest additional funding options may exist by engaging a new cadre of donors not currently involved in global health.
Over the next 50 years, the burden of disease in LMIC is forecast to shift from predominantly communicable to non-communicable diseases.20 Over the same period of time, some LMIC may experience increasing economic growth and the ability to further invest in health infrastructure. Before the widescale implementation of treatment programmes for childhood cancer are possible, however, defining what cancers to treat and where is a necessary first step.
From this point, further evaluating disease specific CEA models, larger sectorial analyses comparing multiple interventions, correlating real versus estimated costs and estimating required start-up investments is necessary. To do these analyses, however, more outcomes and costing data are necessary. Even at centres in LMIC where paediatric oncology treatment is available, challenges with rigorous, complete and long-term data collection is a known barrier to clinical research.32 Also, among the few published studies available, event-free and overall survival rates are usually the primary outcome with costs rarely reported.
Ultimately, the central long-term question for paediatric cancer in LMIC is how to implement national treatment programmes. The predominant paradigm to implement and improve paediatric cancer care delivery in LMIC over the past 20 years has been the twinning model.33 Twinning, or formalised partnerships between healthcare institutions in HIC and LMIC, has reduced cancer-related mortality significantly where implemented.34 These twinning programmes have included activities such as implementing locally adapted treatment protocols, education programmes, leadership development, telemedicine and fundraising support. However, there are few examples in LMIC of how to implement paediatric oncology services at the national level. While twinning may work well for pilot programmes and specific paediatric cancer units, new macro implementation strategies that directly engage government ministries are needed. In addition, transnational collaborative networks modelled after Children's Oncology Group in the USA, the Brazilian Childhood Cooperative Group for ALL Treatment or the Société Internationale d'Oncologie Pédiatrique, could also benefit areas of Latin America, Asia and Africa through shared educational opportunities and larger research cohorts. Finally, a focus on local quality improvement is needed. Simple interventions, such as parental education at the start of treatment, are inexpensive yet very effective strategies that can yield a significant return on investment by increasing the cost-effectiveness thresholds through higher event-free survival rates.35 ,36
The calculated cost-effectiveness thresholds in this paper suggest that it is feasible and very cost effective to treat certain childhood cancers in LMIC. Further exploration of this topic will help guide health policymakers incorporate paediatric oncology into their national health priority decision-making process.
We would like to thank Sheena Mukkada for her editing assistance.
Contributors NB is the guarantor of this article. All other authors have reviewed and contributed to this submitted manuscript.
Funding ALCM is funded by a fellowship from the University of Sydney. SG is supported by a doctoral fellowship award from the Canadian Institutes for Health Research. SCH is funded by the American Lebanese Syrian Associated Charities and the NIH Cancer Center Core Grant CA 21765.
Competing interests None.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.