The comparatively high cost of health care, the stubbornly high burden of communicable and growing burden of non-communicable disease, and a sluggish economy puts mounting pressure on the South African government to use available financial and human resources in the most effective manner.
Globally, there is a buzz about Precision Medicine and how it can be used to enhance the pharmacoeconomics of medical treatments. Despite its merits in principle the field appears to be somewhat ‘lost in translation’ owing to the ambiguous use of terminology (Precision vs Stratified vs Personalised) and a focus on high-profile research efforts as opposed to demonstrating the benefits through end-user (patient) centric initiatives. To the uninitiated, it may seem as if the Genomic Medicine hype of old has been relabeled while the underlying issues haven’t been resolved.
As a result, the benefits that could be gained from ‘Precision Medicine’ initiatives, including for resource-constrained and developing economic settings, may not materialize. As a consequence, funders may be reluctant to put financial support behind dedicated programs, creating a lose-lose scenario for all stakeholders that could be avoided!
This post explores the development of a tailored medical treatment program based on the preemptive use of existing SNP (Single Nucleotide Polymorphism)-genotyping testing assays in a target population of 200,000. It explores the full-economic cost implications of such a program, including end-user (patient) distribution, not just the cost of producing raw data (e.g. genome sequences).
It is also an attempt to build a framework for analyzing the cost-benefit implications of Precision Medicine efforts that are based on using information-rich, big-data generating applications in Medicine and Health.
The case presented is based on the following premises:
- It assumes that a systematic testing program of national interest is run in a systematic fashion but looks at an initial target population of 200,000 for illustration purposes. It is an economic cost-benefit analysis rather than a business case.
- The effectiveness of a drug treatment is subject to the presence or absence of certain genetic markers in DME (drug metabolising enzyme) genes and varies from patient to patient (a person’s individual DME genotype). These SNP markers can be mapped by way of using adequate DNA testing methodologies, a discipline called Pharmacogenetics.
- Tailoring type and dosage of a drug treatment to an individual patient’s phenotype is one of the hallmarks of Precision Medicine, generating benefits to patients, medical aid providers, and drug makers. The case therefore aims to explore benefits for a number of stakeholders.
- SNP-genotyping assays can be used in a customised and cost-effective fashion, adopting assay content to local needs and carrying out tests in a central lab that is supplied with samples from a number of access points across the country.
- While existing genotype-phenotype correlations have been largely established in non-African populations, a sufficient subset of these correlations will have utility in the South African context. For the purpose of this case study, a 75% effectiveness of existing panels has been assumed; 25% of the panel constitutes waste, initially.
- In order to enhance the effectiveness of SNP genotyping panels used in African populations, further dedicated research may be necessary. The case presented below is based on the sequencing of DME genes in 5% of a 200,000 size target population. Therefore, the research effort is treated as a means to an end, not an end in itself.
- SNP-genotyping assays are carried out in a preemptive, not ad-hoc, fashion in the target population, making the relevant information readily available at the point-of-care/point-of-decision making.
- Relevant clinical information is gathered as part of the program. In contrast to the pharmacogenetic information, the corresponding clinical data will be gathered in an ad-hoc fashion since these will depend on the type of drug treatment employed. This will require efficient clinical information management systems and proper data storage and management. A program of this nature will benefit from a standard list of complementary assays (measuring other biomarkers in blood or urine) that can be performed with respect to specific drug treatments.
- As a crude approximation, benefits of such a program can be measured in the form of savings made from using effective and avoiding the use of ineffective treatments, and from a decrease in the number of sick days. In both instances, a positive effect on GDP can be determined. Consequently, the net benefit of the program can be determined. Other benefits, though not considered in further detail, can materialize in the form of R&D investment inflows (made by pharma companies) and spin-off innovation (in the form of new IP generated and being commercialised, start-up companies being formed or new employment being created) that may emanate from utilising large genotype-phenotype data-sets.
- The case is not perfect but meant to stimulate discussion and creative thinking. Detailed data used in the formulation of the financial models below can be made available upon request.
- Further analysis, in particular related to high end Genomics applications (next-generation sequencing), is needed and will be done.
Health & pharmaceutical spending in SA
In 2011, total spend on health was USD 36 -billion – or around 9%of GDP , way above the 5% recommended by the World Health Organisation (WHO).  Expenditure for Pharmaceuticals was in the order of USD 3,74 billion in the same year.  At a population size of 52 M , the per capita cost of health expenditure in 2011 has been USD 720,-, and the per capita cost of drug treatments was USD 75,-.
Figure 1: Health & pharmaceutical expenditure in South Africa
For national treasuries, health care providers and medical insurers choosing the best possible treatment for a given indication is a major objective when assessing the relative merits of two or more treatments (e.g. high blood pressure). The discipline that deals with this practically is Pharmacoeconomics . It compares the cost-benefit of using one product in favour of another in view of distinct clinical outcomes, price and available budget. The example below is for illustrative purposes and has been derived and modified from . In the example, treatment A is preferable over B owing to the bottom line benefit it creates.
Table 1: Illustrative Pharmacoeconomics example
Precision medicine refers to the tailoring of medical treatment to the individual characteristics of each patient. It does not literally mean the creation of drugs that are unique to a patient, but rather the ability to classify individuals into subpopulations that differ in their susceptibility to a particular disease, in the biology and/or prognosis of those diseases they may develop or in their response to a specific treatment . Often used interchangeably with ‘Stratified Medicine’ or ‘Personalized Medicine’, the core concept is increasingly relevant for researchers, drug developers, policy makers and health care providers alike .
While this is a challenging paradigm for drug developers in view of expiring drug patents, poor pipeline performance and increasing pressure from payers to demonstrate the cost-benefit of a new drug application, arguments have been made that end-users stand to benefit from the new paradigm, but also the developers themselves (this is due to the fact that a potential loss in sales volume in a stratified medicine scenario will be compensated by increased patient (and doctor) compliance, or better inclusion of otherwise under-served patients) .
Lost in Translation?
I see at least 2 issues with Precision Medicine right now: (i) it’s a niche effort despite claims that the concept has been around for some time. The language that is used by its proponents isn’t designed to facilitate broad-based engagement, which is not assisted by that same community’s ongoing struggle to define or agree on the terminology it uses (Stratified vs Personalized vs Precision vs P4 Medicine ); (ii) it lacks precise case-making in terms of the impact it can potentially make for patients, payers and the public at large. This is perhaps due to the fact that it is currently centered on large-scale efforts of human genome sequencing (not exclusively but to a large degree), to decipher in more detail the biological foundation of human diseases such as cancer, as opposed to projects that clearly demonstrate how large-scale systematic efforts of genetic screening (or similar) can make an impact right away, right now.
The problem with drugs is that they usually work only in a fraction of patients. This lack of effectiveness varies with disease indication and drug application but it can affect anywhere between 38% and 75% of a target patient population . Figure 2 (derived from (12) illustrates this point.
Figure 2: Percentage of the patient population for which a particular drug is ineffective, on average
Pharmacogenomics is the study of genetic variations that influence individual response to drugs. Pharmacogenomics combines traditional pharmaceutical sciences such as biochemistry with an understanding of common DNA variations in the human genome. The most common variations in the human genome are called single nucleotide polymorphisms (SNPs). There is estimated to be approximately 11 million SNPs in the human population, with an average of one every 1,300 base pairs. An individual’s response to a drug is often linked to these common DNA variations. In a similar manner, susceptibility to certain diseases is also influenced by common DNA variations .
To put this into context, the cost alone of adverse drug reactions in the US is more than US 20 bn per annum . Therefore, Pharmacogenomics is of importance for regulators , the research community , drug makers and national health providers .
The cost-benefit (cost to the payer when a medicine is applied without test vs cost incurred for a medical treatment with a dedicated test) of a Pharmacogenomic application is usually the subject of intensive study. Cost-benefit ratios have been determined in areas such as cancer ; however, such analyses have been difficult to perform for cases with limited data. Cost-benefit ratios have been shown to be small to moderate in the past ; however, this was at a time when genomic implications where in an early stage of maturity and costs-per-data-point where much higher than today.
A schematic explaining drug effectiveness in a target population is provided in Figure 3.
Figure 3: Tailored Medicine effect
Pharmacogenomics in Africa
As bad as the ineffectiveness of existing medicines is in a developed world context, the problem case is exacerbated in emerging countries, and in particular so in Africa. This is mainly due to (i) the distinct degree of genomic complexity and the diversity of African populations, (ii) the extensive lack of knowledge about genotype-drug phenotype relations and (iii) the fact that most existing drugs have been researched and developed against the backdrop of a Caucasian gene pool. Globally, the widespread adoption of PGx testing in the clinic is hampered by regulatory and issue about reimbursement, amongst others . While these factors do play a role in the (South) African context they are certainly exacerbated by issues related to cost, availability of adequate financial resources and the necessary human talent to support effective implementation in often under-capacitated health care settings (12). While other countries have dealt more extensively with the implementation of Pharmacogenetics from a practical point of view (16), a large-scale systematic effort is to date amiss in the African context, despite the benefits it could produce.
The cost/benefit of Precision Medicine in South Africa – an exemplary case
In view of the issues discussed above I have sketched a picture of how a pharmacogenetic testing solution could be implemented practically in South Africa. The case presented below is comprised of 4 components: (i) lab testing, (ii) R&D, (iii) distribution, and, (iv), benefits. Details of what is included in each of the 4 components are provided as a separate Appendix.
A summary of the value flows in the initiative is provided in Figure 4. Briefly, patients with a need (pain) approach doctors who prescribe tailored treatments based on available pharmacogenetic data and, if needed, complementary lab results. In case of the latter, samples will be sent through access points for processing and, importantly, data capturing and storage. Actionable pharmacogenetic reports are available to doctors directly (in case of simple analyses, such as for a pharmacogenetic panel including up to 158 SNPs) or through a dedicated access point. The latter scenario will apply when more complex tests are being performed, e.g. for tumour DNA sequencing. However, for the purpose of this case only relatively ‘simple’ reports with utility for doctors have considered. Nevertheless, all data captured is stored centrally to allow for proper management, continuous updating (in view of expanding genotype-drug phenotype knowledge) and downstream exploitation (‘big data’ research projects).
Figure 4: Value flows in pharmacogenetic testing program
From a SNP-genotyping point of view, the case is based on the systematic SNP (Single Nucleotide Polymorphism) genotyping of DNA samples in a target population of 200,000. In order to account for different input cost scenarios tests employing 16, 34, 128 and 158 SNPs each were considered (component ‘lab testing’). These numbers are based on optimal assay design configurations, not scientific considerations. However, a main assumption underlying this scenario is that pre-emptive genotyping with a higher number of SNPs is beneficial over an ad-hoc testing approach. If one considers an initiative of national interest, run as a pilot initially, it will be better to include a larger number of DME (drug metabolizing enzyme) SNPs to carry out cost-benefit analyses for a number of genotype-drug phenotype pairings; of course, the scenario can be modified for any number of SNPs and/or for a specific medical indication, such as HIV or cancer.
Owing to the fact that existing SNP genotyping panels have been developed against the backdrop of Caucasian populations, I have assumed that 25% of the initially used SNP tests will be ineffective, therefore constituting waste. For the same reason, I have assumed that there is a need for dedicated research (to better understand pharmacogenetics in African populations and to enhance the SNP genotyping panel). I have therefore included an R&D project aimed at the targeted DNA sequencing of relevant pharmacogenes in 5% of the 200,000 target population (component ‘R&D’). If in the course of preparing such a testing program more dedicated SNP genotype information in African population comes available, the R&D component can be omitted or modified.
In addition, I have included costs for the distribution of a PGx solution to the end-user (patient), via 3 access points (one each in the Western Cape, Gauteng and KZN Province) and a network of associated doctors (component ‘distribution’). The main vehicle of delivery are doctors’ offices, in turn supported by ‘genomic medicine centres’ (knowledge access points) to assist with test interpretation and training of doctors, among others. Distribution via pharmacies or other means is conceivable but hasn’t been considered in detail.
Finally, I have considered a few beneficial effects, i.e. drug cost savings in 1% of the target population (assuming that wasteful expenditure won’t occur with a pharmacogenomics testing program in place) and enhanced labor productivity, both translating into a positive effect on GDP (Gross Domestic Product).
As indicated by data shown in Figure 5, the benefits gained from implementing a PGx solution exceed the costs of creating it (shown for the 158 SNP scenario). In essence, over a ten year period a cost of R 307 M is offset by R 712 M in benefits, producing a R 405 M in net benefit. Importantly, the bar graph takes into account the time effect of the benefits that can be gained. Since the DNA information in question won’t change over time the data generated can be re-used repeatedly, generating positive impact in future instances of treatment. Therefore, maintenance, management and distribution of data will be a key success factor if multiplier positive effects are to be gained. Considering a one-year time horizon only, the scheme would be loss-making (not shown).
Figure 5: Cost/benefit of PGx testing over time (cost-benefit effect in ZAR M) 
A comparison of the different SNP panel scenarios indicates that all of them would create a positive return (expressed as net benefit and NPV, net present value).
Table 2: SNP panel scenario benefit comparison (10 year horizon)
The benefits derived from a Precision Medicine program are highlighted in the following non-exhaustive summary:
- For national treasury and health care providers, programs that aim at tailoring the development or use of medical treatments to the genetic traits of a target population, benefits will materialize in the form of cost savings and a positive contribution to GDP (lower number of sick leave days).
- For drug developers, access to a national pharmacogenetic knowledge repository (with good quality genotype-phenotype data) can make South Africa more attractive for the development of new, or the re-purposing of existing drugs. This will result in pharma R&D investment inflows, adding positively to our foreign currency account in the short term and to our trade balance on the long-term (because more pharma R&D value will be produced locally). If access to patients (with supporting longitudinal clinical information) can be provided through dedicated access points, South Africa’s rating as an attractive location for clinical drug development may go up even further.
- Overall, such a scheme could be a boon for researchers when increasingly larger-scale sets of data become available for interrogation. As suggested above the scheme can become a basis for further genomic research efforts in academia but also to stimulate pre-clinical drug development efforts by pharmaceutical industry. Overall, the understanding of disease development and treatment in local populations will go up, benefiting society as a whole.
- For medical insurers, extensive drug genotype-phenotype information will facilitate the optimisation of treatment prioritisation algorithms, benefitting the insurer and the patient at the same time because new and potentially more expensive treatments can be made available to more patients while costs will be saved in other areas of the business.
 Jones, J. S. (2006). Pharmacoeconomics in South Africa. South African medical journal= Suid-Afrikaanse tydskrif vir geneeskunde,96(2), 96-96.
 National Research Council (US) Committee on A Framework for Developing a New Taxonomy of Disease. (2011). Toward precision medicine: building a knowledge network for biomedical research and a new taxonomy of disease. National Academies Press (US).
 Collins, F. S., & Varmus, H. (2015). A new initiative on precision medicine. New England Journal of Medicine.
 Trusheim, M. R., Berndt, E. R., & Douglas, F. L. (2007). Stratified medicine: strategic and economic implications of combining drugs and clinical biomarkers. Nature Reviews Drug Discovery, 6(4), 287-293.
 New England Health Institute. Preventing medication errors: a $21 billion opportunity. www.nehi.net/bendthecurve/sup/documents/Medication_Errors_%20Brief.pdf.
 Case studies and cost-benefit analysis of HER2 and TPMT in Four EU member states. http://ftp.jrc.es/EURdoc/eur22214wp2.pdf
 Stallings, S. C., Huse, D., Finkelstein, S. N., Crown, W. H., Witt, W. P., Maguire, J., … & Ginsburg, G. S. (2006). A framework to evaluate the economic impact of pharmacogenomics. Future Medicine
 Lam, Y. W. (2013). Scientific challenges and implementation barriers to translation of pharmacogenomics in clinical practice.ISRN pharmacology, 2013.
 Pharmacogenetic data are created in year 1 of the initiative, forming the basis of continuous intervention and further lab testing over the remaining 9 years of the program time horizon. All data will be centrally stored and managed but made available to relevant users (access points, doctors, patients) through adequate distribution channels (e.g. web-based or mobile device interfaces).
 Van Driest, S. L., Shi, Y., Bowton, E. A., Schildcrout, J. S., Peterson, J. F., Pulley, J., … & Roden, D. M. (2014). Clinically actionable genotypes among 10,000 patients with preemptive pharmacogenomic testing. Clinical Pharmacology & Therapeutics, 95(4), 423-431.
 For simplicity, running costs were extrapolated from Y1 (Year 1) to Y2, and Y3, at an inflationary rate. R&D costs where not extrapolated from Y1 to Y2 and Y3; the same applies to the penalty charge for ineffective genotyping (assuming that the R&D outcome from Y1 will be used to optimise a genotyping panel in Y2). This is because the DNA-based information generated in Y1 will be relevant in the same form in Y2 and Y3, resulting in same or similar benefits without incurring the same costs in the original target population. Of course, in Y2 another population of 200,000 could be tested, doubling the benefit produced in Y1. If one would work on the premise that the entire target population would undergo a different treatment in Y2 or Y3 (and assuming that relevant DME SNPs were included in the original test panel), significant cost saving effects could be achieved for other treatments as well (not shown).
 NPV, Net Present Value. NPV was calculated over a 10-yaer time horizon, using a 10% risk free rate.