Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Jan 1.
Published in final edited form as: Lancet Infect Dis. 2017 May;17(5):482. doi: 10.1016/S1473-3099(17)30192-5

Methodological concerns regarding a PrEP model

Sally Blower 1,*, Laurence Palk 1
PMCID: PMC5748340  NIHMSID: NIHMS928232  PMID: 28447954

To the Editors

We would like to raise a concern regarding the recent publication by Nichols and colleagues1 entitled “Cost-effectiveness analysis of pre-exposure prophylaxis for HIV-1 prevention in the Netherlands: a mathematical modelling study”. Specifically, we have found that using the methodology they describe in their paper we are unable to replicate their results. In their modeling study the authors divide the population into four risk groups based on sexual behavior. They used a standard method, Monte Carlo filtering2-5, for the first stage of their analysis: to calibrate their model. To conduct the Monte Carlo filtering the authors assigned ranges to specify the size of the four risk groups, and then sampled each of these four ranges 100,000 times. The ranges the authors used were: 5 to 15% for the highest risk group, 30 to 50% for the second highest, 10 to 35% for the third highest and 4 to 46% for the lowest risk group. We have rerun the authors' analysis using the same ranges and sampling each of the four ranges 100,000 times. However, we have been unable to replicate the authors results. Our results are shown in the Figure. We found that none of the 100,000 samples could divide the population into four risk groups. In ∼56,000 of our samples (red data) the population would have to be divided into at least five risk groups; this can be seen as the percentage of the population in the four risk groups adds up to less than 100% of the population. In the other ∼44,000 of our samples (blue data) there are too many individuals assigned to at least one of the four risk groups; this can be seen as the percentage of the population in the four risk groups adds up to greater than 100% of the population. Mathematical modeling studies can be extremely useful for predicting the impact of pre-exposure prophylaxis on controlling HIV epidemics in both resource-rich3 and resource-constrained4 countries. However, it is extremely important that studies can be replicated in order to assess their validity. Using the methodology described in the study by Nichols and colleagues1, we have been unable to replicate their findings.

Figure. Results generated from 100 000 samples.

Figure

Open in a new tab

References

  • 1.Nichols BE, Boucher CA, van der Valk M, Rijnders BJ, van de Vijver DA. Cost-effectiveness analysis of pre-exposure prophylaxis for HIV-1 prevention in the Netherlands: a mathematical modelling study. Lancet Infect Dis. 2016;16(12):1423–1429. doi: 10.1016/S1473-3099(16)30311-5. [DOI] [PubMed] [Google Scholar]
  • 2.Rose KA, Smith EP, Gardner RH, Brenkert AL, Bartell SM. Parameter sensitivities, Monte Carlo filtering, and model forecasting under uncertainty. J Forecast. 2006;10:117–133. [Google Scholar]
  • 3.Supervie V, García-Lerma JG, Heneine W, Blower S. HIV, transmitted drug resistance, and the paradox of pre-exposure prophylaxis. Proc Natl Acad Sci USA. 2010;107(27):12381–6. doi: 10.1073/pnas.1006061107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Supervie V, Barrett M, Kahn JS, Musuka G, Moeti TL, Busang L, Blower S. Modeling dynamic interactions between pre-exposure prophylaxis interventions & treatment programs: predicting HIV transmission & resistance. Nature Sci Rep. 2011;1:185. doi: 10.1038/srep00185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Gerberry DJ, Wagner BG, Garcia-Lerma JG, Heneine W, Blower S. Using geospatial modelling to optimize the rollout of antiretroviral-based pre-exposure HIV interventions in Sub-Saharan Africa. Nat Commun. 2014;5:5454. doi: 10.1038/ncomms6454. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES