The cost and cost-effectiveness of rapid testing strategies for yaws diagnosis and surveillance

Abstract

Background: Yaws is a non-venereal treponemal infection caused by Treponema pallidum subspecies pertenue. The disease is targeted by WHO for eradication by 2020. Rapid diagnostic tests (RDTs) are envisaged for confirmation of clinical cases during treatment campaigns and for certification of the interruption of transmission. Yaws testing requires both treponemal (trep) and non-treponemal (non-trep) assays for diagnosis of current infection. We evaluate a sequential testing strategy (using a treponemal RDT before a trep/non-trep RDT) in terms of cost and cost-effectiveness, relative to a single-assay combined testing strategy (using the trep/non-trep RDT alone), for two use cases: individual diagnosis and community surveillance.

Methods: We use cohort decision analysis to examine the diagnostic and cost outcomes. We estimate cost and cost-effectiveness of the alternative testing strategies at different levels of prevalence of past/current infection and current infection under each use case. We take the perspective of the global yaws eradication programme. We calculate the total number of correct diagnoses for each strategy over a range of plausible prevalences. We employ probabilistic sensitivity analysis (PSA) to account for uncertainty and report 95% intervals.

Results: At current prices of the treponemal and trep/non-trep RDTs, the sequential strategy is cost-saving for individual diagnosis at prevalence of past/current infection less than 85% (81-90); it is cost-saving for surveillance at less than 100%. The threshold price of the trep/non-trep RDT (below which the sequential strategy would no longer be cost-saving) is US$ 1.08 (1.02-1.14) for individual diagnosis at high prevalence of past/current infection (51%) and US$ 0.54 (0.52-0.56) for community surveillance at low prevalence (15%).

Discussion: We find that the sequential strategy is cost-saving for both diagnosis and surveillance in most relevant settings. In the absence of evidence assessing relative performance (sensitivity and specificity), cost-effectiveness is uncertain. However, the conditions under which the combined test only strategy might be more cost-effective than the sequential strategy are limited. A cheaper trep/non-trep RDT is needed, costing no more than US$ 0.50-1.00, depending on the use case. Our results will help enhance the cost-effectiveness of yaws programmes in the 13 countries known to be currently endemic. It will also inform efforts in the much larger group of 71 countries with a history of yaws, many of which will have to undertake surveillance to confirm the interruption of transmission.

Fig 1. Diagram of alternative testing strategies: Combined (panel A) and sequential (panel B).

Boxes represent tests and test results; diamonds represent diagnoses; dotted lines represent discordant treponemal and non-treponemal test results by the trep/non-trep RDT, excluded from Fig 2 for simplicity; in the case of discordant results between the treponemal RDT and the trep/non-trep RDT, the sequential testing strategy takes the result of the trep/non-trep RDT.

Fig 2. Decision model of diagnosis of current infection under the combined testing strategy.

The decision to be made is whether an individual has current infection or not (box); the individual’s combined test result (circles) can be either: dually positive (suggesting current infection), treponemal positive only (past infection), or dually negative (never any infection); this depends on the Sensitivity (Se) and Specificity (Sp) of the treponemal line (Se2a and Sp2a, respectively) and of the non-treponemal line (Se2b and Sp2b), as well as on the prevalence of past/current infection in the total population (Pr1) and the prevalence of current infection in the population of past/current infections (Pr2); the treponemal line provides either a true or false diagnosis of past/current infection; this depends on Positive Predictive Value (PPV) and Negative Predictive Value (NPV) of the treponemal line (PPV2a and NPV2a, respectively); PPV is calculated as (Se×Pr1)÷(Se×Pr1+(1–Sp)×(1–Pr1)); NPV is calculated as Sp×(1–Pr1)÷((1-Se) ×Pr1+Sp×(1–Pr1)); the non-treponemal line provides either a true of false diagnosis of current infection; this depends on PPV and NPV of the non-treponemal line (PPV2b and NPV2b), using the prevalence of current infection in the population of past/current infections (Pr2); red lines indicate pathways to false diagnoses of current infection; note, the treponemal line can give a false positive diagnosis of past/current infection and yet the non-treponemal line can still give a correct overall diagnosis of no current infection.

Fig 3. Decision model of diagnosis of current infection under the sequential testing strategy.

In the sequential strategy, the individual’s first test result (circles) can be either: treponemal positive (suggesting past/current infection), or treponemal negative (never any infection); this depends on the Sensitivity (Se) and Specificity (Sp) of the treponemal RDT (Se1 and Sp1, respectively), as well as on the prevalence of past/current infection in the total population (Pr1); the treponemal line provides either a true or false diagnosis of past/current infection; this depends on Positive Predictive Value (PPV) and Negative Predictive Value (NPV) of the treponemal RDT (PPV1 and NPV1, respectively); PPV is calculated as (Se×Pr1)÷(Se×Pr1+(1–Sp)×(1–Pr1)); NPV is calculated as Sp×(1–Pr1)÷((1-Se) ×Pr1+Sp×(1–Pr1)); those with a true negative result for past/current infection are logically also truly negative for current infection; among those with a false negative result for past/current infection, we assume that the prevalence of current infection among false past/current infection negatives is the prevalence of current infection in the population of past/current infections (Pr2); those with either true or false positive results for past/current infection receive a second test; their second test result can be either: dually positive (suggesting current infection), treponemal positive only (past infection), or dually negative (never any infection); for simplicity, we assume that we use only the non-treponemal line; the non-treponemal line provides either a true of false diagnosis of current infection; this depends on PPV and NPV of the non-treponemal line (PPV2b and NPV2b),using the prevalence of current infection in the population of past/current infections (Pr2); red lines indicate pathways to false diagnoses of current infection.

Fig 4. Cost savings of the sequential testing strategy across different scenarios of prevalence, per 1000 people tested, by use case.

Best—best estimate (median); Low—low estimate (2.5th centile); High—high estimate (97.5th centile); cost savings are expressed per 1000 people tested.

Fig 5. Incremental cost-effectiveness (cost per correct diagnosis gained) of the combined testing strategy across different scenarios of prevalence, by use case.

Best—best estimate (median); Low—low estimate (2.5th centile); High—high estimate (97.5th centile); ICER—Incremental Cost Effectiveness Ratio (cost per correct diagnosis); black rectangles indicate where the sequential strategy may be more costly and more effective; grey areas without an ICER value indicate negative ICERs, where the combined testing strategy is less effective and more costly or more effective and less costly.