Clinical Consequences of Using an Indeterminate Range for Early Infant Diagnosis of HIV: A Decision Model : JAIDS Journal of Acquired Immune Deficiency Syndromes

Secondary Logo

Journal Logo

Clinical Science

Clinical Consequences of Using an Indeterminate Range for Early Infant Diagnosis of HIV: A Decision Model

Salvatore, Phillip PhD, SMa,b; Johnson, Karl BSc; Vojnov, Lara PhDd; Doherty, Meg MD, PhD, MPHd; Dowdy, David MD, PhD, ScMa,e

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes 82(3):p 287-296, November 1, 2019. | DOI: 10.1097/QAI.0000000000002155

Abstract

INTRODUCTION

As programs to prevent mother-to-child transmission of HIV are scaled-up globally, the prevalence of HIV among exposed infants born to HIV-infected mothers will continue to decrease.1 However, with low levels of perinatal transmission and imperfect diagnostic tests, a substantial fraction of positive HIV test results may occur in infants who are truly uninfected,2 with the potential that many of these infants will receive long-term antiretroviral therapy (ART) as a result. The World Health Organization (WHO) recommends that positive HIV test results be confirmed by a second independent test on a new sample.3,4 However, in some high-burden settings, the proportion of infants receiving appropriate confirmatory testing is less than 20%,5–8 creating challenges for clinicians who must weigh the risks of false-positive diagnoses9 against the high mortality faced by infants with untreated HIV.10

In settings where confirmatory testing is not routinely conducted, clinical decisions may be aided by additional semi-quantitative information which accompanies the results of positive nucleic acid amplification tests (NAATs) generally used for early infant diagnosis. This semi-quantitative result, reported as a cycle threshold (Ct), reflects the number of amplification cycles completed before the detection of viral nucleic acid: A low Ct indicates more rapid detection (corresponding to a higher viral load), and a high Ct indicates slower detection (corresponding to a lower viral load).11 Currently, manufacturers recommend that any HIV nucleic acid detected (regardless of the Ct value) may be interpreted as a qualitative “positive” result.12–14 However, classifying the weakest results (ie, highest Ct values) as “indeterminate” may allow programs to triage infants with such results for repeat or confirmatory testing before initiating ART.15–17

The implications of incorporating these semi-quantitative results have not been fully explored. We therefore constructed a decision analysis model to estimate the clinical consequences of adopting guidelines using an indeterminate range for early infant HIV diagnosis in settings of incomplete confirmatory testing of positive results. We present anticipated tradeoffs between false-positive diagnoses and potential missed cases at a variety of Ct values that may be used to define indeterminate results.

METHODS

Population of Interest and Model Structure

Our analysis traced the economic and clinical outcomes of a hypothetical population of 10,000 HIV-exposed infants in sub-Saharan Africa with an HIV prevalence of 2.5% among infants at the time of initial EID testing. In this population, we evaluated the performance of 2 diagnostic and treatment algorithms (Fig. 1): (1) standard of care (SoC) versus (2) the adoption of an Indeterminate Range, defined as a Ct value above a given threshold on NAAT testing for early infant HIV diagnosis.

F1
FIGURE 1.:
Test results and treatment decisions for simulated HIV-exposed infants under standard of care and Indeterminate Range algorithms. A, Infants with HIV-infected mothers are assumed to have a given prevalence of HIV. The result of each infant's initial test is determined by the infant's HIV status, the sensitivity/specificity of the assay, and (in the Indeterminate Range scenario) the distribution of Ct values among nonnegative results. B, Decreasing Ct values (equivalent to increasing viral loads) are plotted from left to right. The red area represents negative results below the limit of detection of the assay. In the standard of care algorithm (C), any nonnegative result (blue area) receives presumptive ART unless (and until) a confirmatory test returns a negative result. In the Indeterminate Range algorithm (D), Ct values below the threshold (more rapid detection, blue) trigger presumptive treatment unless (and until) a confirmatory test returns a negative result, whereas Ct values above the threshold (slower detection, purple) do not receive ART unless a confirmatory test is performed which returns a nonnegative result. The key clinical difference between algorithms—care for infants with high Ct results who do not receive confirmatory testing—is outlined with a dotted box. In the scenario under which the indeterminate range is used to prioritize confirmatory testing, all indeterminate results receive confirmatory testing but no frankly positive results (low Ct) receive confirmatory testing (illustrated by solid paths).

In both algorithms, all HIV-exposed infants received an initial NAAT for HIV diagnosis within 6 weeks of birth. The sensitivity and specificity of this assay, coupled with the infant's HIV status (Fig. 1A), determine the probability of a negative test result (defined as a result with no HIV nucleic acid detected and no Ct reported, depicted as the white area to the left in Fig. 1B). In the SoC algorithm, all detectable (nonnegative) results are classified as positive; in the Indeterminate Range scenario, nonnegative test results are classified as either positive (a Ct value lower than the selected Ct cutoff) or indeterminate (a Ct value higher than the selected Ct cutoff). For each possible Ct cutoff, the probabilities of positive versus indeterminate results for HIV-infected and HIV-uninfected infants (dark gray and light gray areas in Fig. 1B, respectively) are determined by empirical distributions of Ct values from a meta-analysis (see Figure S1, Supplemental Digital Content, https://links.lww.com/QAI/B371 and Luo et al, “Use of an indeterminate range in HIV early infant diagnosis: a systematic review and meta-analysis” submitted with this manuscript).

The SoC diagnostic algorithm used for our analysis was adopted from contemporary WHO guidelines for early infant diagnosis of HIV (Fig. 1C).3 Under the SoC, infants with any initial nonnegative HIV test result (whether due to HIV infection or false-positive results) immediately receive ART and, according to WHO guidelines, should have a confirmatory test performed. However, in the SoC algorithm, a proportion of these infants do not receive confirmatory testing and thus are assumed to carry an HIV diagnosis (and ART recommendation) for life. The remaining infants have samples sent for confirmation, with confirmatory test results arriving after a protracted delay. While the length of these delays may vary across settings, we assume an average delay of 6 months, based on available data.18 The confirmatory test is assumed to be definitive, with ART interrupted after a negative result and continued after a positive result. Infants with an initial negative test are not treated or confirmed.

In the Indeterminate Range algorithm (Fig. 1D), positive results (Ct values lower than the selected cutoff) are treated the same as under the SoC. However, infants with indeterminate test results do not receive presumptive ART and only receive a lifelong HIV diagnosis and ART recommendation if a confirmatory test also returns a nonnegative result. Importantly, infants with indeterminate test results for whom confirmatory testing is not performed do not receive ART (highlighted in orange). We present results from 2 separate comparisons of the Indeterminate Range algorithm to the SoC. In the first, we assume that a given percentage of infants with nonnegative results (15% in the base case) will receive confirmatory testing, regardless of Ct value. However, this strategy may not be the most efficient use of limited confirmation capacity, as those infants with indeterminate results (high Ct values) are most likely to be falsely positive. In the second comparison, we assume that, in the Indeterminate Range scenario, all infants with indeterminate results will receive a confirmatory test, and we compare this to confirmatory testing of an equal number of randomly selected infants in the SoC (reflected in the use of dashed and solid lines in Fig. 1D).

In both scenarios, we assume for simplicity that HIV-infected infants who initially receive a false-negative diagnosis will obtain a correct diagnosis at the age of 2 years (due to, eg, serological testing at the conclusion of breastfeeding).3 We therefore assume that these infants experience a mortality risk associated with untreated HIV for 2 years.10 All HIV-infected infants (and HIV-uninfected infants with false-positive results) who survive to age 2 are assumed to have a lifelong indication for ART and to incur annual costs associated with HIV care throughout life (modeled as an average of 35 years longer for HIV-uninfected infants).19,20 Greater loss to follow-up before confirmatory testing is expected under the Indeterminate Range due to mortality among the additional missed HIV cases; therefore, we assume that such infants experience elevated mortality for 6 months to account for delays in treatment initiation. Further details on treatment and mortality may be found in the Supplementary Methods, Supplemental Digital Content, https://links.lww.com/QAI/B371.

Clinical Outcomes and Cost-Effectiveness Outcomes

To characterize the clinical and economic outcomes of a policy switch from SoC to implementation of an Indeterminate Range algorithm, we calculated the incremental number of infants unnecessarily put on treatment, missed cases, and HIV-related deaths in the Indeterminate Range scenario relative to SoC. Because the use of an Indeterminate Range often reduces the number of infants unnecessarily put on treatment compared with SoC, these are presented as negative—beneficial—values. Our primary outcomes of interest were the incremental number of HIV-uninfected infants unnecessarily receiving long-term ART (false-positives), the incremental number of HIV-related deaths, and the ratio of incremental unnecessary ART courses to incremental HIV-related deaths in the Indeterminate Range scenario compared to the SoC. A higher ratio can be interpreted as more favorable for the Indeterminate Range scenario (preventing more unnecessary treatments for each additional death that occurs). This ratio is highest (most favorable to the Indeterminate Range) when the Ct cutoff is stringent (giving fewer indeterminate results) and the prevalence of HIV is low. As the aspirational value of an Indeterminate Range policy is to repurpose funding saved (by minimizing unnecessary ART) for more effective HIV prevention and control activities, our secondary outcomes of interest included the incremental cost-effectiveness ratio (ICER) measured as the incremental cost per disability-adjusted life year (DALY) averted (see the Supplementary Methods for further details, Supplemental Digital Content, https://links.lww.com/QAI/B371). Because the SoC is expected to have better health outcomes, but to be more expensive, ICERs are reported as the (negative) incremental cost of an Indeterminate Range policy over the (positive) additional DALYs compared with SoC. More negative (larger magnitude) ICERs can be interpreted as more favorable to an Indeterminate Range policy than less negative ICERs.

Model costs included testing, presumptive ART care, and life-long HIV care. Unit costs of testing were based on multi-country estimates assuming 80% instrument use (personal communication, “All-in Roche EID Cost” via the Clinton Health Access Initiative). Costs of ART care in the first 2 years of life were derived from estimates of the mean total cost per patient-year.21 After age 2, each surviving patient receiving ART incurs an estimated (per patient) lifetime cost of HIV care beyond age 2.19 See Table S1, Supplemental Digital Content, https://links.lww.com/QAI/B371 for results from a wide range of lifetime costs.

In each scenario, DALYs are estimated to account for morbidity across the life course (described in the Supplementary Methods, Supplemental Digital Content, https://links.lww.com/QAI/B371). We considered a lifetime time horizon with costs and utilities discounted at 3% annually. Based on the life expectancy and disability weights in our model, the DALYs lost because of one HIV-related infant death are comparable to the DALYs lost because of 19.6 infants receiving ART unnecessarily (95% UR: 17.9–21.7). This threshold is highly dependent on life expectancies, economic discounting, and the subjective valuation of receiving ART unnecessarily. Any such threshold should be interpreted cautiously.

Sensitivity Analyses

For each scenario modeled, 10,000 sets of parameter values were created by probabilistically sampling from the parameter ranges defined in Table 1 (assuming uniform distributions unless noted), and outcomes were simulated for each parameter set. Outcome values are reported for the 50th, 2.5th, and 97.5th percentiles from these simulations.

T1
TABLE 1.:
Model Parameters and Sensitivity Sampling Ranges

Multivariable sensitivity analysis was performed using nonparametric partial rank correlation coefficients with respect to our primary outcome of interest and to our cost-effectiveness outcome (see Figure S2, Supplemental Digital Content, https://links.lww.com/QAI/B371). Secondary one-way sensitivity analyses evaluated the impact of HIV prevalence (see Figure S3, Supplemental Digital Content, https://links.lww.com/QAI/B371) and the independent probability of confirmatory testing (see Figures S4 and S6, Supplemental Digital Content, https://links.lww.com/QAI/B371) on model outcomes.

RESULTS

Clinical Outcomes and Incremental Effectiveness

In our reference scenario, we estimated that, for every 10,000 HIV-exposed infants tested, 250 would be HIV-infected. Under the SoC, we projected that 2.3 (95% UR: 1.8–3.1) of these 250 would test false-negative and 20.0 (95% UR: 18.2–21.9) would die within 2 years. Of the remaining 9750 HIV-uninfected infants, assuming that 15% of all initially nonnegative tests would receive confirmatory testing, 11.6 infants (95% UR: 9.3–15.4) would test false-positive and be placed on treatment unnecessarily under the SoC (Table 2). (For reference, the performance of the SoC with 100% confirmation is presented in Table S2, Supplemental Digital Content, https://links.lww.com/QAI/B371). Adopting an Indeterminate Range algorithm range defined by Ct ≥33 with 15% confirmatory testing resulted in an estimated median of 20.5 initially false-negative diagnoses (95% UR: 16.8–25.0), resulting in 28.1 (95% UR: 25.3–31.2) HIV-related deaths, while reducing the number of unnecessary treatments to 0.8 (95% UR: 0.6–1.1). More limited (higher) Ct cutoff values resulted in fewer HIV-infected infants receiving false-negative diagnoses, but more infants receiving treatment unnecessarily (Table 2).

T2
TABLE 2.:
Clinic Outcomes and Incremental Effectiveness of Indeterminate Range Algorithms

By contrast, if confirmatory testing were performed on all infants with indeterminate results defined by Ct ≥33, an estimated median of 0.9 HIV-uninfected infants would be placed on ART unnecessarily and 21.3 HIV-infected infants would die (Table 2). In the corresponding SoC algorithm with an equal number of confirmatory tests as the Ct ≥33 estimate, we estimate a median of 11.9 uninfected infants would be placed on ART unnecessarily, and 20.0 HIV-infected infants would die. In this scenario, therefore, adopting an Indeterminate Range defined by Ct ≥33 would avert 11.0 (95% UR: 8.8–14.3) unnecessary treatments while incurring 1.2 (95% UR: 0.9–1.7) additional deaths: a ratio of 8.8 unnecessary treatments averted for each excess (95% UR: 6.1–13.4). Figure 2 illustrates incremental clinical outcomes of adopting an Indeterminate Range algorithm, assuming that all infants with indeterminate results are prioritized for confirmatory testing (and compared against an SoC where the same number of confirmatory tests are performed), and Figure 3 illustrates the corresponding ratio of unnecessary treatments averted to excess deaths at different levels of HIV prevalence.

F2
FIGURE 2.:
Impact of prioritizing infants with indeterminate results for confirmatory testing. Clinical outcomes are compared for simulations following the SoC algorithm versus the Indeterminate Range algorithm in which the cutoff for an indeterminate result was varied from a Ct of ≥36 to a Ct of ≥30. Bars represent median estimates (from 10,000 simulations) under the Indeterminate Range algorithm at each cutoff; dotted lines represent the median estimates under the SoC; and arrows and labels indicate the incremental difference between the Indeterminate Range algorithm and SoC. “Unnecessary ART” (in blue) represents the median number of false-positive diagnoses who receive a lifelong indication for ART. “Missed HIV cases” (in green) represents the number of false-negative diagnoses who do not receive presumptive ART. “HIV-Related Deaths” (in purple) represents the number of HIV-infected infants who die before 2 years of age. As lower (more inclusive) Ct values are adopted, more HIV-uninfected infants fall in the Indeterminate Range, resulting in fewer unnecessary ART regimens, but more missed HIV cases and HIV-related deaths.
F3
FIGURE 3.:
Incremental clinical consequences after adoption of an Indeterminate Range. The ratio of additional unnecessary ART regimens averted to additional excess HIV-related deaths (with both values being measured relative to standard of care, SoC) is plotted for the use of an Indeterminate Range with cutoffs from Ct ≥36 to Ct ≥30. A higher value for this ratio indicates more unnecessary ART regimens averted for each additional death due to missed HIV diagnosis, indicating a more favorable tradeoff for using the Indeterminate Range. Results from the 15% confirmatory testing scenario are plotted in (A) for HIV prevalence (among tested infants) of 1%, 2.5%, and 5%. Results for the scenario in which all infants with indeterminate results are prioritized for confirmatory testing are plotted in (B). To illustrate the relative scales of (A and B), the results of the primary analysis using a cutoff of Ct ≥36 and the bounding box in (A) are reproduced in (B). Points represent median values of 10,000 simulations and error bars represent 95% uncertainty ranges.

Sensitivity Analyses and Cost-Effectiveness Analysis

The most important drivers of the ratio of unnecessary ART averted to excess HIV deaths, as assessed by partial rank correlation, were the prevalence of HIV (PRCC = 0.84) and the proportion of test results from HIV-infected infants falling above the Indeterminate Range cutoff (PRCC = 0.82). In addition, ICER values were strongly correlated with the assumed lifetime cost of HIV care and life-expectancy of HIV-infected infants (see Figure S2, Supplemental Digital Content, https://links.lww.com/QAI/B371 for more details). For further results detailing the impact of HIV prevalence, see Figure S3, Supplemental Digital Content, https://links.lww.com/QAI/B371.

Implementing an Indeterminate Range as described in the 15% confirmatory testing analysis (Ct cutoff of ≥33, HIV prevalence 2.5%) would result in an estimated $201,700 (95% UR: $165,200–$249,800) in cumulative savings compared with SoC, and prioritizing indeterminate results for confirmatory testing would result in $128,200 (95% UR: $103,600–$164,400) in savings (Table 3). Most incremental costs resulted from long-term ART, whereas most incremental DALYs reflected HIV-related mortality (see Figure S2, Supplemental Digital Content, https://links.lww.com/QAI/B371). Lowering lifetime costs of ART among HIV-uninfected patients (eg, allowing patients to be lost to care) reduced the incremental savings of the Indeterminate Range scenario, thereby moderately reducing its cost-effectiveness (see Figure S7, Supplemental Digital Content, https://links.lww.com/QAI/B371). Assuming 15% of nonnegative results receive confirmatory testing, the ICERs of the Indeterminate Range relative to SoC ranged from $929 (95% UR: $1102–$793) per DALY (Ct ≥ 30) to $17,508 (2.5th percentile: $179,316) per DALY (Ct ≥36; Table 3 and see Figure S8, Supplemental Digital Content, https://links.lww.com/QAI/B371 for further details). In the scenario where all infants with indeterminate results are prioritized for confirmatory testing, the ICERs of the Indeterminate Range ranged from $2050 (95% UR: $2894–$1568) per DALY (Ct ≥30) to $9451 (95% UR: $32,983–$5010) per DALY (Ct ≥33). At Ct cutoffs above 33, fewer results are classified as indeterminate (and those which are indeterminate have a higher probability of being HIV-uninfected). Therefore, as the cutoff increases, the relative cost-effectiveness of the SoC decreases (Table 3).

T3
TABLE 3.:
Incremental Cost-Effectiveness of Indeterminate Range Algorithms

DISCUSSION

This simplified decision model illustrates the tradeoffs associated with using semi-quantitative HIV testing results to define an Indeterminate Range for early infant HIV diagnosis. Our results suggest that the use of an Indeterminate Range will modestly increase the number of false-negative diagnoses. However, thoughtful, setting-specific selection of the criteria used to define indeterminate results can provide substantial economic savings while minimizing harmful health outcomes. For example, if indeterminate results are used to prioritize infants for confirmatory testing, such a policy can prevent 82 infants from receiving unnecessary lifelong ART for every additional death due to missed HIV. These savings, if effectively repurposed, could result in even greater health gains for infants and other people living with HIV.22 This analysis was presented at a WHO Expert Group meeting; the WHO subsequently released recommendations that NAAT results with an indeterminate classification (at a Ct value of 33 or greater) should be repeat-tested using the same sample before result dispatch.23

It is important to contextualize our results given the WHO recommendations. Our analysis simulates confirmatory testing (of a second, independent sample), whereas WHO recommends repeat-testing (of the same sample). While WHO recommends repeat-testing before dispatch from the lab, thus avoiding the need for clinicians to interpret indeterminate results, our analysis assumes that clinical decisions are informed by Ct values (including indeterminate results) while awaiting confirmatory testing. We did not evaluate repeat testing in this analysis, as there are currently no available data on the distributions of repeat test results to inform our model. If repeat testing generates similar results to confirmatory testing, then our results will be conservatively biased (favoring SoC), as we incorporated a time delay between initial and confirmatory testing that should not occur with repeat testing. However, if repeat tests are more closely correlated with initial results than confirmatory tests, the benefits of performing repeat testing after an indeterminate result may be less than projected here. Future empirical work can help quantify this correlation, enabling more robust modeling of the consequences of repeat testing.

We also provide data on the clinical consequences of adopting Ct cutoffs other than the currently recommended ≥33. These results suggest that higher (more limited) Ct cutoffs (eg, ≥35 or ≥36) classify fewer infants as indeterminate and thus minimize the number of infants misclassified as false-negative—but at the expense of relatively smaller gains in false-positive diagnoses averted. By contrast, lower (more inclusive) Ct cutoffs (eg, ≥30 or ≥31) maximize the number of infants who are appropriately diagnosed as HIV-negative, but accordingly misclassify more infants who are truly HIV-infected as indeterminate. For programs where confirmatory testing is not yet universal and indeterminate results are provided to clinicians, our results can inform the clinical tradeoffs of initiating ART for infants with results that fall below various Ct cutoffs. As HIV prevalence declines and confirmatory testing becomes more widely available in some settings, these results can inform future WHO guidance as epidemiological conditions change.

This analysis also helps to operationalize the use of semi-quantitative Ct results of HIV testing in infants, the value of which has been suggested in earlier work. A recent systematic review of 32 studies across 8 countries found that 16.5% of nonnegative PCR results were indeterminate (Ct ≥33) (see Luo et al, submitted with this manuscript). The South African National Health Laboratory Service (NHLS) has similarly documented that indeterminate results have accounted for an increasing fraction of all nonnegative test results since 2010,24,25 and NHLS has provided formal clinical guidelines for the management of indeterminate results since 2015,26 recommending 1–3 additional confirmatory NAAT and/or viral load tests for infants receiving an indeterminate result. Our results illustrate the potential utility of Indeterminate Range policies in settings where confirmatory testing is not routinely available (or available only after long delays).

The projections presented here depend importantly on estimates for several key parameters. The most influential determinant of the tradeoffs of an Indeterminate Range policy was the prevalence of HIV among HIV-exposed infants—a value that will vary greatly across settings. In addition, our results were dependent on the proportion of nonnegative results among HIV-uninfected infants falling in the Indeterminate Range. Our estimates for this parameter were informed by available data (see Luo et al). Should these values vary substantially across populations or diagnostic instruments, the tradeoff values at specific Ct cutoffs will shift accordingly. The incremental cost-effectiveness of adopting an Indeterminate Range is strongly driven by the cost of HIV care for false-positive patients. If these costs are lower than projected—for example, if treatment costs fall (see, Table S1, Supplemental Digital Content, https://links.lww.com/QAI/B371) or retention in care is poor (see Figure S7, Supplemental Digital Content, https://links.lww.com/QAI/B371)—then adopting an Indeterminate Range may be less cost-effective. Because such future costs are difficult to estimate with precision, our primary analysis focuses instead on the numbers of infants affected.

Necessarily, this model simplifies many complexities of EID testing. For example, our model does not consider transmission after initial testing; although such transmission is expected to be similar under SoC and an Indeterminate Range, the SoC will result in more falsely-positive infants. These infants may receive (unintended) preventive benefits from ART, which may improve the cost-effectiveness of the SoC; however, we also omit social effects (eg, stigmatization of false-positive diagnoses) unfavorable toward the SoC. Our model assumes that all surviving, false-negative HIV-infected infants begin ART by age 2; although this is optimistic, we also incorporate high counterbalancing 2-year mortality risks (up to 57%) for untreated HIV. Finally, we simulated a hypothetical sub-Saharan setting to illustrate the utility of an Indeterminate Range algorithm where confirmatory testing is not widely available. This abstracted context will not represent any specific EID setting (especially those outside of sub-Saharan Africa),27,28 although we evaluate a variety of scenarios of transmission and confirmatory testing in our sensitivity analyses.

In conclusion, this decision model illustrates the tradeoffs associated with adopting an Indeterminate Range for early infant HIV diagnosis, showing how such an algorithm can help to derive maximum benefit from incomplete diagnostic information. We demonstrate that, although optimal clinical outcomes are obtained by ensuring confirmatory testing for all HIV-exposed infants with nonnegative results (an approached previously projected to be cost-saving in South Africa),19 prioritizing infants with indeterminate results (high Ct values) for confirmatory tests is likely to improve diagnosis and save costs when confirmatory testing is not universally conducted. Importantly, the desirability of such a policy may depend on several key factors including the prevalence of HIV, the economic and human costs of unnecessary ART, and the willingness to pay of individual programs and stakeholders. These findings align with current WHO recommendations to perform repeat-testing on all samples with an indeterminate value (at a Ct threshold of 33 or greater) and may inform future WHO guidance as the epidemiology and management of early infant HIV diagnosis continue to evolve.

REFERENCES

1. Gumede-moyo S, Filteau S, Munthali T, et al. Implementation effectiveness of revised (post-2010) World Health Organization guidelines on prevention of mother-to-child transmission of HIV using routinely collected data in sub-Saharan Africa: a systematic literature review. Medicine (Baltimore). 2017;96:e8005.
2. Feucht U, Forsyth B, Kruger M. False-positive HIV DNA PCR testing of infants: implications in a changing epidemic. South Afr Med J. 2012;102:149–152.
3. World Health Organization (WHO). WHO Recommendations on the Diagnosis of HIV Infection in Infants and Children. Geneva, Switzerland; 2010. Available at: http://whqlibdoc.who.int/publications/2010/9789241599085_eng.pdf. Accessed May 15, 2018.
4. World Health Organization (WHO). WHO Reminds National Programmes to Retest All Newly Diagnosed People With HIV. Geneva, Switzerland; 2014. Available at: https://www.who.int/hiv/pub/vct/retest-newly-diagnosed-plhiv-full/en. Accessed May 15, 2018.
5. Flynn D, Johnson C, Sands A, et al. Annex 2: An Analysis of 48 National HIV Testing and Counselling Policies. Consolidated Guidelines on HIV Testing Services. 2015. Available at: http://apps.who.int/iris/bitstream/10665/180208/1/WHO_HIV_2015.19_eng.pdf. Accessed May 15, 2018.
6. National AIDS/STD Control Program (Kenya Ministry of Health). Early Infant Diagnosis Testing Trends 2010–2019. 2018. Available at: https://eid.nascop.org/. Accessed May 15, 2018.
7. Central Public Health Laboratories (Uganda Ministry of Health). Early Infant Diagnosis Initiation Rate. Uganda Early Infant Diagnosis Dashboard. 2018. Available at: https://edash.cphluganda.org/. Accessed May 15, 2018.
8. Ministry of Health (Malawi). Early Infant Diagnosis Monthly Tests Summary. Early Infant Diagnosis National Dashboard. Available at: http://www.eidmalawi.org/index2.php?testingsystem=1. Accessed May 15, 2018.
9. Shanks L, Klarkowski D, Brien DPO. False positive HIV diagnoses in resource limited settings: operational lessons learned for HIV Programmes. PLoS One. 2013;8:e59906.
10. Newell M, Coovadia H, Cortina-Borja M, et al. Mortality of infected and uninfected infants born to HIV-infected mothers in Africa: a pooled analysis. Lancet. 2004;364:1236–1243.
11. Maritz J, Preiser W, van Zyl GU. Establishing diagnostic cut-off criteria for the COBAS AmpliPrep/COBAS TaqMan HIV-1 qualitative test through validation against the amplicor DNA test v1.5 for infant diagnosis using dried blood spots. J Clin Virol. 2012;53:106–109.
12. Abbot Molecular Inc. Abbott RealTime HIV-1 Qualitative Test Insert. Des Plaines, IL: 2017. Available at: https://www.molecular.abbott/download-ifu?controlNumbers=51-608381R5. Accessed May 15, 2018.
13. Roche Molecular Systems Inc. COBAS AmpliPrep/COBAS TaqMan HIV-1 Test Package Insert. Branchburg, NJ: 2007. Available at: https://diagnostics.roche.com/hk/en/products/params/cobas-ampliprep-cobas-taqman-hiv-1-qualitative-test-v2-0.html. Accessed May 15, 2018.
14. Sloma CR, Germer JJ, Gerads TM, et al. Comparison of the abbott RealTime human immunodeficiency virus type 1 (HIV-1) assay to the cobas ampliprep/cobas TaqMan HIV-1 test: workflow, reliability, and direct costs. J Clin Microbiol. 2009;47:889–895.
15. Maritz J, van Zyl GU, Preiser W. Irreproducible positive results on the cobas ampliPrep/cobas taqMan HIV-1 qual test are different qualitatively from confirmed positive results. J Med Virol. 2014;86:82–87.
16. Haeri Mazanderani A, du Plessis NM, Thomas WN, et al. Loss of detectability and indeterminate results: challenges facing HIV infant diagnosis in South Africa's expanding ART programme. South Afr Med J. 2014;104:574–577.
17. Haeri Mazanderani A, Moyo F, Kufa T, et al. Differentiating clearly positive from indeterminate results: a review of irreproducible HIV-1 PCR positive samples from South Africa's early infant diagnosis program, 2010–2015. Diagn Microbiol Infect Dis. 2018;91:248–255.
18. Jani IV, Meggi B, Loquiha O, et al. Effect of point-of-care early infant diagnosis on antiretroviral therapy initiation and retention of patients. AIDS. 2018;32:1453–1463.
19. Dunning L, Francke JA, Mallampati D, et al. The value of confirmatory testing in early infant HIV diagnosis programmes in South Africa: a cost-effectiveness analysis. PLoS Med. 2017;14:e1002446.
20. The World Bank. Life expectancy at birth, total (years). 2017. Available at: https://data.worldbank.org/indicator/SP.DYN.LE00.IN. Accessed May 15, 2018.
21. Kimaro GD, Mfinanga S, Simms V, et al. The costs of providing antiretroviral therapy services to HIV-infected individuals presenting with advanced HIV disease at public health centres in Dar es Salaam, Tanzania: findings from a randomised trial evaluating different health care strategies. PLoS One. 2017;12:e0171917.
22. Kahn JG, Bollinger LA, Stover J, et al. Improving the efficiency of the HIV/AIDS policy response: a guide to resource allocation modeling. In: Major Infectious Diseases. Washington, DC: The International Bank for Reconstruction and Development/The World Bank; 2017. Available at: https://www.ncbi.nlm.nih.gov/books/NBK525179/.
23. World Health Organization (WHO). Updated Recommendations on First-Line and Second-Line Antiretroviral Regimens and Post-Exposure Prophylaxis and Recommendations on Early Infant Diagnosis of HIV. 2018. Available at: https://apps.who.int/iris/bitstream/handle/10665/277395/WHO-CDS-HIV-18.51-eng.pdf. Accessed February 1, 2019.
24. Haeri Mazanderani A, Viro FSA, Moyo F, et al. Declining baseline viremia and escalating discordant HIV-1 confirmatory results within South Africa's early infant diagnosis program, 2010–2016. J Acquir Immune Defic Syndr. 2018;77:212–216.
25. Maritz J, Maharaj JN, Cotton MC, et al. Interpretation of indeterminate HIV-1 PCR results are influenced by changing vertical transmission prevention regimens. J Clin Virol. 2017;95:86–89.
26. Haeri Mazanderani A, Technau KG, Hsiao NY, et al. Recommendations for the management of indeterminate HIV PCR results within South Africa's early infant diagnosis programme. South Afr J HIV Med. 2016;17:451.
27. Penazzato M, Revill P, Prendergast AJ, et al. Early infant diagnosis of HIV infection in low-income and middle-income countries: does one size fit all? Lancet Infect Dis. 2014;14:650–655.
28. Audureau E, Kahn JG, Besson M, et al. Scaling up prevention of mother-to-child HIV transmission programs in sub-Saharan African countries: a multilevel assessment of determinants of performance. BMC Public Health. 2013;13:286.
29. Rahima Moosa Mother, Child Hospital. LEOPARD Birth Testing Statistics, Johannesburg June 2014 to December 2017. Johannesburg, South Africa: WHO; 2018.
    30. Rutstein SE, Kamwendo D, Lugali L, et al. Measures of viral load using Abbott RealTime HIV-1 Assay on venous and fingerstick dried blood spots from provider-collected specimens in Malawian district hospitals. J Clin Virol. 2014;60:392–398.
    31. Marston M, Becquet R, Zaba B, et al. Net survival of perinatally and postnatally HIV-infected children: a pooled analysis of individual data from sub-Saharan Africa. Int J Epidemiol. 2011;40:385–396.
    32. Ciaranello AL, Chang Y, Margulis AV, et al. Effectiveness of pediatric antiretroviral therapy in resource-limited settings: a systematic review and meta-analysis. Clin Infect Dis. 2009;49:1915–1927.
    33. Peacock-Villada E, Richardson BA, John-Stewart GC. Post-HAART outcomes in pediatric populations: comparison of resource-limited and developed countries. Pediatrics. 2011;127:e423–e441.
    34. UNICEF: Levels and Trends in Child Mortality Report. 2017. Available at: http://www.unicef.org/publications/index_101071.html. Accessed April 6, 2018.
      35. GBD 2017 DALYs and HALE Collaborators. Global, regional, and national disability-adjusted life years (DALYs) for 359 diseases and injuries and healthy life expectancy (HALE) for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392:1859–1922.
      Keywords:

      early diagnosis; clinical decision-making; cost effectiveness; mathematical model; indeterminate result; false-positive

      Supplemental Digital Content

      Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc.