Abstract
We evaluated the University of North Carolina-Malawi Risk Screening Score (UMRSS) for detection of acute and early HIV-1 infection (AEHI) in a cohort of Kenyan MSM with approximately 8% annual HIV-1 incidence. Three components of the UMRSS (fever, diarrhea, and discordant rapid HIV tests) were also independent predictors of AEHI in our cohort. The predictive ability (area under the receiver operating characteristic curve, AUC) of the UMRSS was 0.79. A cohort-derived risk score consisting of six characteristics (fever, diarrhea, discordant rapid HIV tests, fatigue, age <30 years, and symptomatic sexually transmitted disease) had a higher AUC of 0.85. Screening for AEHI will have substantial transmission prevention benefits.
MSM have among the highest HIV-1 incidence in sub-Saharan Africa, but targeted interventions for HIV-1 testing of MSM are mostly lacking [1–3]. When HIV-1 is acquired, patients may frequently seek urgent healthcare for symptoms, including fever or unconfirmed ‘malaria’ [4–6], and become extremely infectious during a short period of 3–4 weeks (acute HIV infection; AHI) and highly contagious during the first 6 months (early HIV-1 infection) [7,8]. However, diagnosing acute and early HIV-1 infection (AEHI) remains challenging in resource-limited settings, in part due to a lack of low-cost, point-of-care tests for nucleic acid detection [9]. In Malawi, the UNC Malawi Risk Screening Score (UMRSS) combining discrete clinical and behavioral characteristics has been developed to identify AHI among sexually transmitted diseases (STDs) clinic patients [10]. In this study, STD patients who were HIV-1 negative or had discordant rapid HIV tests received a score of 1 for fever, body ache, and more than one partner; 2 for diarrhea and genital ulcer disease (GUD); and 4 for discordant rapid tests. Using this algorithm, Powers et al.[10] could identify 95% of the AHI cases identified by targeted testing of only patients with a score of 2 or greater (40% of the population studied). As MSM frequently present with an STD, we wanted to validate this UMRSS in our MSM cohort in Coastal Kenya, and compare it to a cohort-derived risk screening score (CDRSS) using our own data and similar methodology.
Since 2005, we have enrolled HIV-seronegative MSM in a cohort study of HIV-1 acquisition, as previously described [1]. Men made either monthly or quarterly scheduled visits at which risk reduction counseling was provided and a medical history and physical examination was performed. HIV-1 seroconversion was diagnosed using two rapid test kits (Determine, Abbott Laboratories, Abbott Park, Illinois, USA; Unigold, Trinity Biotech plc, Bray, Ireland) in parallel. Patients with discordant rapid HIV-1 test results were retested until discordancy was resolved. All seronegative and discordant samples were tested for p24 antigen (Vironostika HIV-1 p24 ELISA, Biomérieux, Ltd, Marcy l’Etoile, France). Up to 1 January 2012, preseroconversion and postseroconversion plasma samples were tested for HIV-1 RNA level (Amplicor Monitor, version 1.5; Roche, Branchburg, New Jersey, USA) with a positive result defined as more than 400 copies/ml [1]. All HIV diagnoses were confirmed by a positive RNA level (n = 67) or by follow-up until both rapid tests were positive (n = 6).
An AEHI visit was defined as a visit with an antibody seroconversion (determined by two positive rapid HIV tests), serodiscordant rapid HIV tests (one rapid test positive, one rapid test negative), or positive p24 antigen test. Eighteen (90%) of 20 patients with a positive p24 antigen test were clinically evaluated within 1 week of the test result [1].
HIV-1 incidence was estimated at 7.5% (95% confidence interval 6.0–9.5) per 100 person years during the study period. The median number of days from estimated date of HIV-1 infection to evaluation was 39 (interquartile range 19–59). At their AEHI visit, 42 patients had a rapid antibody seroconversion (after a documented seronegative result at the last study visit), 11 had a serodiscordant rapid HIV test, and 20 had a positive p24 antigen test. Characteristics reported (including symptoms experienced since the last study visit) at 73 AEHI visits were compared with characteristics reported at 6458 scheduled cohort visits (Table 1).
Table 1. Comparison of clinical and behavior characteristics in MSM at acute and early HIV-1 infection and seronegative clinic visits, Coastal Kenya, 2005–2012.
| Predictor | Acute or early HIV-1 infection visits (n = 73) n (%) | HIV-1 negative visits (n = 6458) n (%) | Unadjusted PORa (95% CI) | Domain-specific model adjusted POR (95% CI)b | Combined model adjusted POR (95% CI)c | Final model adjusted POR (95% CI)d | Cohort-derived Scoree | UMRSS scoref |
| Sociobehavior and sociodemographic | ||||||||
| >1 Sex partner in the past month | 46 (63.0) | 4677 (72.4) | 0.8 (0.5–1.3) | 0.6 (0.4–1.0) | – | – | – | 1 |
| Age <30 yearsg | 64 (87.7) | 4510 (69.8) | 3.3 (1.6–6.8) | 3.1 (1.5–6.2) | 3.0 (1.3–6.9) | 3.3 (1.5–7.5) | 1 | – |
| Symptoms reported in the past month or since last visit | ||||||||
| Feverh | 47 (64.4) | 1037 (16.1) | 9.5 (5.8–15.7) | 1.2 (0.6–2.5) | 2.3 (1.2–4.6) | 2.8 (1.4–5.5) | 1 | 1 |
| Joint or muscle pain | 47 (64.4) | 1230 (19.1) | 7.2 (4.4–11.9) | 1.4 (0.7–2.8) | – | – | – | 1 |
| Head ache | 45 (61.6) | 1244 (19.3) | 6.7 (4.2–11.1) | 1.4 (0.7–2.8) | – | – | – | – |
| Fatigue | 47 (64.4) | 939 (14.5) | 10.9 (6.6–18.0) | 2.2 (1.1–4.6) | 3.8 (1.8–7.9) | 3.5 (1.8–6.8) | 1 | – |
| Loss of appetite | 39 (53.4) | 700 (10.8) | 9.3 (5.8–15.0) | 1.6 (0.9–2.9) | – | – | – | – |
| Night sweats | 30 (41.1) | 449 (7.0) | 9.7 (5.8 –16.2) | 1.5 (0.8 –2.8) | – | – | – | – |
| Sore throat | 26 (35.6) | 563 (8.7) | 5.9 (3.5–9.8) | 1.0 (0.5–2.0) | – | – | – | – |
| Diarrhea | 26 (35.6) | 411 (6.4) | 8.6 (5.2–14.3) | 2.1 (1.2–3.6) | 2.7 (1.5–4.9) | 2.8 (1.6–4.9) | 1 | 2 |
| Swollen glands | 19 (26.0) | 328 (5.1) | 6.6 (3.7–11.7) | 1.2 (0.5–2.6) | – | – | – | – |
| Vomiting | 19 (26.0) | 290 (4.5) | 6.9 (3.9–12.2) | 1.1 (0.6–2.1) | – | – | – | – |
| Oral ulcers | 14 (19.2) | 240 (3.7) | 6.1 (3.2–11.5) | 0.8 (0.4–1.6) | – | – | – | – |
| Too sick to work | 33 (45.2) | 345 (5.3) | 14.7 (8.9–24.5) | 2.6 (1.4–4.7) | – | – | ||
| Urethritis | 8 (11.0) | 392 (6.1) | 1.8 (0.8–3.9) | – | – | – | – | – |
| Proctitis | 13 (17.8) | 273 (4.2) | 6.5 (3.3–12.6) | – | – | – | – | – |
| Genital sores | 10 (13.7) | 186 (2.9) | 6.8 (3.2–14.2) | – | – | – | – | – |
| Any symptomatic STD | 21 (28.8) | 614 (9.5) | 4.2 (2.4–7.3) | 1.7 (0.9–3.1) | 2.0 (1.1–3.8) | 2.3 (1.3–4.0) | 1 | – |
| Clinical examination at visit | ||||||||
| Maculopapular rash | 4 (6.0) | 100 (1.6) | 4.1 (1.4–11.8) | 2.8 (1.0–7.7) | 1.7 (0.5–6.0) | – | – | – |
| Conjunctivitis | 14 (19.2) | 349 (5.4) | 4.2 (2.2–7.8) | 3.1 (1.7–5.9) | 1.0 (0.4–2.3) | – | – | – |
| Genital ulcer | 1 (1.5) | 40 (0.6) | 2.6 (0.3–21.3) | 1.4 (0.2–10.2) | – | – | – | 2 |
| Genital warts | 0 | 34 (0.6) | – | – | – | – | – | – |
| Lymph nodes | 19 (26.0) | 736 (11.4) | 2.8 (1.6–4.8) | – | – | – | – | – |
| Cervical | 12 (17.7) | 277 (4.3) | 5.0 (2.5–9.8) | 2.3 (1.1–4.8) | 1.3 (0.5–3.0) | – | – | – |
| Axillar | 2 (3.0) | 40 (0.6) | 5.3 (1.2–23.8) | 3.5 (0.8–14.8) | – | – | – | – |
| Inguinal | 15 (22.4) | 556 (8.9) | 3.1 (1.7–5.6) | 1.8 (1.0–3.1) | 1.1 (0.5–2.4) | – | – | – |
| Discordant HIV test results | 11 (15.1) | 14 (0.22) | 73.5 (11.7–462.3) | N/A | 53.5 (11.1–258.1) | 45.9 (9.2–229.7) | 4 | 4 |
AEHI, acute and early HIV-1 infection; 95% CI, 95% confidence interval; GEE, generalized estimating equation; POR, prevalence odds ratio; STD, sexually transmitted disease; UMRSS, University of North Carolina-Malawi Risk Screening Score.
aPrevalence odds ratio.
bFactors associated with AEHI at P ≤ 0.05 were included in initial multivariable models for two domains: ‘symptom’ and ‘clinical exam’ findings.
cFactors associated with AEHI at P ≤ 0.05 in the initial domain-specific models were included in a combined model, to which age and discordant HIV test results were added.
dAll variables in final model associated with AEHI at P ≤ 0.05.
ePredictor score is equal to its β coefficient (natural log of the adjusted prevalence odds ratio) from the GEE model, rounded to the nearest integer.
fUMRSS derived from multivariable analysis of predictors of acute HIV infection [10].
gAge <30 years. Younger age in MSM was associated with HIV-1 acquisition in unadjusted multivariable analysis in our cohort [1].
hIncluded patients who had a history of presumptive ‘malaria’ treatment.
Cohort-derived risk screening score
To identify predictors of AEHI in our MSM cohort, we calculated unadjusted prevalence odds ratios for sociodemographic, medical history, and physical examination findings with AEHI as an outcome. We compared characteristics reported at AEHI visits to those reported at all seronegative visits, using generalized estimating equation (GEE) to adjust for intraindividual correlation. We constructed separate models for two specific domains (i.e. symptoms reported in past months and clinical examination findings), similar to the approach of Powers et al.[10]. Characteristics associated with AEHI at P ≤ 0.05 were included in initial multivariable models for two domains: ‘symptom’ and ‘clinical examination’ findings. We constructed a full, combined model including discordant rapid test results, age, fever, and the variables from the reduced, domain-specific models. Fever (or a history of having received treatment for unconfirmed ‘malaria’) was included a priori as this was the most significant reason for care seeking prior to seroconversion in our cohort [4]. The final model retained only predictors associated with AEHI at P ≤ 0.05 in the combined model. Data for men who seroconverted were censored at the AEHI visit.
Comparison of cohort-derived risk screening score with University of North Carolina-Malawi Risk Screening Score
Similar to Powers’ approach, we assigned each variable in the final cohort-derived model a predictor score equal to its β coefficient (natural log of the adjusted prevalence odds ratio) from the GEE model, rounded to the nearest integer (Table 1). Independent predictors for HIV-1 acquisition included in the CDRSS and their corresponding predictor scores were 1 for fever, fatigue, any symptomatic STD, diarrhea or age less than 30 years; and 4 for discordant HIV tests. The maximum score possible using the CDRSS was 9. Predictors in the UMRSS are shown in the last column of the Table 1 and can sum to 11 [10]. We selected the CDRSS cut-off point that optimized sensitivity, specificity, and positive predictive values (details in the supplemental figure) and compared our results with UMRSS results. We calculated the AUC for the predictive ability of each score to identify a patient visit at which AEHI was diagnosed.
The UMRSS with a cut-off point at 2 had a sensitivity of 75.3%, specificity of 76.4%, and positive predictive value of 3.5% to identify AEHI correctly in our study population. Corresponding values for the CDRSS with a cut-off point at 2 were 80.8%, 76.0%, and 3.7%, respectively. The AUCs for the UMRSS and the CDRSS were 0.79 and 0.85, respectively (P < 0.009). When we restricted the CDRSS to include only age less than 30 years, fever, or any symptomatic STD, the AUC became 0.77 (not different from the UMRSS).
To our knowledge, this is the first time that the UMRSS has been validated outside Malawi. Three of the six UMRSS characteristics (history of fever, diarrhea, and discordant rapid HIV tests) identified AEHI independently in our MSM population, and so were included in the CDRSS. The CDRSS had a better performance when all six predictors of AEHI in our cohort were included (additional characteristics: fatigue, age <30 years, and symptomatic STD). Interestingly, a simplified CDRSS including only three characteristics (age <30 years, fever, and any symptomatic STD) had a similar performance to the UMRSS.
Study limitations include possible ascertainment bias if patients overreported symptoms at their AEHI visit due to a seroconversion diagnosis, and the fact that we derived our validation from a very high-risk MSM population followed in a research setting. Although GUD was a predictor of AEHI in the study by Powers et al. [10], GUD was uncommon in this MSM population [1]. Fatigue was not a predictor of AEHI in STD patients in Malawi, but was frequently reported by our population and by men (but not women) in a large study of primary HIV-1 infection [11].
In summary, our study confirmed the importance of fever, diarrhea, and discordant HIV test results for the identification of AEHI in African populations, and demonstrated that targeted screening for AEHI in MSM could be performed using a CDRSS consisting of a limited set of characteristics, including age younger than 30 years, fever, diarrhea, fatigue, any symptomatic STD, and discordant HIV test results. Such screening for AEHI, when supported by risk reduction counseling and combination prevention therapy, will have substantial transmission prevention benefits [3].
Acknowledgements
We thank Kimberly Powers at the University of North Carolina and an anonymous AIDS reviewer for useful comments on an earlier draft of this paper. We thank staff of the KEMRI-HIV Key Populations Studies Cluster, based at the KEMRI-Wellcome Trust Research Programme (KWTRP) in Kilifi, and the International AIDS Vaccine Initiative for supporting the MSM cohort studies. The KWTRP at the Centre for Geographical Medicine Research-Kilifi is supported by core funding from the Wellcome Trust (#077092).
This study is made possible by the generous support of the American people through the United States Agency for International Development (USAID). The contents are the responsibility of the study authors and do not necessarily reflect the views of USAID, the NIH, or the United States Government. This report was published with the permission of KEMRI.
Author contributions: E.W. conducted data management, performed the analysis, and drafted the manuscript; G.F. performed data analysis and manuscript writing; H.S.O. conducted data management and performed the analysis; P.M. provided support to clinical care and manuscript writing; M.A.P. supported study design, data analysis, and manuscript writing; G.M. conducted clinical data collections; A.T. conducted clinical data collections; S.M.G. supported study design, data analysis, and manuscript writing; and E.J.S. designed the study, conducted data analysis and manuscript writing.
Conflicts of interest
There are no conflicts of interest.
Supplementary Material
Footnotes
Correspondence to Dr Eduard J. Sanders, Kenya Medical Research Institute, Centre for Geographic Medicine Research – Coast, PO Box 230, Kilifi, Kenya. Tel: +254 41 752 2133; fax: +254 41 752 2390; e-mail: ESanders@kemri-wellcome.org
References
- 1.Sanders EJ, Okuku HS, Smith AD, Mwangome M, Wahome E, Fegan G, et al. High HIV-1 incidence, correlates of HIV-1 acquisition, and high viral loads following seroconversion among MSM. AIDS 2013; 27:437–446 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Price MA, Rida W, Mwangome M, Mutua G, Middelkoop K, Roux S, et al. Identifying at-risk populations in Kenya and South Africa: HIV incidence in cohorts of men who report sex with men, sex workers, and youth. J Acquir Immune Defic Syndr 2012; 59:185–193 [DOI] [PubMed] [Google Scholar]
- 3.Sullivan PS, Carballo-Dieguez A, Coates T, Goodreau SM, McGowan I, Sanders EJ, et al. Successes and challenges of HIV prevention in men who have sex with men. Lancet 2012; 380:388–399 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sanders EJ, Wahome E, Mwangome M, Thiong’o AN, Okuku HS, Price MA, et al. Most adults seek urgent healthcare when acquiring HIV-1 and are frequently treated for malaria in coastal Kenya. AIDS 2011; 25:1219–1224 [DOI] [PubMed] [Google Scholar]
- 5.Bebell LM, Pilcher CD, Dorsey G, Havlir D, Kamya MR, Busch MP, et al. Acute HIV-1 infection is highly prevalent in Ugandan adults with suspected malaria. AIDS 2010; 24:1945–1952 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Serna-Bolea C, Munoz J, Almeida JM, Nhacolo A, Letang E, Nhampossa T, et al. High prevalence of symptomatic acute HIV infection in an outpatient ward in southern Mozambique: identification and follow-up. AIDS 2010; 24:603–608 [DOI] [PubMed] [Google Scholar]
- 7.Cohen MS, Shaw GM, McMichael AJ, Haynes BF. Acute HIV-1 Infection. N Engl J Med 2011; 364:1943–1954 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Miller WC, Rosenberg NE, Rutstein SE, Powers KA. Role of acute and early HIV infection in the sexual transmission of HIV. Curr Opin HIV AIDS 2010; 5:277–282 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Fiscus SA, Pilcher CD, Miller WC, Powers KA, Hoffman IF, Price M, et al. Rapid, real-time detection of acute HIV infection in patients in Africa. J Infect Dis 2007; 195:416–424 [DOI] [PubMed] [Google Scholar]
- 10.Powers KA, Miller WC, Pilcher CD, Mapanje C, Martinson FE, Fiscus SA, et al. Improved detection of acute HIV-1 infection in sub-Saharan Africa: development of a risk score algorithm. AIDS 2007; 21:2237–2242 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Meditz AL, MaWhinney S, Allshouse A, Feser W, Markowitz M, Little S, et al. Sex, race, and geographic region influence clinical outcomes following primary HIV-1 infection. J Infect Dis 2011; 203:442–451 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.