On-site HIV testing in resource-poor settings: is one rapid test enough?

Objective: To determine the feasibility, accuracy and cost-effectiveness of a rapid, on-site, HIV testing strategy in a rural hospital, and to assess its impact on test turnaround time and the proportion of patients post-test counselled.

Design: Prospective comparison of two testing strategies [double rapid test on-site versus central enzyme-linked immunosorbent assay (ELISA)-based testing], and an economic evaluation.

Setting: Hlabisa Hospital, a rural South African district hospital.

Patients: A total of 454 consecutive adult inpatients requiring and consenting to HIV testing as part of their clinical management.

Main outcome measures: Concordance between rapid tests, and between the rapid and ELISA strategies, test turnaround time, proportion of patients post-test counselled, and cost-effectiveness.

Results: HIV seroprevalence was 49.6%. Both rapid tests were concordant in all patients [one-sided 95% confidence interval (CI) of probability, 99.3-100]. The rapid strategy was 100% sensitive (95% CI, 97.9-100) and 99.6% specific (95% CI, 97.2-100) compared with the ELISA strategy. The mean interval between ordering a test and post-test counselling fell from 21 days prior to the introduction of the rapid test strategy to 4.6 days after its introduction (P < 0.00001). The proportion of patients post-test counselled increased to 96% from 17% after the introduction of the rapid test strategy (P< 0.00001). By using a double rapid test strategy the cost per patient post-test counselled was almost halved to US $11. Accuracy of the rapid strategy was not substantially increased by performing two tests.

Conclusion: In high prevalence, resource-poor settings, rapid, on-site HIV testing is feasible, accurate and highly cost-effective, substantially increasing the number of patients post-test counselled. A single rapid test may be sufficient.

Introduction

Testing for antibodies to HIV in resource-poor settings is important in providing a safe blood supply, as an aid in the diagnosis and management of HIV-related diseases, and for primary prevention of HIV infection [¹].

Sensitivity and specificity of modern tests approach 100% and testing in the most rudimentary settings is possible [²].

To be effective as well as being accurate, results must be available within as short a period of time as possible.

In South Africa, testing tends to be centralized and rural hospitals often experience long delays in receiving test results. Some patients never receive their results [³] and any benefit of testing is therefore lost. Of 664 patients tested between October 1995 and January 1996 in Hlabisa Hospital only 17% ever received their results and were post-test counselled. Average test turnaround time was 21 days and 9% of the results were never received.

The aim of our study was to determine the feasibility, accuracy and cost-effectiveness of a rapid on-site testing strategy for HIV infection in a rural hospital, and to assess the impact of this strategy on test turnaround time and the proportion of patients post-test counselled.

Methods

Setting

Hlabisa health district in South Africa serves a population of 206 000. Most residents are rural Africans living in widely scattered kraals (homesteads), who depend largely on pension remittances, migrant labour and subsistence farming. A major national road and trading route crosses the district and a large township is sited on this road. Prevalence of HIV infection in women attending antenatal clinics in Hlabisa has increased rapidly from 4.2% in 1992 [⁴] to 14% in 1995 (unpublished data).

Subjects

Adult inpatients who required an HIV test in line with current hospital policy were eligible for this study. Current policy recommends that all patients with suspected HIV-related disease should be offered voluntary HIV testing after pre-test counselling and giving informed consent, to make a full diagnosis (and therefore optimize clinical management), and as part of a comprehensive package of individual voluntary testing and counselling, and group education, designed to maximize the opportunity for individual and community behaviour change.

In total, 570 tests were ordered during the 3-month study period (March to May 1996): 52 patients were not tested either because their status was already known (n = 28), blood was not taken (n = 22), or because the patient died before blood could be taken (n = 2). Of the remaining 518 patients, 64 (12%) declined testing. Thus, 454 adults were admitted to the study, and the first 234 (52%) had a detailed evaluation to determine the impact of the rapid testing strategy on test turnaround time and post-test counselling rates.

Testing strategies and counselling

Two commercially available test kits were used for the rapid, on-site strategy. Serum was tested for HIV antibodies using the Capillus HIV-1/HIV-2 rapid test (Cambridge Biotech, Galway, Ireland) and the Abbott Test Pack HIV-1/HIV-2 (Abbott Laboratories, Delkenheim, Germany). The Capillus test is a latex agglutination assay using recombinant proteins representing the major antigens from the envelope proteins of HIV-1 (p24, gp41) and HIV-2 (gp41). The Abbott Test Pack test is based on the enzyme-linked immunosorbent assay (ELISA) principle and uses recombinant peptides representing HIV-1 core and envelope antigens (gp120, gp41, p24) and HIV-2 envelope antigens (gp120, gp34). This two-test strategy complies with World Health Organization (WHO) recommendations for rapid HIV testing [¹]. Specimens were concurrently tested with both kits by two medical technologists as part of routine work and no special training was provided. Remaining serum was stored at 4°C and transported to the Department of Virology of the University of Natal or routine HIV antibody ELISA testing according to current laboratory practice, which conforms to WHO HIV testing strategy 3 [¹]. Specimens negative on the first ELISA in Durban (Abbott Axsym: HIV-1 p24 and p41, HIV-2 p36) were reported as negative and not tested further. Specimens positive on the first ELISA were tested using a second ELISA with a different antigen combination [Ominimed UNIFORM II (HIV-1 p24 and gp160, HIV-2 Env peptide) (Organon, Boxtel, The Netherlands)]. Specimens were reported as positive if both tests were positive; those testing positive on the first ELISA and negative on the second were subjected to a confirmatory antigen test (Abbott Antigen HIV-1 p24: HIV-1 gp41, HIV-2 gp36).

Patients were counselled as most likely to be HIV-positive (blood sent for confirmation) if both rapid tests were positive, and as HIV-negative if both were negative. Patients were not charged for this service.

The study was approved by the Ethics Committee of the Faculty of Medicine, University of Natal, Durban.

Costing

The commercial cost of the rapid kits was obtained from the manufacturers, and the cost of both laboratory staff and all consumables and equipment used was calculated from existing data. For the ELISA strategy an average cost per patient tested was quoted by the Department of Virology, irrespective of the number of ELISA and confirmatory tests required. No attempt was made to cost buildings or overhead costs, because these should not differ much between the two testing sites and they represent a tiny fraction of the costs of either strategy. In addition to the cost per test and cost per patient, the cost-effectiveness was defined as the cost per patient post-test counselled.

Analysis

The aim was to test the equivalence of the rapid test strategy with the ELISA test strategy. Sample size depends on the sensitivity of the ELISA, the prevalence of HIV in the patient population tested, and the degree of confidence required to conclude equivalence between the two strategies. Significance was set at 5% and power at 80%, and the test was a one-sided hypothesis test for equivalence. An ELISA sensitivity and specificity of 99% was assumed when calculating sample size. We estimated that one-half of the patients tested would be HIV-positive. With a sample size of 400 and an HIV prevalence of 46%, equivalence can be assumed if the sensitivity and specificity of the rapid test strategy is greater than 98.4%.

The primary evidence of equivalence in sensitivity and specificity is the probability of discordance between the two strategies [⁵]. Discordance is defined as the probability that the two strategies will give different results in testing groups of HIV-positive and HIV-negative individuals. The smaller the probability, the more likely the sensitivities and specificities will be equivalent. When a one-sided 95% confidence interval (CI) of the probability of discordance is less than a specified value d, the null hypothesis of equivalence of the two strategies will be accepted. This approach is used when the probability of discordance is small. The one-sided 95% CI was calculated from the exact binomial distribution.

The null hypothesis tested was H₀:

Equation (Uncited)

The alternate hypothesis was H_a:

Equation (Uncited)

where d is the stated acceptable difference between the two strategies, S₀ is the sensitivity (specificity) of the ELISA strategy and S_n is the sensitivity (specificity) of the rapid strategy. Proportions were compared using the χ² test, and means by the Students t test.

Categorical data modelling was used to estimate the standard errors of the conditional error rates of using one or two rapid tests. For both the conditional approaches - false-positive and false-negative rate, and misdiagnosis rate (given true HIV status) - the problem was the number of cells with zero counts in the cross tabulations (see Results). To make the analysis tractable, a small positive value (0.01) was added to cells with zero frequencies. The standard errors estimated in the categorical models were then used to calculate the upper limit of one-sided 95% CI.

Results

Serology

Of 454 patients tested, 225 (49.6%) were confirmed HIV-positive by the ELISA-based strategy. Both rapid kits used on-site gave concordant results in all patients (one-sided 95% CI, 99.34-100%). Concordance between the rapid test strategy and the ELISA strategy occurred in all but one case. An elderly man tested HIV-positive by both rapid tests, but was negative on ELISA testing; when he was retested 7 days later, both rapid tests and the repeat ELISA were negative. Sensitivity, specificity, and positive and negative predictive values for the rapid test strategy were close to 100% (Table 1).

Table 1

Testing equivalence

The sensitivity of the ELISA strategy was assumed to be 100%. There were no discordant pairs among the 225 HIV-positive patients, and the one-sided 95% CI for the probability of discordance was 0-1.32. The null hypothesis of equivalence is therefore accepted if the stated acceptable difference is 1.5%. The specificity of the ELISA strategy was also assumed to be 100%. There was one discordant result in the 229 HIV-negative patients, and the one-sided 95% CI for the probability of discordance was 0-2.07. The null hypothesis of equivalence is therefore accepted if the stated acceptable difference is 2.5%.

Patient characteristics

Average age was 34 years (SD, 13 years), 45% were aged 15-29 years, and 53% were women. Most patients (66%) were WHO clinical stage 3 [⁶], and many had confirmed or suspected pulmonary tuberculosis.

Test turnaround time

The mean interval between a rapid HIV test being ordered and post-test counselling being performed was 4.6 days. This compares with a mean interval of 21 days prior to the introduction of the rapid strategy (P < 0.00001).

Post-test counselling

The results of rapid tests performed on all 454 patients were available to the health worker ordering the test, compared with 91% made available prior to the use of the rapid test strategy (P < 0.0001). In all, 224 (96%) patients were post-test counselled at least once. Of the 10 patients not counselled, two died and eight were discharged before they could be counselled. This compares with only 17% of patients post-test counselled prior to the use of the rapid test strategy (P < 0.0001). The mean number of post-test counselling sessions per patient was 2.9, and 80% had more than one post-test counselling session. Prior to the use of the rapid strategy no patients were post-test counselled more than once.

Costs and cost-effectiveness

The cost and cost-effectiveness of each test and the testing strategies are compared in Table 2. Although the double rapid test strategy is expensive, it was highly cost-effective because its use on-site considerably increased the proportion of patients post-test counselled.

Table 2

Is one rapid test enough?

In Table 3 the accuracy of a single rapid test is compared with that of a double rapid test strategy, with the ELISA strategy assumed to be 100% accurate. There are two measures of accuracy that are important. The first is the error rate of the test (the rate of false-positive and false-negative results). The best estimate of the true false-negative rate is zero, but our results are compatible with a false-negative rate of 1 in 4320 if two rapid tests are used. This increases to 1 in 1243 if one test is used. The best estimate of the true false-positive rate is 1 in 229, but our results are compatible with a false-positive rate of 1 in 86 using two tests and 1 in 84 if using one test.

Table 3

The second measure of accuracy is the misdiagnosis rate. For truly HIV-positive persons our best estimate of the rate of misdiagnosing as HIV-negative is zero, but our results are compatible with a misdiagnosis rate of 1 in 1999 if two tests are used, and 1 in 1501 if one test is used. For truly HIV-negative persons our best estimate of the rate of misdiagnosing as HIV-positive is 1 in 229, but our results are compatible with a misdiagnosis rate as HIV-positive of 1 in 87 using two tests and 1 in 54 using one test.

Discussion

The introduction of a rapid, on-site, HIV testing strategy significantly reduced the time for a test result to become available (4.6 versus 21 days), and substantially increased the proportion of patients post-test counselled (96 versus 17%). Despite being more expensive, the rapid test strategy proved to be highly cost-effective [R 45.20 (US $11) versus R 83.80 (US $21) per patient post-test counselled].

Both rapid kits were easy to use in a busy, unsophisticated laboratory, were popular with laboratory staff and were concordant in all cases. In only one case was there discordance with the ELISA-based strategy. Due to the very high sensitivity and specificity of both testing strategies it was difficult to estimate precisely the accuracy of the rapid strategy (Table 3). However, accuracy was not significantly improved by performing two rapid tests instead of one. On-site testing is therefore cheaper - and more cost-effective - if one test is used.

In many resource-poor settings in Africa, HIV tests are provided by foreign donors and cost dictates that only one test is available per patient, despite recommendations to the contrary [¹]. Effectiveness is much improved by on-site testing, with a seemingly small reduction in overall accuracy. It therefore seems reasonable to conclude that in most resource-poor settings a single rapid test is satisfactory. If under these circumstances the result is negative, it would seem reasonable to counsel the patient as HIV-negative. If the result is positive the counselling strategy will depend on the availability of a confirmatory test in a specialized laboratory. There seems to be little benefit in performing a second on-site rapid test. We have adopted the strategy of counselling as HIV-positive and offering a confirmatory test. However, to be at all useful the results of such confirmatory tests should be available very quickly. Although rapid tests are simple and robust, staff do require appropriate training with updates, and laboratory quality control is important.

In conclusion, this study, conducted in a high prevalence setting, has shown that rapid on-site HIV testing is feasible, accurate and highly cost-effective. There seems to be little to be gained by performing two rapid tests instead of one. Such testing strategies require careful evaluation and auditing, and should form an integral part of a broader education, counselling and care service if they are to have maximal impact.

Acknowledgements

We thank D. Mthembu and R. Mathema for performing the rapid tests, and C. Connolly for data management; and Ridge Diagnostics, Johannesburg, South Africa for supplying the Capillus test kits.