During the present efforts to roll out antiretroviral treatment programs in sub-Saharan Africa, accurate and efficient monitoring of HIV-infected patients is essential. Eligibility criteria for treatment are usually based on evidence of HIV-related disease and level of immunodeficiency measured by CD4 cell counts. Viral load preferentially assessed as plasma HIV-RNA is employed to monitor treatment efficacy.1 Measurements of plasma HIV-RNA require expensive reagents and equipment not generally available in sub-Saharan Africa. The HIV-1 antigen p24 has been proposed as an alternative marker of progression in several studies.2,3 Usage of p24 measurements was earlier hampered by high detection limits, but optimization has improved the sensitivity.2 p24 has been shown to correlate with HIV-RNA and to contain prognostic value.4 However, only few studies have examined the prognostic capabilities of p24 measurements in a rural African context with a predominance of infections by non-B subtypes. In Senegal, p24 was found to correlate well with HIV-RNA but the sensitivity of the assay to predict low levels of HIV-RNA was limited.5 Conversely, in Côte d'Ivoire, p24 predicted treatment failure similarly to HIV-RNA.6 In the present study, we investigated the prognostic strength of p24 measurements among HIV-infected antiretroviral treatment-naive individuals in a cohort set in rural Zimbabwe.
The study population has been described previously.7 Briefly, 379 adults were recruited for the Mupfure Schistosomiasis and HIV Cohort between October 2001 and November 2002.
p24 Antigen Analyses
Whole blood was drawn into ethylene diamine tetra-acetic acid (EDTA) tubes. Plasma was separated within 4 hours by centrifugation at 3000 rpm for 10 minutes. Plasma was stored in liquid nitrogen until shipment on dry ice; in Copenhagen, samples were stored at −80°C.
The HIV-1 p24 antigen measurements were performed at the AIDS Laboratory, Rigshospitalet, Copenhagen, using the PerkinElmer p24 Ultrasensitive Assay Protocol (PerkinElmer, Boston, MA).
The method was performed according to the manufacturer's instructions, except for the following modifications: The virus disruption buffer, developed at the Swiss National Center for Retroviruses, described by Schüpbach et al2,8 and Jennings et al,9 was used to optimize the assay performance by adding it to the prediluted kit lysis buffer 1:10, and 250 μL of this buffer was mixed with 50 μL plasma. The samples were then lysed for 10 minutes at room temperature. This procedure was modified from the procedure described by Schüpbach et al,2,8 where the virus disruption buffer was mixed with sample material, then the samples were lysed for 10 minutes at room temperature, and subsequently the kit lysis buffer was added. This modification was attributed to simplify the procedure for easy implementation in field setups.
The samples were subsequently heated at 100°C for 5 minutes. After cooling to room temperature, a volume of 250 μL specimen, control, or standard was added to wells coated with mouse monoclonal antibody to HIV-1 p24. The plate was incubated on a whirlmix shaker at 400 rpm for 2 hours at room temperature and washed in an automated plate washer (DROP washer Seac, Radim Company, Firenze, Italy). Each plate carried 1 blank, 3 negative controls, and 9 prediluted standards. Frozen aliquots of negative controls, prediluted standards, and aliquots of 0.5% Triton X-100 containing the virus disruption buffer were used to reduce interassay variation. Furthermore, frozen aliquots of the 2 streptavidin-horseradish peroxidase solutions and the signal amplification solution were used. Absorbance was read as an end point reading on an Anthos 2010 reader at wavelength 492 nm and reference filter 620 nm (Anthos Labtec Instruments, Wals, Austria). The test was performed as a single test. A lower limit of detection (LLD) corresponding to the mean of the 3 negative controls in the present run plus 3 SDs of the negative controls from all runs was used.
CD4 cell counts were obtained by multiplying the CD4 percentage (CD3, CD4, CD45 Tritest antibodies analyzed on FacsCalibur, Becton-Dickinson, San Jose, CA) with the lymphocyte count (Hematology Analyzer SF 3000, Sysmex, Ramsey, MN). Plasma levels of HIV-1 RNA (Roche Amplicor HIV-1 Monitor Test v1.5, F. Hoffmann-La Roche, Basel, Switzerland) were determined according to the manufacturer's instructions.
At inclusion, participants were interviewed, clinically examined, and classified into clinical stages A, B, and C according to the Centers for Disease Control and Prevention (CDC) classification system for HIV infection.10
Subjects were screened for Schistosoma (S). haematobium and Schistosoma mansoni, and infected individuals were randomized to receive a single oral dose of praziquantel, 40 mg/kg, at baseline (early intervention) or 3 months later (delayed intervention).
In December 2005 to January 2006, a survival survey was performed and times of deaths of deceased participants were recorded. For 13 of the 58 deceased participants, only the month of death was known and the date of death was then registered as the 15th of that month.
The Medical Research Council of Zimbabwe and the Central Medical Scientific Ethics Committee of Denmark approved the present study. Informed consent was obtained from all participants. Participants with CD4 cell counts below 200 cells/μL were offered treatment with sulfamethoxazole and trimethoprim. Secondary infections were treated according to Zimbabwean guidelines. Pregnant women were referred to mother-to-child prevention programs and were not included in the study. No antiretroviral treatment programs were available in the region at the time of study.
p24 and HIV-RNA measurements were log transformed to approximate normal distribution. Univariate linear regressions were performed with p24 as the dependent variable and HIV-RNA or CD4 cell count as predictors. Back transformed regression coefficients, hence, representing the fold change in p24 with a 1-unit change in the predictor, were reported. Univariate and multivariate logistic regressions with participants stratified according to CD4 cell count or clinical stage and p24, HIV-RNA, or CD4 cell count as predictors were performed and odds ratios (ORs) reported.
Survival times were compared between p24 strata by log-rank tests. Univariate and multivariate Cox proportional hazard models were performed. Results are shown as hazard ratios. For all analyses, 95% confidence intervals (CIs) were reported.
Baseline characteristics of the cohort have been described previously in detail.7,11 Briefly, the cohort included 198 HIV-infected participants of whom 83% were female. The median age was 31 years among females (range 19-59) and 32 years among males (range 20-54). The median CD4 cell count was 319 cells/μL (interquartile range: 185-503) and the median HIV-RNA was 63,250 copies/mL (interquartile range 12,800-180,000).
Among the 198 HIV-infected participants, 90% had quantifiable p24 antigen at baseline while measurements were below the LLD in 20 (10%) of the measurements. Two (1%) HIV-RNA measurements were below the LLD. The levels of HIV-RNA corresponding to the p24 measurements below the LLD had a median of 4000 copies/mL (interquartile range: 861-12,075 copies/mL). However, the median CD4 count among patients with p24 below the LLD was 647 cells/μL (interquartile range: 394-849), and only one of these had a CD4 cell count below 200 cells/μL. The samples with undetectable p24 antigen or HIV-RNA were assigned the value of the LLD. As measurements of the negative controls varied between runs, the LLD also varied between runs. The conclusions of all the following tests were not changed when the standard curves were extended below the LLD, and actual optical densities for measurements below the LLD were used to calculate the p24 values.
A univariate regression analysis was performed with plasma p24 antigen values as the dependent parameter and HIV-RNA as the predictor. p24 correlated with HIV-RNA (Fig. 1). When the analysis was repeated without measurements below the LLD, similar results were obtained (Fig. 1).
p24 also correlated with CD4 cell count (R2: 0.23, P < 0.0001, mean fold change: 1.4 with CD4 cell count 100 cells/μL lower, CI: 1.3 to 1.5). A slightly stronger correlation between HIV-RNA and CD4 cell count was found (R2: 0.24, P < 0.0001, mean fold change: 1.6 with CD4 cell count 100 cells/μL lower, CI: 1.5 to 1.8).
Participants were categorized into 2 strata according to baseline CD4 cell counts below or above 200 cells/μL. p24 predicted CD4 stratum as evaluated by logistic regression (P < 0.0001, OR: 2.8 with p24 10-fold higher, CI: 1.7 to 4.4). Similar results were obtained in a multivariate analysis with age and sex adjustments. Likewise, levels of HIV-RNA predicted CD4 group (P < 0.0001, OR: 3.8 with HIV-RNA 10-fold higher, CI: 2.1 to 6.9). In a multivariate analysis with p24 and HIV-RNA as predictors and adjustments for age and sex, there was effect of HIV-RNA (P = 0.01, OR: 2.6 with HIV-RNA 10-fold higher, CI: 1.3 to 5.5) but not of p24 (P = 0.31, OR: 1.4 with p24 10-fold higher, CI: 0.7 to 2.6). To evaluate the sensitivity and specificity of p24 to predict CD4 count below 200 cells/μL, a receiver operating curve (ROC) for p24 was drawn (Fig. 2). An ROC for HIV-RNA predicting CD4 group was drawn for comparison (Fig. 2). Both were adjusted for age and sex. Areas under the curves were 0.75 and 0.77, respectively.
Logistic regressions were performed with p24 as a predictor for CDC categories at baseline (asymptomatic: A, symptomatic: B or C). p24 antigen predicted symptomatic CDC category (Fig. 3A). To evaluate whether the predictive strength of p24 differed according to high or low CD4 cell count, we performed the analysis stratified on CD4 count. Stratification on the CD4 count median of 319 cells/μL was chosen because only 53 participants had a CD4 count of below 200 cells/μL. p24 predicted CDC category in participants with CD4 cell counts below and above the median (Figs. 3B, C). HIV-RNA also predicted CDC category (P < 0.001, OR, CDC B/C vs A with HIV-RNA 10-fold higher: 2.2, CI: 1.4 to 3.4). When a model including p24 antigen, HIV-RNA, and CD4 cell count was fitted, there were no effects of any of the parameters. In a model including p24 and CD4 cell count, there was an effect of p24 (P < 0.01, OR, CDC B/C vs A with p24 10-fold higher: 2.0, CI: 1.3 to 3.3) but not of CD4 cell count (P = 0.34). Adjustments for age and sex did not alter the results.
p24 measurements were additionally performed on samples from 20 HIV-seronegative individuals from the same cohort. All were below the LLD.
p24 measurements in the 3-month follow-up samples correlated tightly with baseline measurements as expected with no significant intrapersonal change (paired t test: P = 0.88, mean fold change, follow-up vs baseline: 1.01, CI: 0.88 to 1.16). Moreover, the change in p24 between baseline and the 3-month follow-up correlated with the change in HIV-RNA as assessed by univariate linear regression (P < 0.0001, Pearson r = 0.32).
A thorough analysis of the mortality in the cohort has previously been reported.12 Briefly, the total follow-up time was 631 years, and 58 deaths were recorded. Among the 46 deaths with available information, the causes included diarrhea, respiratory illness, wasting, and cancer, and only one participant was reported to have died from a nonnatural cause (suicide). Among 180 HIV-seronegative participants in the Mupfure Schistosomiasis and HIV cohort, only 2 deaths were recorded of which at least one was from nonnatural cause (murdered). The 156 schistosomiasis and HIV-infected individuals were treated with praziquantel at baseline or at the 3-month follow-up, and schistosomiasis infection at baseline or schistosomiasis intensity did not predict mortality.12 In a univariate Cox proportional hazards analysis, p24 predicted mortality (P < 0.0001, hazard ratio, p24 10-fold higher: 2.3, CI: 1.6 to 3.1). In a multivariate model including p24, CD4 cell count, age, and sex, mortality increased with higher p24 and lower CD4 cell count (Table 1). Similarly, in a model with HIV-RNA substituted for p24, mortality increased with HIV-RNA (Table 1). HIV-RNA was stronger than p24 as evaluated by P value (<0.0001 vs <0.05). There was no interaction between CD4 cell count and HIV-RNA or between CD4 count and p24 antigen in any model. The assumption of proportionality was met in all models.
We found p24 to correlate well with plasma HIV-RNA with a correlation coefficient similar to those previously reported.2-5,8,13 Participants in the present study were antiretroviral naive, and the majority had measurable p24 antigen using the optimized assay. Ten percent of the patients had p24 antigen below the LLD at baseline. One percent of the HIV-RNA measurements was rendered undetectable, and p24 measurements below the LLD corresponded to a broad range of HIV-RNA loads, but only one of these patients had a CD4 count of less than 200 cells/μL.
In line with previous reports, we found that p24 antigen contained significant information in the analyses, where patients were stratified according to CD4 count (< or >200 cell/μL) and clinical stages (asymptomatic or symptomatic), and in survival analyses. However, HIV-RNA was stronger than p24 in predicting CD4 cell counts below 200 cells/μL when both p24 and HIV-RNA were included in a multivariate analysis. We evaluated the sensitivity and specificity of both p24 and HIV-RNA as predictors for CD4 cell count below 200 cells/μL. The optimal operational point in the ROC is debatable. However, type 2 errors, where individuals are falsely categorized to have a CD4 count of above 200 cells/μL, could be detrimental to the patient. As an example, a cutoff value for p24 of 50 pg/mL would result in a sensitivity of 81% and a specificity of 60%. Given that the CD4 cell count is the golden standard in determination of when to institute treatment, not only p24 but also HIV-RNA were far from perfect. The efficacy of p24 may be superior as a marker of treatment failure. As this cohort only included antiretroviral-naive volunteers, this would have to be investigated in other studies. This is an obvious limitation to this study. However, an alternative end point in the current study could be to identify the individuals with a higher than 10% risk of death within the first year. An ROC with the p24 cutoff adjusted to a sensitivity of 90% of predicting death within 1 year had a specificity of 46%. The corresponding ROC for HIV-RNA had a specificity of 52%. Hence, with death within 1 year as the end point, HIV-RNA performed slightly but insignificantly better.
In this study, we included a cutoff corresponding to the mean of the 3 negative controls in the present run plus 3 SDs of the negative controls from all runs. Participants in this study were known to be infected with HIV, and inclusion of participants above or below the LLD did not change results. Elimination of negative controls would not impact results and only save space for another 3 patient samples. However, if the test was not only to be used to identify participants in need of treatment but also treatment failure, then it might be relevant with a precise definition of the LLD. The interassay variation in the current definition of the LLD would be problematic if used to ascertain treatment failure. Different measures, for example, dilution of standards in negative plasma instead of PBS, could be applied to optimize the assay and it is hoped that it would minimize variations from assay to assay and intra-assay variation.
p24 was surprisingly superior to CD4 cell counts at predicting symptomatic CDC category. p24 predicted mortality in univariate analysis. However, it was inferior to HIV-RNA and CD4 counts in multivariate analysis.
The p24 antigen assay has previously been evaluated in few studies including patients with subtype C infections. The assay proved to be a useful diagnostic test for vertical HIV transmission among infants infected with subtype C HIV-1.14 Another study investigated the changes in p24 antigen levels among patients infected with subtype C HIV receiving antiretroviral treatment relative to the changes in HIV-RNA and found a strong correlation between the changes in the 2 markers,15 suggesting that p24 can be used to monitor therapy.
To our knowledge, this study is the first investigating the prognostic strengths of p24 in a region with predominance of HIV-1 subtype C. p24 correlated with HIV-RNA and was stronger than CD4 cell count at predicting CDC symptomatic stages. p24 predicted mortality, although with less predictive strength than HIV-RNA and CD4 cell count.
We thank the Mupfure Community for their willing participation; Mupfure Secondary School for accommodation; and the technical team for their tireless hard work.
2. Schüpbach J, Tomasik Z, Knuchel M, et al. Optimized virus disruption improves detection of HIV-1 p24 in particles and uncovers a p24 reactivity in patients with undetectable HIV-1 RNA under long-term HAART. J Med Virol
3. Boni J, Opravil M, Tomasik Z, et al. Simple monitoring of antiretroviral therapy with a signal-amplification-boosted HIV-1 p24 antigen assay with heat-denatured plasma. AIDS
4. Ledergerber B, Flepp M, Boni J, et al. Human immunodeficiency virus type 1 p24 concentration measured by boosted ELISA of heat-denatured plasma correlates with decline in CD4 cells, progression to AIDS, and survival: comparison with viral RNA measurement. J Infect Dis
5. Ondoa P, Dieye TN, Vereecken C, et al. Evaluation of HIV-1 p24 antigenemia and level of CD8+ CD38+ T cells as surrogate markers of HIV-1 RNA viral load in HIV-1-infected patients in Dakar, Senegal. J Acquir Immune Defic Syndr
6. Tehe A, Maurice C, Hanson DL, et al. Quantification of HIV-1 p24 by a highly improved ELISA: an alternative to HIV-1 RNA based treatment monitoring in patients from Abidjan, Cote d'Ivoire. J Clin Virol
7. Kallestrup P, Zinyama R, Gomo E, et al. Schistosomiasis and HIV-1 infection in rural Zimbabwe: implications of coinfection for excretion of eggs. J Infect Dis
8. Schüpbach J, Boni J, Bisset LR, et al. HIV-1 p24 antigen is a significant inverse correlate of CD4 T-cell change in patients with suppressed viremia under long-term antiretroviral therapy. J Acquir Immune Defic Syndr
9. Jennings C, Fiscus SA, Crowe SM, et al. Comparison of two human immunodeficiency virus (HIV) RNA surrogate assays to the standard HIV RNA assay. J Clin Microbiol
10. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep
11. Kallestrup P, Zinyama R, Gomo E, et al. Schistosomiasis and HIV-1 infection in rural Zimbabwe: effect of treatment of schistosomiasis on CD4 cell count and plasma HIV-1 RNA load. J Infect Dis
12. Erikstrup C, Kallestrup P, Zinyama R, et al. Predictors of mortality in a cohort of HIV-1-infected adults in rural Africa. J Acquir Immune Defic Syndr
13. Schüpbach J, Tomasik Z, Nadal D, et al. Use of HIV-1 p24 as a sensitive, precise and inexpensive marker for infection, disease progression and treatment failure. Int J Antimicrob Agents
14. Zijenah LS, Tobaiwa O, Rusakaniko S, et al. Signal-boosted qualitative ultrasensitive p24 antigen assay for diagnosis of subtype C HIV-1 infection in infants under the age of 2 years. J Acquir Immune Defic Syndr
15. Stevens G, Rekhviashvili N, Scott LE, et al. Evaluation of two commercially available, inexpensive alternative assays used for assessing viral load in a cohort of human immunodeficiency virus type 1 subtype C-infected patients from South Africa. J Clin Microbiol