Screening for HIV: Systematic Review to Update the 2005 U.S. Preventive Services Task Force Recommendation
Download the PDF here|
Ann Intern Med. 20 November 2012
Roger Chou, MD; Shelley Selph, MD, MPH; Tracy Dana, MLS; Christina Bougatsos, MPH; Bernadette Zakher, MBBS; Ian Blazina, MPH; and P. Todd Korthuis, MD, MPH
Background: A 2005 U.S. Preventive Services Task Force (USPSTF) review found good evidence that HIV screening is accurate and that antiretroviral therapy (ART) for immunologically advanced disease is associated with substantial clinical benefits, but insufficient evidence to determine the effects on transmission or in less immunologically advanced disease.
Purpose: To update the 2005 USPSTF review on benefits and harms of HIV screening in adolescents and adults, focusing on research gaps identified in the prior review.
Data Sources: MEDLINE (2004 to June 2012) and the Cochrane Library (through the second quarter of 2012).
Study Selection: Randomized trials and observational studies that compared HIV screening strategies and reported clinical outcomes, evaluated the effects of starting ART at different CD4 cell count thresholds and long-term harms, or reported the effects of interventions on transmission risk.
Data Extraction: 2 authors abstracted and checked study details and quality using predefined criteria.
Data Synthesis: No study directly evaluated the effects on clinical outcomes of screening versus no screening for HIV infection. A randomized trial and a subgroup analysis from a randomized trial found that ART initiation at CD4 counts less than 0.250 x 109 cells/L was associated with a higher risk for death or AIDS-defining events than initiation at CD4 counts greater than 0.350 x 109 cells/L (hazard ratios, 1.7 [95% CI, 1.1 to 2.5] and 5.3 [CI, 1.3 to 9.6]). Large, fair-quality cohort studies also consistently found that ART initiation at CD4 counts of 0.350 to 0.500 x 109 cells/L was associated with lower risk for death or AIDS-defining events than delayed initiation. New evidence from good-quality cohorts with longer-term follow-up confirms a previously observed small increased risk for cardiovascular events associated with certain antiretrovirals. Strong evidence from 1 good-quality randomized trial and 7 observational studies found that ART was associated with a 10- to 20-fold reduction in risk for sexual transmission of HIV.
Limitations: Only English-language articles were included. Observational studies were included. Studies done in resource-poor or high-prevalence settings were included but might have limited applicability to general screening in the United States.
Conclusion: Previous studies have shown that HIV screening is accurate, targeted screening misses a substantial proportion of cases, and treatments are effective in patients with advanced immunodeficiency. New evidence indicates that ART reduces risk for AIDS-defining events and death in persons with less advanced immunodeficiency and reduces sexual transmission of HIV.
Primary Funding Source: Agency for Healthcare Research and Quality.
In 2008, an estimated 1.2 million persons in the United States were living with HIV, and approximately 1 in 5 were unaware of their status (1 - 3). Incidence of HIV in the United States is approximately 50 000 cases per year (1,4), with an estimated 20 000 such cases believed to be due to transmission from persons who are unaware that they are infected (5 - 6). Screening for HIV antibodies can detect infection in asymptomatic patients, who might benefit from interventions to reduce risk for AIDS-related clinical events and transmission.
In 2005, the U.S. Preventive Services Task Force (USPSTF) recommended screening all adolescents and adults at increased risk (defined as persons who reported HIV risk factors or were evaluated in settings with an HIV infection prevalence >1%) (7) on the basis of an earlier evidence review (8 - 10) that found a high yield from screening these patients, good evidence that HIV screening tests are accurate (sensitivity and specificity each >99%), and good evidence that treating HIV infection at immunologically advanced stages of disease (defined as CD4 counts <0.200 x 109 cells/L) with antiretroviral therapy (ART) markedly reduces risk for AIDS-related clinical events and death. Although the USPSTF found that ART was associated with short-term adverse events and an increased risk for long-term cardiovascular events, it determined that benefits substantially outweighed harms.
The USPSTF made no recommendation for or against screening for HIV in adolescents and adults who were not at increased risk for HIV infection (7). Because of the lower prevalence of HIV infection in such persons, it determined that the benefits of screening would be smaller than in higher-risk populations. The USPSTF found insufficient evidence to estimate benefits from screening persons at less immunologically advanced stages of disease (CD4 counts >0.200 x 109 cells/L) or on the effects of screening and subsequent interventions on HIV transmission.
In 2006, the Centers for Disease Control and Prevention (CDC) issued its revised guideline (11) recommending routine voluntary HIV screening of all persons aged 13 to 64 years, unless the prevalence of undiagnosed HIV infection was less than 0.1%. A key reason for this recommendation was evidence showing that 20% to 26% of patients with HIV infection report no risk factors (12), suggesting that risk-based screening strategies miss an important proportion of infected persons. Other reasons for the differences between recommendations include that the CDC placed greater weight on studies showing reductions in self-reported risky behaviors after HIV diagnosis, accepted modeling studies to estimate effects of HIV diagnosis on transmission risk, and placed greater weight on modeling studies that showed acceptable incremental cost-effectiveness ratios for screening versus no screening in low-prevalence populations (7).
This report updates the previous USPSTF review on HIV screening in nonpregnant adolescents and adults. It focuses on key research gaps identified in the earlier review with the potential greatest effect on assessment of benefits and harms associated with screening in persons not known to be at higher risk, including effects of screening, counseling, and ART use on HIV transmission risk; effectiveness of ART for HIV-infected persons with CD4 counts greater than 0.200 x 109 cells/L, and long-term harms of ART. The full report (13) provides detailed methods and data for the review, including search strategies and evidence tables with quality ratings of individual studies. Additional key questions about various screening strategies, the effects of knowledge of HIV-positive status and use of ART on risky behaviors, and associations between viremia or risky behaviors and HIV transmission are reviewed in the full report (13) but are not presented here.
Scope of the Review
We followed a standardized protocol and developed an analytic framework (Figure) that included the following key questions:
What are the benefits of universal or targeted HIV screening versus no screening in asymptomatic, nonpregnant adolescents and adults on disease transmission, morbidity, mortality, and quality of life?
What is the yield (number of new diagnoses) of HIV screening at different intervals in nonpregnant adolescents and adults?
How effective is ART for reducing transmission of HIV in nonpregnant adolescents and adults with chronic HIV infection?
How effective is behavioral counseling for reducing transmission of HIV in nonpregnant adolescents and adults with chronic HIV infection?
In asymptomatic, nonpregnant adolescents and adults with chronic HIV infection, what are the effects of initiating ART at different CD4 cell count or viral load thresholds on morbidity, mortality, and quality of life?
What are the longer-term harms associated with ART for nonpregnant adolescents and adults with chronic HIV infection?
We defined "universal" testing to mean routine testing of all persons aged 13 to 64 years, unless the prevalence of HIV infection has been documented to be less than 0.1% (11) and "targeted" screening to mean routine screening of persons who have risk factors or are in high-prevalence (>1%) settings (7).
Data Sources and Searches
We searched Ovid MEDLINE from 2004 to June 2012 and the Cochrane Library through the second quarter of 2012 and reviewed reference lists to identify relevant articles published in English.
At least 2 reviewers independently evaluated each study to determine inclusion eligibility. Papers were selected for full review if they were about HIV screening or treatments in nonpregnant adolescents and adults, were relevant to a key question, and met the predefined inclusion criteria (Appendix Table 1). For treatment interventions, we focused on ART and counseling to reduce transmission risk. Outcomes were mortality, AIDS-related events, HIV transmission risk, and long-term (defined as ≥2 years after initiation of treatment) cardiovascular harms associated with ART. We included randomized, controlled trials and cohort studies for all key questions. We also included systematic reviews published since 2010 that met all predefined quality criteria (14).
Data Abstraction and Quality Rating
One investigator abstracted details about the study design, patient population, setting, screening method, interventions, analysis, follow-up, and results. A second investigator reviewed data abstraction for accuracy. Two investigators independently applied criteria developed by the USPSTF (15) to rate the quality of each study as good, fair, or poor. Discrepancies were resolved by consensus.
We assessed the aggregate internal validity (quality) of the body of evidence for each key question as good, fair, or poor by using methods developed by the USPSTF, on the basis of the number, quality, and size of studies; consistency of results among studies; and directness of evidence (15). Meta-analysis was not attempted, although we reported meta-analyses from published systematic reviews that met our quality criteria.
Role of the Funding Source
This research was funded by the Agency for Healthcare Research and Quality (AHRQ) under a contract to support the work of the USPSTF. Investigators worked with USPSTF members and AHRQ staff to develop and refine the scope, analytic framework, and key questions; resolve issues arising during the project; and finalize the report. The AHRQ had no role in study selection, quality assessment, synthesis, or development of conclusions. The AHRQ provided project oversight; reviewed the draft report; and distributed the draft for peer review, including to representatives of professional societies and federal agencies. The AHRQ performed a final review of the manuscript to ensure that the analysis met methodological standards. The investigators are solely responsible for the content and the decision to submit the manuscript for publication.
The Appendix Figure shows the results of the search and study selection process.
Clinical Benefits of Universal or Targeted Screening
No randomized trial or observational study compared clinical outcomes between adults and adolescents screened and not screened for HIV infection.
Yield of HIV Screening at Different Intervals
No randomized trial or observational study evaluated the yield of repeated HIV screening compared with 1-time screening or compared the yield of different strategies for repeated screening (such as risk-based repeated screening vs. a routinely repeated test).
Effectiveness of ART for Reducing HIV Transmission
A good-quality systematic review (16) evaluated the association between use of ART and risk for HIV transmission from HIV-positive persons to uninfected sexual partners. It included 1 good-quality randomized, controlled trial (17) and 7 observational studies (18 - 24) (Appendix Table 2).
The randomized, controlled trial (HIV Prevention Trials Network study 052) compared early ART initiation (started at enrollment) with delayed therapy (after CD4 count decreased to <0.250 x 109 cells/L or onset of symptoms) in HIV-infected patients with baseline CD4 counts of 0.350 to 0.550 x 109 cells/L and an HIV-negative partner (17). Fifty-four percent of the 1763 couples were from Africa, with the remainder from Brazil, India, Thailand, and the United States. Ninety-seven percent of couples were heterosexual, and 94% were married. All couples received condoms and counseling. The trial was designed to follow patients for 5 years but was terminated early after meeting prespecified criteria for efficacy in interim analyses. At a median follow-up of 1.7 years, risk for seroconversion in HIV-negative partners was much lower in the early-therapy group than in the delayed-therapy group (0.3 vs. 2.2 per 100 person-years; hazard ratio [HR], 0.11 [95% CI, 0.04 to 0.32]). When restricted to cases that were genomically linked to the HIV-infected patient enrolled in the trial, the HR was 0.04 (CI, 0.01 to 0.27).
Results of the 7 observational studies (18 - 24) included in the systematic review (16) were consistent with the randomized trial (17). Sample sizes ranged from 93 to 3408 couples, with typical follow-up between 1 and 3 years (range, 3 months to 9 years). All were cohort studies of HIV-serodiscordant, heterosexual couples from Africa, Italy, Spain, Brazil, or China. Six studies (18 - 22,24) were rated fair-quality and the seventh (23) was a conference abstract. Three studies (19,21,24) adjusted for possible confounding variables, such as age, sex, condom use, or frequency of sexual intercourse.
Six (18 - 23) of the 7 observational studies reported that persons receiving ART had a lower risk for HIV transmission than untreated persons, for a pooled HR of 0.34 (CI, 0.13 to 0.92; I2 = 73%) (16). Exclusion of 1 study with inadequate person-time data (24) and 1 older study that included persons treated with monotherapy (21) resulted in a pooled HR of 0.16 (CI, 0.07 to 0.35) and eliminated statistical heterogeneity (I2 = 0%). The treatment effect was also more pronounced when the analysis was restricted to couples in which the HIV-infected person had a CD4 count less than 0.200 x 109 cells/L (pooled HR, 0.06 [CI, 0.01 to 0.54]) (18 - 20,22).
Effectiveness of Behavioral Counseling for Reducing Transmission
The previous USPSTF review (8 - 10) found no randomized trials or controlled observational studies on the effects of counseling HIV-positive persons about risky behaviors on HIV transmission risk.
There remains little direct evidence on the effects of testing and counseling about risky behaviors on HIV transmission. Two studies on effects of counseling regarding sexual behaviors in HIV serodiscordant couples (25 - 26) were not designed to assess effects on transmission rates and were severely underpowered (5 new HIV diagnoses were observed in each study). No study estimated the effects of counseling HIV-positive persons about injection drug use behaviors on transmission rates.
Effectiveness of Initiating ART at Different CD4 Cell Count Thresholds on Clinical Outcomes
The previous USPSTF review included good-quality randomized, controlled trials (27 - 29) and observational studies (30 - 37) that consistently found a lower risk for AIDS events and death with ART than with placebo or less-intensive regimens in patients with CD4 counts less than 0.200 x 109 cells/L. Evidence showing benefits of starting ART at higher CD4 cell counts was limited. Although a Swiss cohort study (38) found that starting ART at CD4 counts greater than 0.350 x 109 cells/L was associated with a lower risk for death and progression to AIDS than starting at less than 0.350 x 109 cells/L, 3 U.S. cohort studies (35 - 37) found no difference in risk between starting ART at CD4 counts between 0.350 and 0.500 x 109 cells/L versus delaying until CD4 counts were between 0.200 and 0.350 x 109 cells/L.
Two good-quality randomized trials (17,39) published since the previous USPSTF review and 1 subgroup analysis (40) from another randomized trial evaluated the effects of initiating ART at different CD4 cell count thresholds (Table 1). Five observational studies (reported in 7 publications) (41 - 45,48 - 49), each of which combined data from 12 to 23 U.S., European, and Australian cohorts (ranging from 9000 to >60 000 participants and 2- to 5-year follow-up, with substantial overlap in the cohorts included in the studies), also evaluated the effects of starting ART at different CD4 cell count thresholds (Table 1). All of the observational studies were rated fair-quality. None reported blinding of outcome assessors or persons analyzing data, and attrition rates were often not reported or were unclear. Although all studies adjusted for confounders, most provided insufficient information to determine baseline comparability of patients starting or not starting ART at different CD4 cell count strata.
A retrospective subgroup analysis of 477 patients in the SMART (Strategies for Management of Antiretroviral Therapy) randomized trial who were treatment-naive or had stopped therapy for at least 6 months found that ART initiation at CD4 counts less than 0.250 x 109 cells/L was associated with a higher risk for death or AIDS events than initiation at counts greater than 0.350 x 109 cells/L after a mean of 18 months (HR, 5.3 [CI, 1.3 to 9.6]) (40). The SMART trial was done in 33 primarily non-resource-poor countries. The HIV Prevention Trials Network study 052, conducted in 1763 patients from primarily resource-poor countries, also found initiation at CD4 counts less than 0.250 x 109 cells/L associated with a higher risk for death or AIDS events than initiation at counts greater than 0.350 x 109 cells/L after a median of 1.7 years (HR, 1.7 [CI, 1.1 to 2.5]) (17). Another randomized trial (39) with 816 participants found that ART initiation at CD4 counts less than 0.200 x 109 cells/L was associated with higher mortality than initiation at 0.201 to 0.350 x 109 cells/L (HR, 4.0 [CI, 1.6 to 9.8]; P = 0.001), but this trial was conducted in Haiti and evaluated lower CD4 count cutoffs for treatment than those in the United States.
Four observational studies (42 - 45,48) consistently found that ART initiation at CD4 counts between 0.350 and 0.500 x 109 cells/L was associated with a lower risk for death than deferred or no ART. One other study (49) found a reduction in risk that was not statistically significant. The HIV-CAUSAL (HIV Cohorts Analyzed Using Structural Approaches to Longitudinal data) collaboration (43), the largest study in our review (62 760 participants from 12 cohorts), found that ART initiation at CD4 counts of 0.350 to 0.500 x 109 cells/L was associated with a lower risk for death than noninitiation at these counts after 3.3 years of follow-up (adjusted HR, 0.55 [CI, 0.41 to 0.74]). Similarly, the NA-ACCORD (North American AIDS Cohort Collaboration on Research and Design) (44), with 17 517 participants from 22 cohorts, found that ART initiation at CD4 counts of 0.351 to 0.500 x 109 cells/L was associated with a lower risk for death than deferred treatment at these CD4 cell counts after 3 years of follow-up (adjusted RR, 0.61 [CI, 0.46 to 0.83]). In 2 studies (45,49), ART initiation at CD4 counts greater than 0.350 x 109 cells/L was also associated with a lower risk for the combined outcome of AIDS-defining events and death than deferred or no ART initiation. One other study (49) found a reduction in risk that was not statistically significant.
Studies on ART initiation at CD4 counts greater than 0.500 x 109 cells/L were less consistent. The NA-ACCORD cohort study (44) found that ART initiation at CD4 counts greater than 0.500 x 109 cells/L was associated with lower mortality than deferred therapy (adjusted RR, 0.54 [CI, 0.35 to 0.83]) and the HIV-CAUSAL collaboration (43) found a lower mortality risk that was not statistically significant (adjusted HR, 0.77 [CI, 0.58 to 1.0]). Another analysis from the HIV-CAUSAL Collaboration (42) that directly compared ART initiation at CD4 counts greater than 0.500 x 109 cells/L with initiation at greater than 0.350 x 109 cells/L found no difference in mortality (HR, 0.99 [CI, 0.89 to 1.2]). Two other large cohort studies found that ART initiation at CD4 counts greater than 0.500 x 109 cells/L was associated with no difference in risk for death when compared with noninitiation (48) or slightly delayed initiation (49). In all 4 studies, absolute mortality rates were low (2% to 5%) in patients with CD4 counts greater than 0.500 x 109 cells/L.
Results were also mixed for the combined outcome of death plus AIDS-defining events (not reported in the NA-ACCORD study 44). The HIV-CAUSAL collaboration (42) found that ART initiation at CD4 counts greater than 0.500 x 109 cells/L was associated with a lower risk for AIDS-defining events or death than initiation at greater than 0.350 x 109 cells/L (HR, 0.72 [CI, 0.64 to 0.81]). Two other studies (48 - 49) found no clear association between starting or not starting ART at CD4 counts greater than 0.500 x 109 cells/L and risk for AIDS-defining events or death.
Longer-Term Harms Associated With ART
The 2005 USPSTF review included results from the large, ongoing DAD (Data Collection on Adverse Events of Anti-HIV Drugs) study (23 468 participants), which found that increased risk for myocardial infarction was associated with longer exposure to ART (adjusted RR, 1.3 per year of exposure [CI, 1.1 to 1.4 per year of exposure]), although absolute event rates were low (3.5 per 1000 person-years) (50).
Subsequent analyses from the DAD study (51 - 53) and 3 other cohort studies (54 - 56) reported cardiovascular harms associated with ART through 4 to 6 years of follow-up (Appendix Table 3). Sample sizes ranged from 2952 to more than 30 000 persons. All of the studies were rated good-quality except 1, which was rated fair-quality because of lack of detail about baseline patient characteristics and blinding of study personnel (54). All studies adjusted for multiple confounders.
Like the earlier DAD results, the most recent analysis found that longer exposure to indinavir alone (adjusted RR, 1.1 per year of exposure [CI, 1.1 to 1.2 per year of exposure]), ritonavir-boosted indinavir (adjusted RR, 1.2 per year of exposure [CI, 1.1 to 1.3 per year of exposure]), and ritonavir-boosted lopinavir (adjusted RR, 1.1 per year of exposure [CI, 1.0 to 1.2 per year of exposure]) were each associated with a slightly higher risk for myocardial infarction than nonuse (53). No other protease inhibitor was associated with increased myocardial risk.
Evidence on the association between the nucleoside reverse transcriptase inhibitor abacavir and risk for myocardial infarction is mixed. Although 2 studies (53,55) found that abacavir was associated with increased risk (adjusted RRs, 1.7 and 2.0), 2 others (54,56) found no association (adjusted HRs, 0.6 and 1.2).
The DAD study also found that recent didanosine use was associated with increased myocardial infarction risk (adjusted RR, 1.4 [CI, 1.1 to 1.8]), but found no association when analyses were based on cumulative didanosine exposure (53). No association was found between use of other nucleoside reverse transcriptase inhibitors or the nonnucleoside reverse transcriptase inhibitors nevirapine or efavirenz and increased risk for cardiovascular events (53).
As in the 2005 USPSTF review (8 - 10), we found no direct evidence on the effects of screening for HIV infection versus no screening on clinical outcomes. Table 2 summarizes the other evidence reviewed in this update.
The 2005 review found good evidence that HIV screening tests are accurate and that identifying undiagnosed HIV infection and treating immunologically advanced disease (CD4 count <0.200 x 109 cells/L) are associated with substantial clinical benefits. However, it found insufficient evidence to estimate the effects of diagnosis and subsequent interventions on transmission risks or the clinical benefits of ART in patients with less immunologically advanced disease. New studies included in this update (17,39 - 40,43 - 45,48 - 49) provide strong evidence for the effectiveness of initiating ART at CD4 counts between 0.350 and 0.500 x 109 cells/L, although evidence showing benefit is less consistent for ART initiation at greater than 0.500 x 109 cells/L (43 - 44,48 - 49). Recent studies indicate that about 54% of patients present for initial HIV care with CD4 counts less than 0.350 x 109 cells/L (57) and about 75% were diagnosed at CD4 counts less than 0.500 x 109 cells/L (58), suggesting that many patients identified by screening would benefit from immediate ART initiation. Additional research (51 - 53,55) confirms previous findings of a small but statistically significant increase in risk for long-term cardiovascular harms associated with use of certain protease inhibitors. In the DAD study, the absolute increase in risk per year of exposure with certain older protease inhibitors was about 0.3 myocardial infarctions per 1000 person-years (53), compared with an absolute decrease in mortality of about 3.2 to 20 per 1000 person-years after initiating ART, depending on the CD4 cell count at baseline (43). Whether current first-line protease inhibitors and other antiretrovirals are also associated with increased cardiovascular risk is not yet established. Long-term ART is also associated with other harms, including osteoporotic fractures (59) and lipodystrophy (60), that were not addressed in this review.
Strong evidence from a randomized trial and multiple observational studies (16 - 17) indicates that ART use is associated with a 10- to 20-fold reduction in risk for sexual transmission. Recent evidence showing that counseling interventions were relatively ineffective in reducing risky behaviors in HIV-infected persons (61) suggests that the beneficial effects of screening on transmission are probably driven by use of ART.
Our study has limitations. We excluded non-English-language articles, which could have resulted in language bias, although we identified no non-English-language studies that would have met our inclusion criteria. We did not search for studies published only as abstracts and could not formally assess for publication bias by using graphical or statistical methods because of the few studies for each key question and the differences in study design, populations, and outcomes assessed. We included observational studies, which are more susceptible to bias and confounding than well-conducted randomized trials, although we focused on results from studies that performed statistical adjustment for potential confounding. We also included studies conducted in resource-poor and high-prevalence settings, which could limit the applicability of our findings to U.S. practice.
Additional research may further clarify benefits and harms of screening. Continued follow-up of patients receiving ART is needed to further understand the effects of long-term exposure, because many patients receive treatment for far longer than the 6 years evaluated in the longest studies to date. No clinical study has evaluated the yield of repeated HIV screening, which probably depends on the incidence of new infections in a population (61 - 63). The START (Strategic Timing of Antiretroviral Treatment) randomized trial (64), which compares ART initiation at CD4 counts greater than 0.500 x 109 cells/L with deferred treatment until CD4 counts decrease to less than 0.350 x 109 cells/L, is currently recruiting and should help further clarify the effects of very early ART initiation.
The main area of discrepancy between HIV screening guidelines is whether to routinely screen populations not known to be at increased risk (11,65). Screening tests for HIV are highly accurate, but targeted screening misses a substantial proportion of infected persons because of undisclosed or unknown risk factors. Evidence published since the 2005 USPSTF review shows highly beneficial effects of ART for reducing sexual transmission of HIV and risk for AIDS-defining events and death in persons with less immunologically advanced stages of disease.