You Are Here:
Roger Chou, M.D.; Laurie Hoyt Huffman, M.S.; Rongwei Fu, Ph.D.; Ariel K. Smits, M.D., M.P.H.; and P. Todd Korthuis, M.D., M.P.H.
Address correspondence to Roger Chou, MD, Oregon Health & Science University, Mail Code BICC, 3181 SW Sam Jackson Park Road, Portland, OR 97239; E-mail, firstname.lastname@example.org.
This article was first published in Ann Intern Med 2005:143:55-73.
To update its 1996 recommendations, the U.S. Preventive Services Task Force (USPSTF) commissioned a new systematic review of the risks and benefits of testing for anti-HIV antibodies in asymptomatic adolescents and adults.
Infection with HIV-1 is estimated to affect 850,000 to 950,000 persons in the United States.1 Of those infected, 25% (180,000 to 280,000) are thought to be unaware of their status.1 Almost all patients with untreated HIV infection eventually develop AIDS.2 In the United States, more than 500,000 patients with AIDS have died; approximately 18,000 died in 2003.3 AIDS is the seventh leading cause of death in persons 15 to 24 years of age, and the fifth leading cause in persons 25 to 44 years of age.4 Since 1992, 40,000 new HIV infections have been diagnosed annually.5 Statistical modeling suggests that approximately half of HIV-infected persons in the United States acquire their infection by 25 years of age.6
Infection with HIV causes immune deficiency to a large extent by decreasing the level and function of CD4 T lymphocytes. In untreated patients with CD4 cell counts less than 0.200 x 109 cells/L, the chance of clinical progression or death over 3 years is approximately 86%.7 A higher HIV-1 viral load also predicts faster disease progression.7-10
To update its 1996 recommendations, the U.S. Preventive Services Task Force (USPSTF) commissioned a new systematic review of the risks and benefits of testing for anti-HIV antibodies in asymptomatic adolescents and adults.11 Another article in this issue reviews screening in pregnant women.12
Figure 1 summarizes the analytic framework and key questions for this review. Key question 1 addresses direct evidence on the effects of screening on clinical outcomes. The other key questions address the chain of evidence necessary to estimate the effects of screening on clinical outcomes if direct evidence is insufficient. Appendix A discusses the scope and methods used for this review in more detail.
Briefly, we identified relevant studies from MEDLINE® (1983 through 30 June 2004) and the Cochrane Clinical Trials Registry (2004, issue 2), reference lists, hand searches of relevant journals, and suggestions from experts (Appendix B). We selected studies that provided evidence on the benefits and harms of screening, risk factor assessment, accuracy of testing, followup testing, interventions, acceptability of HIV testing, and cost-effectiveness of screening in outpatient settings in the highly active antiretroviral therapy (HAART) era. For interventions, we focused on studies of HAART.13,14 We also reviewed studies on the effectiveness of counseling on risky behaviors15 and prophylaxis against opportunistic infections.16 A separate report17 reviews the effectiveness of other interventions (immunizations, more frequent Papanicolaou testing, and routine monitoring and followup).
We assessed the internal validity and relevance of included studies using predefined criteria developed by the USPSTF (Appendix C).18,19 We rated the overall body of evidence for each key question using the system developed by the USPSTF.
We used the results of the evidence review to construct an outcomes table estimating the effects of one-time screening for HIV infection in hypothetical cohorts of adolescents and adults. We calculated numbers needed to screen (NNS) and treat (NNT) to prevent 1 case of clinical progression or death or to cause 1 cardiovascular complication for each cohort. The point estimates and 95% CIs for NNS and NNT were based on Monte Carlo simulations.
This research was funded by the Agency for Healthcare Research and Quality (AHRQ) under a contract to support the work of the USPSTF. Agency staff and USPSTF members participated in the initial design of the study and reviewed interim analyses and the final report. Draft reports were distributed to 25 content experts for review. Agency approval was required before this manuscript could be submitted for publication, but the authors are solely responsible for the content and the decision to submit it for publication.
Does Screening for HIV Infection in Asymptomatic Adolescents and Adults Reduce Premature Death and Disability or Spread of Disease?
No studies compared clinical outcomes between patients in the general population who were screened or not screened for HIV.
Can Clinical or Demographic Characteristics Identify Subgroups of Asymptomatic Adolescents and Adults at Increased Risk for HIV Infection Compared to the General Population?
A substantial proportion of Americans report behaviors that could put them at risk for HIV infection (Table 1).20 A recent U.S. telephone survey (n = 33,913) found that 11% of sexually active respondents reported multiple partners within the last year, and 4.2% reported other high-risk behaviors.21 Adolescents,22,23 men who have sex with men,24 and persons attending sexually transmitted disease clinics also report high rates of recent risky behaviors.25 Even in settings with good access to health care, high-risk behaviors often remain undetected26 or fail to lead to testing despite identification.27
The largest (n = 1,281,606) U.S. study found that 20% to 26% of HIV-infected people identified at federally funded testing sites reported no risk factors.28 Other studies in a variety of settings indicated that 7% to 51% of HIV-positive patients reported no risk factors.26,29-36 The rate of HIV positivity in patients reporting no risk factors was lower in low-prevalence (0.1% to 2.0%) than in high-prevalence (>5%) sites (0.2% to 0.8% vs. 1.4% to 5.7%).28
One good-quality prospective study in a sexually transmitted disease clinic evaluated different methods of selective screening, such as screening only persons with reported risk factors, those with reported risk factors or those in high-prevalence demographic groups, or screening everybody. In this study, screening only persons who reported risk factors (5.8% of those tested) would have resulted in 74% (79 of 107) missed diagnoses. A broader strategy (70% tested) of also screening persons in high-prevalence demographic groups (black men or persons > 30 years of age) would have resulted in substantially fewer (8%) missed diagnoses.37 Two retrospective studies found that similar selective strategies would have resulted in 33% to 41% of the population being tested and 7% (1 of 14)38 to 13% (192 of 1,474)39 missed diagnoses. Four U.S. studies in high-prevalence (>1%) settings demonstrated an increased yield after the implementation of routine voluntary HIV screening.40-43
What Are the Test Characteristics of HIV Antibody Test Strategies?
The use of repeatedly reactive enzyme immunoassay followed by confirmatory Western blot or immunofluorescent assay remains the standard method for diagnosing HIV-1 infection.44,45 A large study of HIV testing in 752 U.S. laboratories reported a sensitivity of 99.7% and specificity of 98.5% for enzyme immunoassay,45 and studies in U.S. blood donors reported specificities of 99.8% and greater than 99.99%.46,47 With confirmatory Western blot, the chance of a false-positive identification in a low-prevalence setting is about 1 in 250 000 (95% CI, 1 in 173,000 to 1 in 379,000).48
Three rapid (results available in 10 to 30 minutes) HIV tests are in use in the United States; 2 (Uni-Gold Recombigen, Trinity Biotech Plc., Bray, Ireland, and OraQuick Advance, OraSure Technologies, Bethlehem, Pennsylvania) for true point-of-care testing,49 and 1 (Reveal G2, MedMira Laboratories, Inc., Halifax, Nova Scotia, Canada) performed in a laboratory. Three good-quality and 10 fair-quality studies evaluated accuracy of rapid tests on blood specimens against standard HIV testing.50-55 Ten were reported in manufacturer inserts.50-52 Most studies reported the accuracy of rapid tests before confirmatory testing because patients may be notified of results before confirmation is available.56
For the OraQuick test, 3 good-quality studies found sensitivities ranging from 96% to 100% and specificity greater than 99.9%.53-55 Three fair-quality studies found sensitivities ranging from 99.6% to 100%, with specificity 100% in all.50 For the Uni-Gold and Reveal tests, 7 fair-quality studies reported sensitivities ranging from 94% to 100% and specificities greater than 99%.50,52 The positive predictive values for the Reveal and Uni-Gold tests were calculated at 25% to 50% in settings with a prevalence of 0.3%, and at 85% to 95% in settings with a prevalence of 5%.57 One good-quality study among 5744 U.S. pregnant women (prevalence, 0.59%) found a positive predictive value of 90% (4 false-positive results) and a negative predictive value of 100% for the OraQuick test using blood.53
Two large (n = 3,570 and n = 4,442), good-quality studies of the OraSure Oral Specimen Collection Device (Epitope, Inc., Beaverton, Oregon) measured sensitivities of 99.9% and 99.2% and specificities of 99.9% and 99.2%.58,59 Urine HIV tests generally appear less accurate than standard testing and are not in widespread use in the United States.60-63 A good-quality (n = 1,255) study of the only U.S. Food and Drug Administration-approved home collection kit (Home Access, Home Access Health Corp., Hoffman Estates, Illinois) found that the sensitivity and specificity obtained with use of finger-stick blood spot samples were both 100% compared to standard testing.64 More than 98% of participants in 2 studies obtained adequate samples for testing.64,65
No studies have evaluated the optimal frequency of HIV screening, which partly depends on the incidence and the prevalence of undetected HIV infection in the group being tested.66
What Are the Harms Associated with Screening?
Information on the frequency and consequences (anxiety, labeling) of false-positive test results is anecdotal67-69 False- and true-negative results could provide false reassurance if high-risk behaviors are continued.
True-positive HIV test results are associated with important harms, including fears of rejection, abandonment, verbal abuse, and physical assault.70 A substantial proportion (20% to 25%) of Americans continue to agree with stigmatizing statements about HIV.71,72 Four percent of 142 patients with recently diagnosed HIV infection reported losing a job because of their status, 1% had been asked to move, and 1% had been assaulted.73
Notification of a positive HIV test result can lead to emotional and psychological distress. On the other hand, receipt of a negative HIV test result is associated with reduced anxiety in at-risk individuals74 Although earlier studies reported high suicide rates after a positive test result,75-78 no studies have addressed suicide risk after an HIV diagnosis in the HAART era. A large prospective cohort study through 1993 found that suicide rates after routine screening were similar between HIV-positive and HIV-negative military recruits.79 Counseling may reduce distress after a positive test result.80-83
Both HIV-negative and HIV-positive persons appear to have similar rates of intimate partner violence when matched for high-risk behaviors.84-86 One prospective cohort study found that rates of abuse declined after disclosure of HIV status (87). Several small observational studies did not find an increased rate of partnership dissolution after a positive diagnosis.87-89
Is Screening Acceptable to Patients?
In the United States, as of 2002 approximately half (43.5%) of persons age 18 to 64 years had been tested at least once for HIV.90 The proportion of tested female adolescents is substantially lower at 25%.91 Among persons reporting high-risk behaviors, recent studies found that 20% to 30% had never been tested.25,92,93
A good-quality systematic review of 62 studies reported that acceptance rates of voluntary HIV testing varied widely (from 11% to 91%) in the United States, even within similar health care settings.94 In general, low-prevalence settings were associated with lower acceptance rates. Higher acceptance rates were associated with the client's perception of HIV risk, acknowledgment of risk behaviors, confidentiality protections, and the provider's belief that testing would be beneficial.
One United Kingdom study of "opt-out" testing (in which an HIV test is considered routine and is performed unless the patient declines) in nonpregnant persons found that uptake increased from 35% to 65%.95 In several studies, anonymous testing was associated with increased testing rates96-98 or higher mean CD4 cell count at diagnosis,99 although others did not find a clear association.100-102 In Connecticut, testing rates in adolescents doubled after removal of a parental consent requirement.103
No clinical trials have evaluated the incremental acceptability of alternative testing (rapid test, home sampling, or oral sampling) compared with standard testing. A recent observational study found that 29% to 69% of patients in different settings accepted rapid testing.104 Another found that all 150 patients being treated for substance abuse who accepted testing chose an oral fluid test over a blood test.105 In studies of patients who accepted home sample collection106,107 or oral fluid sampling,108 a substantial proportion (22% to 33% for home sampling and 58% for oral fluid sampling) had not been previously tested.
How Many Newly Diagnosed HIV-Positive Patients Meet Criteria for Antiretroviral Treatment or Prophylaxis against Opportunistic Infections?
In asymptomatic HIV-positive patients, viral load and CD4 cell count testing are used to determine eligibility for HAART and opportunistic infection prophylaxis.13,16 Antiretroviral therapy is currently recommended for patients with CD4 cell counts less than 0.200 x 109 cells/L. Antiretroviral therapy can also be considered for other asymptomatic patients at high risk for disease progression (CD4 cell count < 0.350 x cells 109/L or viral load > 100,000 copies/mL). Interventions are generally less effective in persons with advanced immune deficiency,109 although some benefit is seen.110,111
No studies report both CD4 cell count and viral load in newly diagnosed patients. Seven U.S. studies in different settings found that the proportion of patients with CD4 cell counts less than 0.200 x 109 cells /L at diagnosis or when establishing care ranged from 12% to 43%, and the proportion with CD4 cell counts less than 0.500 x 109 cells/L ranged from 46% to 80%.26,41,112-116
Screening could identify a higher proportion of persons whose CD4 cell counts have not decreased below thresholds for interventions. In addition, patients with an adequate response to HAART can safely discontinue prophylaxis against certain opportunistic infections.16 We identified no studies estimating the effects of screening or treatment on the proportion of patients qualifying for different interventions.
How Many HIV-Positive Patients Who Meet Criteria for Interventions Receive Them?
Patients positive for HIV who meet criteria for interventions may not receive them. Ten percent to 44% of tested patients do not have a post-test counseling session or fail to return for test results,117-119 although most (79% to 93%) positive patients are eventually located.30,120 Two recent studies of routine testing in urgent care centers found that 74% to 82% of patients learned of their positive results.40-41
Rapid testing was associated with a higher rate of HIV-positive persons learning their status than was standard testing in an anonymous testing clinic (100% vs. 86%),121 sexually transmitted disease clinic (97% vs. 79%),121 and emergency department setting (73% vs. 62%).122 In noncomparative studies, rapid testing resulted in more than 98% of patients learning their status104,123 Of 174 316 persons submitting home samples, 95% to 96% called for results.106
Patients positive for HIV may delay medical care or not receive care at all. In 1996, 36% to 63% of HIV-positive patients were regularly seeing a non-emergency department provider.124 Studies in the United States found that 17% to 29% of patients had delayed entry into care for at least 3 months,125,126 and 11% to 39% delayed it for at least 1 year.126-128 A study of rapid testing found that entry into care within 6 months ranged from 100% (in a sexually transmitted disease clinic) to 22% (in a jail).104
No prospective studies measured the proportion of newly diagnosed HIV-positive persons who received appropriate treatment. Four large (n = 1,411 to 9,530) U.S. surveys found that 53% to 85% of HIV-positive patients were receiving antiretroviral therapy according to then-current guidelines.129-132
How Effective Are Interventions in Improving Clinical Outcomes?
Currently, HAART regimens with 3 or more antiretroviral agents, usually from at least 2 different classes, are the standard of care for HIV-infected persons receiving antiretroviral therapy.13,14 A good-quality systematic review of 54 randomized, controlled trials with 16 684 HIV-infected patients with limited or no antiretroviral experience found that 3-drug therapy was more effective than 2-drug therapy (odds ratio, 0.62 [95% CI, 0.50 to 0.78]).133 Observational studies indicate that HAART can result in sustained (up to 4 to 5 years) improvements in CD4 cell counts and viral loads,134-136 although long-term clinical outcomes data are not yet available.
Large, good-quality cohort studies from the United States137-140 and Europe141-143 parallel the findings of the systematic review regarding the effectiveness of HAART. In addition, studies have consistently found a marked decline in morbidity and mortality among U.S. HIV-infected patients that coincided with the widespread adoption of HAART.138-140,144-149 In 2 U.S. studies, for example, mortality rates declined from 20.2140 and 29.4138 per 100 person-years to 8.4 and 8.8 per 100 person-years, respectively.
Few trials have adequately assessed the effect of HAART on quality of life or functional status (such as ability to work).133 Four fair-quality trials of 3-drug vs. 2-drug regimens reported conflicting results for differences in quality-of-life outcomes.150-153
The use of HAART could decrease the spread of HIV from infected persons by decreasing viral loads.154 On the other hand, increases in risky behaviors by patients receiving HAART could offset the beneficial effects of viral suppression.155-158 A recent good-quality meta-analysis of 25 studies found no association between receipt of HAART or having an undetectable viral load and unprotected sex.159 Among both seronegative and seropositive persons, however, unprotected intercourse was associated with optimistic beliefs about HAART or an undetectable viral load (odds ratio, 1.82 [CI, 1.52 to 2.17]).
No studies have estimated the effects of HAART on horizontal transmission rates. One cohort study found that heterosexual transmission from monogamous zidovudine-treated men was lower than that from untreated men (relative risk, 0.5 [CI, 0.1 to 0.9]).160 An epidemiologic study estimated that the annual HIV transmission rate from HIV-seropositive persons in the United States declined from 13% in 1987 (the year zidovudine was introduced) to 5.5% in 1989, and has remained steady at approximately 4.2% since 1990.161 This study was not designed to assess the relative contribution of antiretroviral therapy, changes in high-risk behaviors, or other factors to changes in transmission rates.
Because the incidence of new HIV infections has remained steady while mortality due to AIDS has declined, the number of persons living with HIV infection in the United States continues to increase.3 A substantial proportion of HIV-infected persons report behaviors that increase the risk for transmitting infection.15,24,126,162-164 Data on the link between sexual behaviors and reduced risk for HIV transmission are strongest for consistent use of condoms for prevention of heterosexual transmission.165,166 Good-quality systematic reviews found that testing plus counseling is most effective in reducing risky behaviors among serodiscordant heterosexual couples and those testing HIV-positive, with less evidence for beneficial effects in other populations.167-169 Several recent fair-quality observational studies reported decreased self-reported risky behaviors after patients had HIV testing or received a positive diagnosis.170-173 Some174-178 but not all179-182 fair-quality randomized trials found that targeted (tailored to participant needs) or more intensive counseling was associated with greater reductions in risky behaviors than standard or less intensive counseling, but counseling methods varied greatly across trials.
No clinical trials evaluated the impact of testing and counseling compared to no testing and counseling on HIV transmission rates. One prospective U.S. study of 144 serodiscordant heterosexual couples who received counseling and reported reduced risky behaviors found no seroconversion after 193 couple-years of followup.183 A prospective African study found that the rate of seroconversion among uninfected female partners of HIV-positive men was 6 to 9/100 person-years, compared with 22/100 person-years in women with untested partners.184 Two observational studies found that testing plus counseling was associated with a moderate (about 33%) decrease in sexually transmitted diseases among those who tested positive but that it increased the risk among those who tested negative (relative risk, 1.27 to 2).185,186 Two good-quality randomized, controlled trials found that more interactive counseling was more effective than standard counseling in reducing sexually transmitted disease rates among HIV-positive women176 and seronegative heterosexual persons,187 although there were too few new HIV infections to detect differences in HIV rates.187
No studies have estimated the effects of counseling HIV-positive persons regarding injection drug use behaviors on HIV transmission rates. Although cross-sectional studies found that HIV-positive drug users reported less risky behaviors than those untested or not infected,188-190 1 randomized trial191 and 1 prospective study192 found that testing plus counseling was not associated with decreased drug behaviors. On the other hand, 2 randomized trials found that more intense counseling reduced drug use behaviors more than did standard counseling.174,193
Prophylaxis Against Opportunistic Infections
Table 2 summarizes 2 good-quality systematic reviews194,195 and 3 clinical trials196-198 of primary prophylaxis against Pneumocystis carinii pneumonia. Prophylaxis was associated with a nonsignificant mortality benefit.194 Several medications used for prophylaxis against P. carinii pneumonia are also effective for toxoplasmosis prophylaxis.16,195
Two good-quality systematic reviews199,200 found that isoniazid prophylaxis was effective at preventing tuberculosis (risk reduced by 60% to 86%) and death (risk reduced by 21% to 23%) in HIV-positive patients with a positive tuberculin skin test result.16
Table 3 summarizes 4 good-quality placebo-controlled trials201-203 and 2 head-to-head trials204,205 of primary prophylaxis against disseminated Mycobacterium avium intracellulare complex infection. Only clarithromycin was associated with a significant mortality benefit.202
Two placebo-controlled trials of ganciclovir for cytomegalovirus prophylaxis found mixed results for reducing invasive CMV infection, no mortality benefit, and significant adverse events.206,207
In Asymptomatic Patients with HIV Infection, Does Immediate Antiretroviral Treatment Result in Improvements in Clinical Outcomes Compared to Delayed Treatment until the Patient Is Symptomatic?
Initiation of HAART in asymptomatic patients must be weighed against potential harms, including effects on quality of life, long-term adverse events, and the development of resistance. Current U.S. guidelines recommend that all asymptomatic patients with CD4 cell counts less than 0.200 x 109 cells/L be offered HAART.13 Recommendations for other asymptomatic patients are less firm.
Twelve observational studies evaluated the risk for disease progression or death in asymptomatic patients initiating HAART at different CD4 cell count thresholds above 0.200 x 109 cells/L. All lasted less than 4 years and could underestimate long-term risks for immediate treatment. Other limitations of studies include not controlling for lead-time bias208 and not accounting for important confounders, such as the level of adherence209 or physician experience.110
Four fair-quality observational studies controlled for lead-time bias by identifying cohorts of patients at initial CD4 cell count strata and evaluating outcomes according to when they received HAART.210-213 Three U.S. studies found no significant benefit associated with starting HAART at CD4 cell counts between 0.350 and 0.500 x 109 cells/L versus between 0.200 and 0.350 x 109 cells/L (Table 4).210,212,213 A Swiss study reported a benefit for starting at CD4 cell counts above 0.350 x 109 cells/L but did not stratify results of patients starting at CD4 cell counts above or below 0.200 x 109 cells/L.211 Six109,214-218 of 8209,219 other observational studies that did not control for lead-time bias or used novel methodologic approaches found a benefit or trend toward benefit from initiation of treatment at CD4 counts above versus below 0.350 x 109 cells/L.
A randomized clinical trial (the SMART [Strategies for Management of Anti-Retroviral Therapies] study)220 comparing viral suppression in asymptomatic patients with a CD4 cell count less than 0.350 x 109 cells/L with delay until counts decrease below 0.250 x 109 cells/L is in progress, with preliminary results expected in 5 to 7 years.221
What Are the Harms Associated with Antiretroviral Therapy?
Individual antiretroviral drugs, drug classes, and drug combinations are all associated with specific adverse event profiles.13 Retrospective U.S. cohort studies found that 61% of patients had changed or discontinued their initial HAART regimen by 8 months222 and that the median duration of the initial regimen was less than 2 years;223 40% to 50% discontinued the initial regimen because of adverse events. Many antiretroviral-associated adverse events, however, are short-term or self-limited, and effective alternatives can often be found.14,134 Detailed and regularly updated guidelines review adverse events associated with specific antiretroviral drugs, drug classes, and combinations.13 Certain drugs and combinations are not recommended because of associated adverse events.
A recent good-quality systematic review found that 26 of 54 trials of antiretroviral therapy reported drug-related withdrawals, a marker for intolerable or severe adverse events.133 Among trials comparing 3-drug and 2-drug regimens, dropout rates were similar if both regimens either included protease inhibitors or were protease inhibitor-sparing. In a large (n = 1,160), good-quality Swiss cohort study of adverse events in clinical practice, 47% of patients reported a clinical adverse event that was probably or definitely attributed to HAART within the previous 30 days.224 Among these, 9% were graded as serious or severe.
The use of HAART is associated with metabolic disturbances (lipodystrophy syndrome, hyperlipidemia, and diabetes) that are related to an increased risk for cardiovascular events.225,226 The largest prospective study on the risk for cardiovascular events associated with both protease inhibitor-based and non-protease inhibitor-based combination regimens was a good-quality study of 23,468 patients in 11 cohorts.227 It found that the incidence of myocardial infarction increased with longer exposure (adjusted relative rate per year of exposure, 1.26)227. The relative risk for the combined outcome of myocardial infarction, invasive cardiovascular procedures, or stroke was similarly increased, although the event rate was higher (5.7 events/1000 person-years vs. 3.5 events/1000 person-years for myocardial infarction alone).229 Other studies primarily evaluating the cardiovascular risk associated with protease inhibitors also generally found an increased risk.230-238
Studies evaluating trends over time reported mixed findings regarding the rate of cardiovascular events in HIV-infected patients since the introduction of HAART. These studies are limited by potential confounding from changes in clinical practice and the demographic characteristics of persons surviving with HIV infection.239-242
Estimates of the Numbers Needed To Screen and Treat
Table 5 estimates outcomes after 3 years from 1-time screening for HIV in 3 hypothetical cohorts of 10 000 asymptomatic persons (0.3% prevalence, 1% prevalence, and 5% to 15% prevalence [high risk]) (Appendix Table 1 gives base-case assumptions). Because no trials directly compare 3-drug regimens to placebo, we indirectly calculated (Appendix A) a relative risk for clinical progression or death of 0.35 (CI, 0.25 to 0.47).133 For all cohorts, the number of cases of clinical progression or deaths that were prevented greatly outweighed the number of cardiovascular adverse events caused by antiretroviral therapy. Evidence was insufficient to estimate the effects of screening on transmission rates.
What Is the Cost-Effectiveness of Screening for HIV Infection?
In 2 good-quality studies, the cost-effectiveness of one-time HIV screening in outpatients with 1% prevalence compared to no screening was $38,000 to $42,000 per quality-adjusted life-year.243,244 One of these studies found that the cost-effectiveness improved to $15,000 per quality-adjusted life-year when secondary transmission benefits were directly incorporated into cost-effectiveness ratios, and they remained less than $50,000 per quality-adjusted life-year even when screened populations had HIV prevalences substantially lower than seen in the general population.243 The other study, which did not directly incorporate secondary transmission benefits into cost-effectiveness ratios, found that the incremental cost-effectiveness of one-time screening in the general population was greater than $100,000 per quality-adjusted life-year.244
Neither study incorporated long-term cardiovascular risks associated with HAART into their models. The study by Sanders and colleagues found that the model was sensitive to the effects of screening on secondary transmission and the benefits of early identification and therapy.
The 1996 USPSTF guidelines recommended screening persons who report high-risk behaviors.11 Neither of the 2 reviewed studies evaluated the incremental cost-effectiveness of a strategy of screening only higher-risk persons compared to broader screening strategies in different populations. One of the studies found that the incremental cost-effectiveness of testing every 5 years compared to one-time screening exceeded $50,000 per quality-adjusted life-year.243
There is no direct evidence on benefits of screening for HIV infection in the general population. Other evidence obtained for the systematic review (summarized in Table 6) indicates that testing is extremely accurate, a high proportion of patients receive a diagnosis at immunologically advanced stages of disease, and interventions (particularly HAART) are effective in reducing morbidity and mortality in patients with immunologically advanced disease. Although long-term HAART is associated with cardiovascular complications, absolute rates are low.
Reasonable screening strategies might be to screen patients with acknowledged risk factors, all patients in settings with a higher prevalence of HIV infection, or all patients in the general population. Studies that have assessed risk factor assessment to guide screening indicate that targeted screening misses a substantial proportion of HIV-positive patients. On the other hand, universal screening would result in large numbers of patients screened for each clinical outcome prevented.
An important gap in the literature is the inadequate evidence with which to accurately estimate the benefits from identification of HIV-positive patients at earlier stages of disease who do not initially qualify for HAART, particularly since screening could lead to higher rates of earlier diagnosis. In these patients, other interventions, such as counseling to reduce transmission, assume greater relative importance. Despite evidence that knowledge of HIV-positive status reduces some high-risk behaviors, there is insufficient evidence with which to accurately estimate the effects on transmission rates. The relationship between HAART use and beliefs, risky behaviors, and transmission rates also needs to be explored further. The case for screening, particularly in lower-risk populations, would be greatly strengthened by studies showing that identification at earlier stages of disease is associated with decreased transmission rates. When available, results of the SMART trial221 will provide important information about the effectiveness of HAART in asymptomatic patients with higher CD4 cell counts.
Other studies are needed on methods to improve risk assessment, effects of streamlined or targeted counseling, methods to improve entry into medical care and uptake of recommended interventions, and effects of newer testing and sampling methods. In addition, data with which to estimate the magnitude of screening harms and on methods to minimize their risk are limited. Continued attention to adverse events as patients continue receiving HAART will help clarify long-term risks.
Despite continuing HIV education efforts and the availability of effective interventions, incidence of HIV remains steady in the United States, and HIV infection continues to place an enormous burden on the health care system. Further implementation and evaluation of screening programs could have an important impact on the morbidity and mortality associated with this disease.