AIDS Behav DOI 10.1007/s10461-014-0745-8

ORIGINAL PAPER

Cost-Effectiveness of Interventions to Prevent HIV and STDs Among Women: A Randomized Controlled Trial Jennifer Prah Ruger • Arbi Ben Abdallah • Nora Y. Ng • Craig Luekens • Linda Cottler

Ó Springer Science+Business Media New York 2014

Abstract Injection drug use is a leading transmission route of HIV and STDs, and disease prevention among drug users is an important public health concern. This study assesses cost-effectiveness of behavioral interventions for reducing HIV and STDs infections among injection drug-using women. Cost-effectiveness analysis was conducted from societal and provider perspectives for randomized trial data and Bernoullian model estimates of infections averted for three increasingly intensive interventions: (1) NIDA’s standard intervention (SI); (2) SI plus a well woman exam (WWE); and (3) SI, WWE, plus four educational sessions (4ES). Trial results indicate that 4ES was cost-effective relative to WWE, which was dominated by SI, for most diseases. Model estimates, however, suggest that WWE was cost-effective relative to SI and dominated 4ES for all diseases. Trial and model results agree that WWE is

Electronic supplementary material The online version of this article (doi:10.1007/s10461-014-0745-8) contains supplementary material, which is available to authorized users. J. P. Ruger (&) University of Pennsylvania Perleman School of Medicine, 3401 Market Street, Suite 320, Philadelphia, PA 19104, USA e-mail: [email protected] A. B. Abdallah Washington University School of Medicine, St. Louis, MO 63110, USA N. Y. Ng  C. Luekens Yale University Schools of Public Health and Medicine, New Haven, CT, USA L. Cottler Department of Epidemiology, College of Public Health and Health Professions and College of Medicine, University of Florida, Gainesville, FL, USA

cost-effective relative to SI per hepatitis C infection averted ($109 308 for in trial, $6 016 in model) and per gonorrhea infection averted ($9 461 in trial, $14 044 in model). In sensitivity analysis, trial results are sensitive to 5 % change in WWE effectiveness relative to SI for hepatitis C and HIV. In the model, WWE remained cost-effective or cost-saving relative to SI for HIV prevention across a range of assumptions. WWE is cost-effective relative to SI for preventing hepatitis C and gonorrhea. WWE may have similar effects as the costlier 4ES. Keywords Cost-effectiveness  Bernoullian model  Substance abuse  HIV  STDs

Introduction Nineteen million new cases of sexually transmitted diseases (STDs) arise each year in the US; such diseases, including HIV, cost an estimated $12–20 billion in lifetime direct medical costs [5]. The World Health Organization estimated the occurrence of 340 million new curable cases of STDs in 1999, including 12 million of syphilis, 92 million of chlamydia, and 62 million of gonorrhea [35], in addition to the estimated 3 % of the world’s population infected with hepatitis C [34]. These diseases have complex interactions—an estimated 35 % of HIV-positive patients in the US and Europe also have hepatitis C, a figure perhaps as high as 50–90 % among HIV-positive injection drug users [32]. The need for policy-relevant costeffectiveness analyses of HIV and STD prevention programs is clear. Injection drug use is a leading route of HIV transmission in men and women. The Women Teaching Women intervention was designed to reduce high-risk sexual behaviors

123

AIDS Behav

and substance use among drug-using women. Our randomized controlled trial examined three Women Teaching Women interventions (n = 454) [25]: (1) a modified National Institute on Drug Abuse (NIDA) cooperative agreement standard intervention (hereafter SI), (2) SI and a field-based well woman exam (hereafter WWE), and (3) SI, WWE, and four educational sessions (hereafter 4ES). This study uses previously published cost estimates [25] to conduct cost-effectiveness analyses and cost-utility analyses for trial data and estimates derived from a Bernoullian mathematical model on preventing: HIV, total STDs, hepatitis C, syphilis, chlamydia, and gonorrhea. A companion study analyzes the cost-effectiveness of these interventions for cocaine and alcohol abuse [27]. The authors conducted this study in accordance with US Panel on Cost-Effectiveness in Health and Medicine recommendations [13]. To our knowledge, this is the first study to report incremental cost-effectiveness ratios of HIV infections averted and quality-adjusted life years saved from preventing STDs for interventions targeting drug-using women, while also uniquely comparing trial and model data. Targeting STDs is an effective way of averting HIV’s spread, and STD prevention programs’ cost-effectiveness is underestimated when reducing HIV transmission is not taken into account [7]. Integrating complex factors in the spread of HIV and STDs is a distinctive feature of this study.

(n = 144), WWE (n = 153), or 4ES (n = 157). Each intervention was conducted on separate subsamples of patients. Study participants were 90 % racial/ethnic minorities. On average, they were 39 years old, had approximately 11 years of education, and worked for about 7 months when interviewed at baseline. The number of arrests ranged from an average of 4.9 for WWE participants to 9.4 for the SI. Analysis was based on intention to treat. Intervention Conditions and Costs In SI, a peer facilitator—a female in drug recovery with at least 1 year’s sobriety—provided: (1) 20 min of HIV pretest counseling, blood collection and the NIDA SI; and (2) 2 weeks later, test results and HIV post-test counseling. The WWE intervention incorporated the SI plus an additional breast and routine pelvic examination with cervical cytological testing (Pap smear) provided by a nurse practitioner, who also obtained a short medical history. The 4ES intervention comprised of the SI, WWE, and 4ES delivered by a peer facilitator paired with a health professional, based on the Health Belief Model. The facilitator used a holistic approach emphasizing substance abuse, HIV/AIDS, health and nutrition, and stress and coping. Intervention costs are derived from our previous study developing a microcosting methodology and conducting cost analyses [25]. Effectiveness Measures

Methods Ethics Statement This study has been approved by the Institutional Review Board Human Studies Committee at Washington University Hilltop (03-20), Washington University Medical Center (04-0285) and Yale University Human Investigation Committee (27312). The randomized controlled trial in the parent study has been registered with ClinicalTrials.gov and the registration number is 01235091. Data from the parent study were analyzed anonymously; written consent was obtained for the effectiveness data and for the cost data. Recruitment, Design, and Sample Employing proven street-outreach methods, health workers recruited participants from targeted areas. Eligibility criteria included: (1) age 18 or older; (2) sexual activity in the prior 4 months; (3) cocaine, heroin, amphetamines, or other injection drug use; and (4) St. Louis area residency during the study period (2000–2006). Participants had not entered drug treatment in the past 30 days. Study participants were randomized to one of three intervention conditions: SI

123

At baseline and 12 months, we tested trial participants for HIV, hepatitis C, syphilis, chlamydia, and gonorrhea. Prevalences of infections are taken from randomized controlled trial data except for HIV; per-act transmission probabilities derive from literature (Table 1). STD prevalence was based on baseline assessment. Condom effectiveness was set at 90 % [18]. The total number of primary infections prevented by the intervention is AP = N (PP0 PPn), where N is the number of participants, PP0 is the probability of primary infection at baseline, and PPn is the probability of primary infection at 4 or 12 months [20]. The randomized trial also aimed to increase protected sex and reduce unprotected sex and the number of sexual partners and transactions among trial participants (see Supplementary Material 1). To help assess the impact of sexual behavioral change on HIV and STDs prevention, we used a Bernoullian mathematical model to convert reported sexual behavioral changes into estimates of the number of infections averted [19]. The model provides estimates at 4, 12 months, and between 4 and 12 months (see Supplementary Material 2). The probability of becoming infected for a woman who only has sex with men due to n1 acts of unprotected receptive anal intercourse, n2 acts of condom-protected receptive anal

AIDS Behav Table 1 Bernoullian model input parameters Base case

Sensitivity analysis

Source

0.02

0.008–0.032

f

0.002

0.0008–0.0032

f

0.001

0.0005–0.0015

f

0.0001

0.00005–0.00015

f

0.01

N/A

Ratio of incidence to prevalenceg

Transmission probabilities HIV Unprotected receptive anal a

Condom-protected receptive anal Unprotected receptive vaginal

a

Condom-protected receptive vaginal Hepatitis C

b

Syphilis

0.15

N/A

h

Chlamydia

0.35

N/A

h

Gonorrhea

0.22

N/A

h

HIV

0.03

0.001–0.15d

Estimatei

Hepatitis C

0.20, 0.23, 0.23

0.016

Baseline dataj

Syphilis

0.04, 0.04, 0.03

0.00016

Baseline datak [average]

Chlamydia

0.06, 0.04, 0.06

0.02

Baseline datak [average]

Gonorrhea

0.01, 0.02, 0.03

0.014

Baseline datak [average]

HIV

5.83

3.38–14.63

l

Hepatitis C

2.38

Prevalencec

Quality-adjusted life years lost

Chlamydia

N/A

m

e

N/A

h

e

0.0085

Gonorrhea

0.0085

N/A

h

Syphilis

N/A

N/A

N/A

HIV

$330,000

$150,000–400,000

n

Hepatitis C

$8,200

$7,000–25,000

o

Syphilis

$572

$286–858

p

Chlamydia

$315

$158–473

p

Gonorrhea

$343

$172–515

p

Lifetime costs

Average, where we took the average of the four zip codes with the highest number of cases for each disease, using data from the City of St. Louis Department of Health a Condom-protected HIV transmission probabilities are derived from unprotected transmission probabilities, based on the stipulated condom effectiveness of 90 % b

Hepatitis C sexual transmission probability is calculated using the method of taking the ratio of incidence to prevalence, from [30]. The St. Louis 2004 hepatitis C incidence rate is 162.8/100,000 or 0.001628; the prevalence rate in the randomized controlled trial study group is 0.2. 0.001628/0.2 is 0.008, which rounds upto 0.01 c

Prevalence for STDs for base case analysis taken from pre-treatment testing levels. Listed by arm in the following order: Standard Intervention, Well Woman Exam, 4 Educational Sessions (Table 2). This assumes that prevalence among participants’ partners is equivalent to that of participants where secondary/total infections are estimated (HIV only) d 0.15 is a rounded estimate of the Centers for Disease Control and Prevention (CDC) HIV prevalence estimate for female injection drug users in 1997 (the most recent statistic reported). It is also the CDC prevalence for injection drug users age 30–39 in 1997 e

Figures for chlamydia and gonorrhea ‘‘Quality-Adjusted Life Years Saved’’ are for men, over 12 months

f

Ref. [15, 23]

g

Ref. [8, 30]

h

Ref. [33]

i

Ref. [4, 19, 31]

j

Ref. [1]

k

Ref. [9]

l

5.83 is the average discounted (3 %) QALY saved per HIV infection averted reported by Farnham et al [12]. Ref. [20]

m

Ref. [2]

n

$330,000 is the average lifetime treatment cost reported by Farnham et al [12] and Schackman et al. [29]

o

Ref. 28

p

Ref. [5]

123

AIDS Behav Table 2 Effectiveness measures by treatment group: HIV and STDs, trial and Bernoullian model data

Treatment outcome

Standard intervention

Well woman exam

4 educational sessions

Mean

(SD)

Mean

(SD)

Mean

(0.09)

0.01

(0.12)

0

(SD)

Randomized clinical trial data HIV data HIV status, mean infectionsa Pre-treatment (n = 454)

0

End of treatment, 12 months (n = 421)

0.01

0

0

STD data, mean infectionsa STDs, total Pre-treatment (n = 327)

0.307

(0.52)

0.308

(0.52)

0.349

(0.54)

0.307

(0.52)

0.383

(0.60)

0.311

(0.56)

Pre-treatment (n = 384)

0.198

(0.40)

0.226

(0.42)

0.229

(0.42)

End of treatment, 12 months (n = 384)

0.207

(0.41)

0.234

(0.42)

0.237

(0.43)

End of treatment, 12 months (n = 327) Hepatitis C

Syphilis Data are mean Standard Intervention NIDA’s Standard Intervention, Well Woman Exam Standard Intervention plus Well Woman Exam, 4 Educational Sessions Standard Intervention, Well Woman Exam plus four educational sessions a Mean infections are calculated by dividing the number of infections by the number of participants in each trial arm b

Data based on Bernoullian model, using sexual behavior data from Supplementary Table 2 and parameters from Table 1. Standard deviations are not applicable

c Quality-adjusted life years saved per: HIV infections averted are 5.83; per hepatitis C infection averted are 2.38; per chlamydia and gonorrhea infection averted are 0.0085. Utility weights or qualityadjusted life year estimates for syphilis are unavailable

Pre-treatment (n = 379)

0.035

(0.18)

0.037

(0.12)

0.031

(0.17)

End of treatment, 12 months (n = 379)

0.070

(0.26)

0.074

(0.26)

0.047

(0.21)

Chlamydia Pre-treatment (n = 374)

0.057

(0.23)

0.038

(0.19)

0.058

(0.23)

End of treatment, 12 months (n = 374)

0.024

(0.15)

0.053

(0.22)

0.041

(0.20)

Gonorrhea Pre-treatment (n = 374)

0.008

(0.09)

0.023

(0.15)

0.025

(0.16)

End of treatment, 12 months (n = 374)

0.008

(0.09)

0.008

(0.09)

0.017

(0.13)

Bernoullian model data HIV infections averted, primaryb End of treatment, 12 months

0.353

0.178

End of treatment, 12 months

0.563

1.536

0.732

0.676

2.058

1.038

3.281

8.954

4.268

11.898

31.443

17.459

0.061

0.063

0.048

0.010

0.040

0.022

Total QALYs savedb,c HIV, primary infections End of treatment, 12 months HIV, total infections End of treatment, 12 months Hepatitis C End of treatment, 12 months Chlamydia End of treatment, 12 months Gonorrhea End of treatment, 12 months

intercourse, n3 acts of unprotected receptive vaginal intercourse, and n4 acts of condom-protected receptive vaginal intercourse with each of m partners is approximately, PP ¼ 1 fð1  pÞ þ p ð1  p1 Þn1 ð1  p2 Þn2 ð1  p3 Þn3 ð1  p4 Þn4 gm where p is the estimated infection prevalence among the participant’s partners and pk is the relevant per-act transmission probability.

123

0.116

HIV infections averted, totalb

We also expanded the Bernoullian model to include secondary infections, which the trial did not measure. Although this has typically been performed when some participants are HIV-positive to account for risk reduction among their partners, we accounted specifically for HIV infections averted among participants’ partners. Theoretically, this application can expand to subsequent levels of infection as well. Pinkerton et al. [21] adjust the original model to account for three factors: estimated overlap of

AIDS Behav

partners (k); the total number of partners (m); and the percentage of susceptible, or uninfected, partners (1-p). Thus, the probability of secondary infections is approximately, PS ¼ PP fm ð1  kÞ ð1  pÞg. We assumed that participants’ partners have the same number of partners as the participants themselves, and estimated a 25 % degree of overlap (k) [22]. The total number of secondary infections is, as above, AS ¼ N ðPS0 PSn Þ. The number of total infections averted is, AS ¼ N ðPS0 PSn Þ. We estimated incremental effects by a two-step difference-in-differences approach to compare results across trial arms. First, we calculated the difference in effectiveness from 12 months to baseline within the same intervention, for trial and model data separately. Second, differences within one intervention were compared to those within the next least costly intervention to obtain the difference in differences, or incremental effects. Since an intervention is effective when the mean decreases from baseline, we calculated incremental effects to reflect the improvement, converting the relevant change to a positive value. Incremental Cost-Effectiveness Analyses Comparing WWE to SI and 4ES to WWE, we calculated the ratio of the difference in intervention costs to the incremental effects for both trial and model results. The resulting incremental cost-effectiveness ratio shows the cost of achieving an additional unit of outcome compared to the next least costly intervention. An intervention is dominated if it is more costly and less effective than the alternative; and it is extended dominated if it is not technically dominated but has a worse marginal cost-effectiveness than its alternative [24]. Since each subsequent intervention adds significant components, an incremental cost-effectiveness analysis is appropriate. An incremental cost utility analysis based on the HIV and STD data, from both the trial and model results, revealed the cost per additional quality-adjusted life year saved. The incremental cost-utility ratio equals ðiC  iATÞ=iAQ, where iC is the incremental cost, iA the incremental infections averted, T the lifetime cost of treating a single case of infection, and Q the number of quality-adjusted life years saved by preventing an infection, the latter two obtained from the literature (Table 1). Sensitivity Analyses Costs and effects were varied separately to determine the percentage change required to relieve or reverse domination. The sensitivity of cost-utility threshold values was assessed by varying HIV model parameters. We examined bivariate and multivariate sensitivity of WWE’s modelbased HIV-related cost-utility ratio compared to SI. We

further constructed acceptability curves through bootstrapping as an alternative to confidence intervals to address uncertainty in the trial results [16].

Results Randomized Controlled Trial Results Table 2 reports effectiveness by treatment group for HIV and STDs prevention. Table 3 reports the incremental costeffectiveness analyses of the same data. At 12 months from baseline for HIV, WWE was dominated by SI, while 4ES was cost-effective relative to WWE at $94,230 per additional infection averted, and cost-saving per additional QALY saved. For STDs, with the exception of hepatitis C and gonorrhea, WWE continued to be dominated by the SI, and 4 education sessions remained cost-effective relative to WWE. $8,359 was the smallest incremental cost-effectiveness ratio (ICER) (for STDs total, 4ES), while the largest ICER was posted for hepatitis C under WWE ($109,308). Costs per additional QALY saved ran higher— over $3 million for chlamydia under 4ES. Bernoullian Model Results Table 4 reports incremental cost-effectiveness and costutility results from the Bernoullian model. Model data are directly comparable with the randomized trial data at 12 months. Model and trial results often show conflicting findings. The most prominent difference between trial and model results is that WWE in the model outperforms WWE in the trial, relative to both SI and 4ES. At 12 months in the model, WWE was cost-effective relative to SI for all diseases, and cost-saving for HIV (primary and total infections) and hepatitis C per additional QALY saved. The model also suggests that 4ES is dominated by WWE across the board for both ICERs and cost per additional QALY saved. The smallest ICER is $6,016 for hepatitis C, and the largest ICER is $208,316 for HIV primary infections, both under WWE. Cost per additional QALY saved was as high as $20 million for chlamydia under WWE. The model results are consistent with trial results, however, in that WWE was cost-effective relative to SI for hepatitis C and gonorrhea; compared to the trial, the model reports a smaller ICER for hepatitis C ($6,016 vs. $109,308) and a larger ICER for gonorrhea ($14,044 vs. $9,461). One-Way Sensitivity Analyses To evaluate sensitivity, Table 5 indicates the percentage change in the mean at which each base case ICER, excluding STD model results, achieves a switching point

123

123

–

Gonorrheaf

0.0

0.015

-0.048

-0.002

-0.075 0.001

-0.007

0.032

0.022

0.113 -0.0003

0.01

–

–

–

– –

–

$9,461

D

D

D $109,308

D

D

$29,636

$43,760

$8,359 D

$94,230

4 Educational sessions

–

–

–

– –

–

Standard intervention

$1,072,760

D

N/A

N/A $42,482

D

Well woman exam

D

$3,449,495

N/A

N/A D

CS

4 Educational sessions

Incremental cost-utility analysis (Cost per additional quality-adjusted life year saved, $)

f

e

d

c

b

a

Quality-adjusted life years saved due to averting one infection s 0.0085

Quality-adjusted life years saved due to averting one infection is 2.38

Quality-adjusted life years saved due to averting one infection is 5.83

Incremental improvement achieved when mean decreases

Positive number indicates incremental improvement. Each value is the difference of the differences from Table 2

Costs from [25]

HIV and STDs (from randomized controlled trial data)

Incremental cost-effectiveness ratio is the difference in cost divided by the difference in effectiveness as compared with the next least costly intervention, and indicates cost per additional outcome achieved; D dominated, which indicates that the intervention is more costly and less effective than the alternative; CS cost-saving

Standard Intervention, NIDA’s Standard Intervention, Well Woman Exam, Standard Intervention plus Well Woman Exam, 4 Educational Sessions, Standard Intervention, Well Woman Exam, plus four educational sessions

–

–

Chlamydiaf

– –

Syphilis

–

STDs, Total Hepatitis Ce

Well woman exam

Standard intervention

4 Educational sessions

Standard intervention

Well woman exam

Incremental cost-effectiveness ratio (Cost per additional infection averted, $)

Incremental effectsb (Difference in effects between one intervention and its next least costly comparator)

HIV Statusd

HIV and STDsc

Baseline to 12 months

Table 3 Incremental cost-effectiveness analysis

AIDS Behav

–

Chlamydia

Gonorrhea

0.237

3.518

0.287

2.892

8.212

0.973

-0.175

-2.035

-1.754

-4.134

-5.875

-0.804

–

–

–

–

–

–

1.383

0.030

0.002

N/A

19.545

5.673

-0.017

-0.015

N/A

-13.983

-4.686

-1.021

4 Educational sessions

–

–

–

–

–

$14,044

$172,303

$17,082

$6,016

$50,774

$208,316

$49,403

–

–

Well woman exam

Standard intervention

D

D

D

D

D

D

$160,189

4 Educational sessions

Incremental cost-effectiveness ratios (Cost per additional infection averted, $)

–

–

–

–

–

–

–

Standard intervention

$1,611,898

$20,233,913

N/A

CS

CS

CS

$49,403

Well woman exam

D

D

N/A

D

D

D

$160,189

4 Educational sessions

Incremental cost-utility analysis (Cost per additional quality-adjusted life year saved, $)

a

Positive number indicates incremental improvement

Standard Intervention, NIDA’s Standard Intervention, Well Woman Exam, Standard Intervention plus Well Woman Exam, 4 Educational Sessions, Standard Intervention, Well Woman Exam, plus four educational sessions

Incremental cost effectiveness ratio is the difference in cost divided by the difference in effectiveness as compared with the next least costly intervention, and indicates cost per additional outcome achieved. Cost-utility analysis indicates the cost per quality-adjusted life year; D dominated, which indicates that the intervention is more costly and less effective than the alternative; CS cost-saving; ED extended dominated, which indicates that while the intervention is not technically dominated, it has a worse marginal cost-effectiveness than its alternative

–

–

Syphilis

–

Hepatitis C

STDs

–

–

Primary

Total

HIV

Baseline to 12 months

Total incremental costs

Well woman exam

Standard intervention

4 Educational sessions

Standard intervention

Well woman exam

Incremental quality-adjusted life years (Difference in quality-adjusted life years saved between one intervention and its next least costly comparator)

Incremental effectsa (Difference in effects between one intervention and its next least costly comparator)

Table 4 Incremental cost-effectiveness analysis and cost-utility analysis: HIV and STDs (from Bernoullian model based on sexual behavior data from randomized controlled trial)

AIDS Behav

123

AIDS Behav Table 5 One-way sensitivity analyses of switching points, HIV and STD data from randomized controlled trial, and HIV model: outcomes, baseline to 12 months Switching pointsa Standard intervention

Well woman exam

4 Educational sessions

% Change

Incremental cost-effectiveness ratio at switching Point

% Change

?5

$288,900

NP

$86,658

Incremental cost-effectiveness ratio at switching Point

STDs, from RCT data HIV

–

STDs, Total

–

?20

Hepatitis C

–

-5

Syphilis

–

?5

$99,675 $475,180

Chlamydia

–

?90

Gonorrhea

–

-200

HIV, from Bernoullian model

-40 ?5

$81,956

-50 -80 ?45

$1,664,503

b

Primary infections

–

-95

?45

$28,910,128

Total infections

–

-150

?85

$1,461,003

If the intervention was dominated, we report the % increase of the mean that is required to make it no longer dominated, and the corresponding incremental cost-effectiveness ratio. If the intervention was not dominated, the % reduction of the mean needed to make it dominated is reported NP not possible, which indicates that a switching point cannot be obtained because it would require the mean to become less than zero, NA not available, which indicates that the base case mean is zero, so a relative percentage change is not possible a

Analyses were performed by varying the original mean value at 5 % increments A negative percentage change indicates a positive change in the probability of becoming infected, which is the outcome of the Bernoullian model given at each endpoint

b

(where a dominated intervention becomes cost-effective or vice versa). STD trial data showed varying sensitivity. For trial results at 12 months, both WWE and 4ES proved highly sensitive to small effectiveness changes in preventing hepatitis C (5 %). WWE is also quite sensitive for HIV and syphilis (5 %), but very robust for chlamydia (90 %) and gonorrhea (200 %). Except for hepatitis C, 4ES results are robust for other diseases, requiring at least a 40 % change to cause or overcome domination. Model HIV results require large changes in effectiveness to ca domination for WWE (at least 95 %) and to relieve domination for 4ES (at least 45 %)—all ICERs at switching points exceed $1 million.

acceptability curves, which display the probability of costeffectiveness for interventions over a range of willingnessto-pay values (the maximum amount payer is willing to pay per additional infection averted). Compared to WWE in preventing total STDs, 4ES has a 0.80 probability at $30,000, versus WWE’s less than 0.20 probability relative to SI. For preventing hepatitis C, the probabilities of being cost-effective are considerably higher for WWE than 4ES (0.44 vs. 0.10 at $10,000), but at $20,000, the two curves begin to report increasingly similar probabilities.

Bivariate and Multivariate Sensitivity Analyses

Our cost-effectiveness analysis evaluates HIV and STDs in a randomized controlled trial targeting injection drug-using women, and compares trial data with Bernoullian model estimates at 12 months. Both trial and model data show WWE to be cost-effective relative to SI for hepatitis C and gonorrhea, but overall the differences between trial and model results are striking. In the trial, 4ES was costeffective relative to WWE, which was dominated by SI, for most diseases. In the model, WWE was cost-effective relative to SI and dominated 4ES for all diseases. Why were model results so different from trial results? Model results were driven by the following factors: disease prevalence; condom effectiveness (stipulated at 90 %); per act transmission probabilities; number of sex acts; and

Table 6 reports bivariate and multivariate sensitivity for WWE relative to SI in preventing HIV at 12 months, as estimated by the Bernoullian model. WWE is cost-effective or cost-saving relative to SI in all model scenarios based on the sensitivity analysis ranges presented in Table 1. Above HIV prevalence of 1 %, WWE is almost always cost-saving. Acceptability Curves Figure 1 presents the uncertainty of the randomized controlled trial results for total STDs and hepatitis C with

123

Discussion

AIDS Behav Table 6 Bivariate and multivariate sensitivity analyses, comparing well woman exam with standard intervention Incremental cost-utility ratio, $ per additional quality-adjusted life year saved Baseline to 12 months Unprotected anal transmission probability

Unprotected vaginal transmission probability

Lifetime treatment costs

Quality-adjusted life years saved

Worst case

Best case

p

Base case

0.008

0.032

0.0005

0.0015

$150,000

$400,000

3.38

14.63

0.001, Primary

1,012,873

1,416,083

782,764

1,393,673

793,840

1,043,748

1,000,867

1,747,057

403,626

1,089,167

917,094

0.001, Total

198,351

292,286

144,174

291,174

145,388

229,226

186,345

342,127

79,042

169,419

210,626

0.005, Primary

157,360

238,011

111,332

233,511

113,558

188,235

145,353

271,423

62,707

123,276

175,244

0.005, Total

CS

13,405

CS

13,177

CS

25,433

CS

CS

CS

CS

34,064

0.01, Primary

50,421

90,752

27,403

88,491

28,523

81,295

38,414

86,968

20,092

2,539

82,513

0.01, Total

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

11,995

0.03, Primary

CS

CS

CS

CS

CS

10,003

CS

CS

CS

CS

20,693

0.03, Total

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

0.05, Primary 0.05, Total

CS CS

CS CS

CS CS

CS CS

CS CS

CS CS

CS CS

CS CS

CS CS

CS CS

8,328 CS

0.10, Primary

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

0.10, Total

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

0.15, Primary

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

0.15, Total

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

CS

‘‘Worst case’’ uses high transmission probabilities, high lifetime treatment costs, and low quality-adjusted life years saved ‘‘Best case’’ uses low transmission probabilities, low lifetime treatment costs, and high quality-adjusted life years saved HIV (from Bernoullian model based on sexual behavior data) CS cost saving a

Varies the well woman exam and standard intervention values simultaneously

Fig. 1 Acceptability curves. Abbreviations: SI standard intervention, WWE well woman exam, 4ES 4 educational sessions

number of partners [lifetime medical costs and QALYs lost were used for ICUA with both trial and model data]. With the exception of HIV prevalence, condom effectiveness and per act transmission probabilities, all other model inputs were based on randomized trial data. The model, then, was substantially driven by trial data, yet showed very different results. One potential explanation is that the

number of sex partners as a model input has greater influence over model estimates than number of sex acts, per act transmission probabilities, and disease prevalence, and WWE outperformed both SI and 4ES in reducing the number of sex partners at 12 months (Supplementary Material 1, reporting sexual behavior outcomes). This is supported by the improved performance of 4ES relative to WWE between 4 and 12 months, during which 4ES outperformed WWE in reducing the number of sex partners (Supplementary Material 2, reporting 4–12 months results). Discrepancies between model and trial results may be a consequence of the model inadequately reflecting the reality of study participants in its assumptions and/or its weighing of inputs, or of the model failing to include all relevant factors at work in real life. Another possibility may be that the differences between 12 months and baseline across arms in the trial data were small, such that the more costly interventions are dominated by their cheaper alternatives—it might be that, if outcomes were measured at additional and different timeframes, there could be more convergence between trial and model results. The sensitivity of WWE trial results for HIV and the small ICERs

123

AIDS Behav

for reducing the number of partners, preventing sexual transactions, and reducing unprotected anal and vaginal intercourse (not shown) indicate that the trial results are likely to undervalue the benefits WWE achieves. Our results may be considered next to [6] incremental cost-effectiveness analysis that compared a more comprehensive educational intervention designed to prevent HIV and STDs to a condom skills-only intervention (also a component of the comprehensive intervention), determining through model estimates based on behavior change an incremental cost per HIV case prevented exceeding $ million and the incremental cost per quality-adjusted life year to be $124,562. Our study’s 4ES intervention shares similarities with the more comprehensive intervention above, and relative to the WWE, the trial reports an incremental cost per HIV infection averted of $94,230; 4ES was cost-saving per QALY saved (Table 3). In our model, however, 4ES was dominated by WWE. This is actually in line with the conclusion reached by others [6], who noted that most of the measured effects of the comprehensive education intervention were attributable to the condom skills component, and that despite the apparent costeffectiveness of the comprehensive intervention, its additional benefits may not be worth the extra costs. Other studies show that shorter and simpler interventions performed at least as well as longer, more comprehensive ones in preventing STDs, even for high-risk populations [3, 14]. An intervention’s impact on HIV and STDs also affects co-infection. HIV co-infection with hepatitis C is well known [11], but HIV co-infection with syphilis, gonorrhea, and chlamydia is also a growing concern [10, 17]. Analyzing the prevention of only one kind of co-infection thus provides an incomplete evaluation. Our trial and model results at 12 months agree that WWE is cost-effective for preventing hepatitis C and gonorrhea, relative to SI. Model data further suggest that WWE is cost-effective for preventing HIV and all STDs at 12 months. Moreover, accounting for HIV infections among partners—a clear implication of preventing primary infections—expands this study’s relevance. Greater standardization and comparability among economic evaluations in health and medicine, especially among vulnerable populations such as pregnant women will enhance the impact of this work in policy analysis and clinical practice [26]. We note several limitations in the statistical analyses: (1) the Bernoullian model predicting infection rates based on sexual conduct assumes independence; (2) nesting of overlapping partners was not considered; (3) the number of partners used in predicting secondary infections is based on participants’ number of partners. Our study has additional limitations. First, its location in a single urban center limits its generalizability. Second, data constraints prevented incorporating future costs, except for those from the

123

literature included in the Bernoullian model (e.g. lifetime HIV and STD treatment costs), suggesting an underestimation of the intervention’s results. Third, by focusing on HIV and STD prevention and substance abuse, the intervention incorporated significant interactions, but might have sacrificed the clarity possible in studying HIV and/or STDs exclusively. Fourth, the model results were highly sensitive to input parameters, especially prevalence and transmission probability. Fifth, baseline assessment determined STD prevalence, which may underestimate the prevalence among partners because the participants were all HIV-negative, suggesting they have fewer STDs than their peers. Sixth, quality-adjusted life year measures and their estimation have many disadvantages when used for health policy evaluation. This study adds to the literature on cost-effectiveness for preventing HIV and STDs among drug-using women and targets complex interactions of activities and diseases, providing a unique comparison between testing data and Bernoullian model estimates. Acknowledgments This work was supported by the National Institutes of Health (NIH) [National Institute on Drug Abuse (NIDA) Grant R01DA11622, and K01DA01635810 to J.P.R.], and the Patrick and Catherine Weldon Donaghue Medical Research Foundation (Grant DF06-112 to J.P.R.).

References 1. Armstrong GL, Wasley A, Simard EP, McQuillan GM, Kuhnert WL, Alter MJ. The prevalence of hepatitis C virus infection in the United States, 1999 through 2002. Ann Intern Med. 2006;144: 705–14. 2. Australian Commonwealth Department of Health and Ageing. 2002. Return on investment in needle and syringe programs in Australia. Available at http://www.health.gov.au/internet/main/ publishing.nsf/Content/2C9410E4A867F2EBCA257650000077B9/ $File/roirep.pdf. Accessed 22 Feb 2012. 3. Branson BM, Peterman TA, Cannon RO, Ransom R, Zaidi AA. Group counseling to prevent sexually transmitted disease and HIV: a randomized controlled trial. Sex Transm Dis. 1998;25: 553–60. 4. Centers for Disease Control and Prevention. 2007. HIV prevalence among selected populations: high-risk populations. Available at http://www.cdc.gov/hiv/topics/testing/resources/reports/ hiv_prevalence/high-risk.htm. Accessed 22 Feb 2012. 5. Chesson HW, Collins D, Koski K. Formulas for estimating the costs averted by sexually transmitted infection (STI) prevention programs in the United States. Cost Eff Res Alloc. 2008;6:10. 6. Chesson HW, Greenberg JB, Hennessy M. The cost-effectiveness of the WINGS intervention: a program to prevent HIV and sexually transmitted diseases among high-risk urban women. BMC Infect Dis. 2002;2:24. 7. Chesson HW, Pinkerton SD. Sexually transmitted diseases and the increased risk for HIV transmission: implications for costeffectiveness analyses of sexually transmitted disease prevention interventions. J Acquir Immune Defic Syndr. 2000;24:48–56.

AIDS Behav 8. City of St. Louis Department of Health. 2005. Program monitoring. Available at http://stlouis.missouri.org/citygov/health/pdf/ cd/CDDec2005.pdf Accessed 22 Sep 2010. 9. City of St. Louis Department of Health. 2007. Public Health: Understanding Our Needs. St. Louis. Available at http://stlouis-mo. gov/government/departments/health/documents/upload//UON_ VOL_3_20072.pdf.Accessed 22 Feb 2012. 10. Diaz A, Junquera ML, Esteban V, Martı´nez B, Pueyo I, Suarez J, STI Study Group and EPI-VIH Group. HIV/STI co-infection among men who have sex with men in Spain. Euro Surveill. 2009;14, pii 19426. 11. Dionne-Odom J, Osborn MK, Radziewicz H, Grakoui A, Workowski K. Acute hepatitis C and HIV coinfection. Lancet Infect Dis. 2009;9:775–83. 12. Farnham PG, Holtgrave DR, Gopalappa C, Hutchinson AB, Sansom SL. Lifetime costs and quality-adjusted life years saved from HIV prevention in the test and treat era. J Acquir Immune Defic Syndr. 2013;64:e15–8. 13. Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-Effectiveness in Health and Medicine. New York: Oxford University Press; 1996. 14. Kamb ML, Fishbein M, Douglas JM Jr, Rhodes F, Rogers J, Bolan G, Peterman TA. Efficacy of risk-reduction counseling to prevention human immunodeficiency virus and sexually transmitted diseases: a randomized controlled trial. Project RESPECT study group. J Am Med Assoc. 1998;280:1161–7. 15. Katz MH, Gerberding JL. Postexposure treatment of people exposed to the human immunodeficiency virus through sexual contact or injection-drug use. N Engl J Med. 1997;336:1097–100. 16. Lothgren M, Zethraeus N. Definition, interpretation and calculation of cost-effectiveness acceptability curves. Health Econ. 2000;9:623–30. 17. Onen NF, Shacham E, Stamm KE, Overton ET. Comparisons of sexual behaviors and STD prevalence among older and younger individuals with HIV infection. AIDS Care. 2010;22:711–7. 18. Pinkerton SD, Abramson PR. Effectiveness of condoms in preventing HIV transmission. Soc Sci Med. 1997;44:1303–12. 19. Pinkerton SD, Abramson PR. The Bernoulli-process model of HIV transmission: applications and implications. In: Holtgrave DR, editor. Handbook of economic evaluation of HIV prevention programs. New York: Plenum Press; 1998. p. 13–32. 20. Pinkerton SD, Holtgrave DR. A method for evaluating the economic efficiency of HIV behavioral risk reduction interventions. AIDS Behav. 1998;2:189–201. 21. Pinkterton SD, Holtgrave DR, Leviton LC, Wagstaff DA, Abramsom PR. Model-based evaluation of HIV prevention interventions. Eval Rev. 1998;22:155–74. 22. Pinkterton SD, Johnson-Masotti AP, Otto-Salaj LL, Stevenson LY, Hoffman RG. Cost-effectiveness of an HIV prevention

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

intervention for mentally ill adults. Mental Health Serv Res. 2001;3:45–55. Pinkerton SD, Martin JN, Roland ME, Katz MH, Coates TJ, Kahn JO. Cost-effectiveness of postexposure prophylaxis after sexual or injection-drug exposure to human immunodeficiency virus. Arch Intern Med. 2004;164:46–54. Resch S, Altice FL, Paltiel AD. Cost-effectiveness of HIV screening for incarcerated pregnant women. J Acquir Immune Defic Syndr. 2005;38:163–73. Ruger JP, Abdallah AB, Cottler L. Costs of HIV prevention among out-of-treatment drug-using women: results of a randomized controlled trial. Public Health Rep. 2010;125(suppl 1):83–94. Ruger JP, Lazar CM. Economic evaluation of drug abuse treatment and HIV prevention programs in pregnant women: a systematic review. Addict Behav. 2012;37(1):1–10. Ruger JP, Abdallah AB, Luekens C, Cottler L. Cost-effectiveness of peer-delivered interventions for cocaine and alcohol abuse among women: a randomized controlled trial. PLoS One. 2012;7:e33594. Salomon JA, Weinstein MC, Hammitt JK, Goldie SJ. Costeffectiveness of treatment for chronic hepatitis C infection in an evolving patient population. J Am Med Assoc. 2003;290:228–37. Schackman BR, Gebo KA, Walensky RP, Losina E, Muccio T, Sax PE, Freedberg KA. The lifetime cost of current human immunodeficiency virus care in the United States. Med Care. 2006;44:990–7. Singer ME, Younossi ZM. Cost effectiveness of screening for hepatitis C virus in asymptomatic, average-risk adults. Am J Med. 2001;111:614–21. Tempalski B, Lieb S, Cleland CM, Cooper H, Brady JE, Friedman SR. HIV prevalence rates among injection drug users in 96 large US metropolitan areas, 1992–2002. J Urban Health. 2009;86:132–54. Verucchi G, Calza L, Manfredi R, Chiodo F. Human immunodeficiency virus and hepatitis C virus coinfection: epidemiology, natural history, therapeutic options and clinical management. Infection. 2004;32:33–46. Wilson DP, Heymer KJ, Anderson J, O’Connor J, Harcourt C, Donovan B. Sex workers can be screened too often: a costeffectiveness analysis in Victoria, Australia. Sex Transm Infec. 2010;86:117–25. World Health Organization. Global prevalence and incidence of selected curable sexually transmitted infections: overview and estimates. Geneva: World Health Organization Department of HIV/AIDS; 2001. World Health Organization. Global Alert and Response. Geneva: World Health Organization. Hepatitis C; 2002.

123