Tuesday, October 26, 2010: 1:00 PM
Grand Ballroom Centre (Sheraton Centre Toronto Hotel)

* Candidate for the Lee B. Lusted Student Prize Competition

Session Chairs:
Uwe Siebert, MD, MPH, MSc, SD and A. David Paltiel, PhD
1:00 PM
Karen Schneider1, Thanyawee Puthanakit2, June Ohata2, Stephen Kerr3, Matthew Law1, David Cooper1, Basil Donovan1, Nittaya Phanuphak4, Jintanat Ananworanich5 and David Wilson1, (1)National Centre in HIV Epidemiology and Clinical Research, Sydney, Australia, (2)The HIV Netherlands Australia Thailand Research Collaboration, Bangkok, Thailand, (3)National Centre in HIV Epidemiology & Clinical Research, Sydney, Australia, (4)The Thai Red Cross Aids Research Centre, Bangkok, Thailand, (5)South East Asia Research Collaboration with Hawaii, Bangkok, Thailand

Purpose:    This study aimed to assess the cost-effectiveness and -utility of different frequencies of monitoring viral load of HIV-positive children initiating antiretroviral treatment (ART) in a resource-limited setting.

Method:    An agent-based simulation model of virological and immunological outcomes of HIV-infected children on ART was built and directly informed by a longitudinal cohort study of 304 HIV-infected children starting ART in Thailand between October 2001 and May 2009. Rates of virological failure, suppression, and rebound, as well as CD4 progression, on first- and second-line ART were captured in the cohort and represented by the model. The model simulated expected clinical outcomes of CD4 percentage (CD4%) and viral load over time among children on ART according to different frequencies of viral load monitoring and initiation of second-line therapies where appropriate. Cost-utility, expressed as cost per quality adjusted life-years (QALYs) saved, and cost-effectiveness, expressed as cost per year of virological failure averted was calculated across 11 monitoring frequencies.  

Result:      Compared with the status quo, of no viral load monitoring, all frequencies of monitoring and access to second-line ART led to significant reductions in the number of children failing ART after 10 years. A single screening during the first year of ART led to an estimated 43.9% reduction in ART failures after 10 years, with repeated viral load monitoring leading to an average 75.1% reduction. The cost per year of virological failure averted, including antiretroviral drug costs, ranged from US$2,332 (IQR: US$2,311−US$2,365) for a one-off screening 30 weeks after the initiation of ART to US$4,098 (IQR: US$4,057−US$4,134) for screening children every 6 months.    Most individuals initiated first-line treatment with a low CD4% and recovered immunologically regardless of monitoring frequency; therefore only a modest gain in QALYs was observed. The incremental cost per QALY gained, attributed to monitoring alone, ranged from US$1,109 (IQR: US$859−US$1,557) for a one-off screening during the first year of ART to US$15,817 (IQR: US$13,961−US$19,855) for screening children every 6 months.

Conclusion:    Even very infrequent viral load monitoring is likely to provide substantial clinical benefit to HIV-infected children on ART and be moderately cost-effective. Without access to second-line drugs and regular viral load monitoring the current efficiency of first-line therapies is likely to be compromised, ultimately leading to a reduction in future drug options for Thailand’s population.

1:15 PM
April D. Kimmel, PhD, MSc1, Stephen C. Resch, PhD, MPH2, Xavier Anglaret, MD, PhD3, Norman Daniels, PhD2, Sue J. Goldie, MD, MPH2, Christine Danel, MD, PhD4, Kenneth A. Freedberg, MD, MSc5 and Milton C. Weinstein, PhD2, (1)Weill Cornell Medical College, New York, NY, (2)Harvard School of Public Health, Boston, MA, (3)Inserm U897, Bordeaux, France, (4)PACCI Program, Abidjan, Ivory Coast, (5)Massachusetts General Hospital, Boston, MA

Purpose: In resource-limited settings, increasing numbers of HIV-infected individuals are initiating ART and remaining in care longer. Many HIV budgets, however, are flattening or decreasing. By modeling a policy of discontinuing ART after treatment failure, we aimed to highlight trade-offs among competing policy goals of optimizing individual health outcomes, population health outcomes, and the number receiving treatment.

Method: We assessed three HIV treatment strategies: (1) no ART; (2) never discontinue ART (Status Quo); and (3) discontinue ART after failure (Alternative). We used a state-transition model (CEPAC-International) to simulate annual probabilities of survival and receiving ART for treatment-eligible, HIV-infected individuals in the absence of treatment constraints. These estimates were then fed into a population-level linear programming model that included constraints on treatment capacity. For simplicity, we assumed that incidence of new patients and treatment capacity were constant over time. Data were derived from clinical trials and cohort studies conducted in Côte d’Ivoire, West Africa. Treated individuals received two sequential ART regimens; switching to 2nd-line ART and discontinuation of 2nd-line ART (Alternative strategy only) occurred upon detection of antiretroviral failure, defined as a 50% decrease in peak CD4 count. Individuals receiving a failed ART regimen continued to experience some treatment benefit, including decreased risk of AIDS-related mortality. At the population level, we assumed an analytic time horizon of 5 years and the number of treatment-eligible cases (100,000/year) exceeded treatment capacity (25,000/year).

Result: At the population level, including treated and untreated individuals, the Alternative strategy increased total life-years by 10,000 (+0.8%) to 1.23 million compared to the Status Quo strategy. The Alternative strategy increased the average number initiating ART annually by 180 individuals (+13.3%) to 1,530 compared to the Status Quo. Although more individuals received treatment under the Alternative strategy, life expectancy for treated individuals decreased by 0.8 years (-4.6%) to 16.4 years compared to the Status Quo. Among patients receiving ART over the 5-year period, 20.7% died under the Alternative strategy compared to 18.7% under the Status Quo. Results were sensitive to the timing of detection of ART failure, number of ART regimens, and level of treatment capacity.

Conclusion: With limited HIV treatment resources, trade-offs emerge between maximizing health outcomes for individual patients receiving treatment and maximizing health outcomes and access to treatment at the population level.

1:30 PM
Julie Levison, MD, MPHIL1, Robin Wood, FCP, MMed, DTM&H2, Callie Scott, MSc3, Andrea Ciaranello, MD, MPH3, Elena Losina, PhD1, Kenneth A. Freedberg, MD, MSc3 and Rochelle P. Walensky, MD, MPH3, (1)Massachusetts General Hospital and Brigham and Women's Hospitals, Boston, MA, (2)University of Cape Town, Cape Town, South Africa, (3)Massachusetts General Hospital, Boston, MA

Purpose:   In resource-limited settings, antiretroviral treatment (ART) options are limited. At 1st-line ART failure, genotype drug resistance testing can identify patients with non-resistant (wild-type) virus who have failed due to poor medication adherence. These patients may safely and effectively continue 1st-line ART, avoiding premature switches to costlier 2nd-line ART.

Methods:   We used a state-transition, Monte Carlo simulation of HIV disease and treatment (the CEPAC-International model) to project per-person life expectancy (LE) and mean lifetime HIV care costs (2006 US$) for no genotype vs. genotype strategies at 1st-line ART failure. In the no genotype strategy, all patients switched to 2nd-line ART at diagnosis of 1st-line ART failure. In the genotype strategy, patients with wild-type virus remained on 1st-line ART with an adherence intervention, and those with resistant virus switched to 2nd-line ART. Model inputs were derived using clinical and cost data from South Africa (mean age 32.8y, mean CD4 307/µl, genotype cost $300/test, 1st-line ART $7/month, 2nd-line ART $63/month); 20% of patients had wild-type virus at 1st-line ART failure. A strategy was considered “very cost-effective” if  the incremental cost-effectiveness ratio (ICER) was below the 2006 South African per capita gross domestic product ($5,400/year of life saved [YLS]). In sensitivity analyses, we examined the impact of variations in prevalence of drug resistance at 1st-line ART failure, ART efficacies, and both genotype and ART costs.

Results: At 1st-line ART failure, projected LE with no genotype was 152.4 months and increased to 155.9 months with genotype. Per-person lifetime costs were $11,690 and $11,740, respectively.  Compared to no genotype, the ICER for genotype was $160/YLS. Genotype was very cost-effective under plausible variations in efficacies of 1st- and 2nd-line ART. In a 2-way sensitivity analysis, genotype was very cost-effective at the base case test cost ($300), if prevalence of wild-type virus remained >2% (Figure, circle). At the base case prevalence of wild-type virus (20%), genotype was very cost-effective at test costs <$2,600 (Figure, square); at 20% wild-type prevalence, results were not sensitive to the genotype cost since this was offset by the savings in 2nd-line ART costs.

[Figure. Incremental cost-effectiveness ratio (ICER, 2006 USD/YLS) of genotype vs. no genotype strategy].  

Conclusion: In South Africa, drug resistance testing at 1st-line ART failure increases the survival benefits of HIV treatment and is very cost-effective.  

1:45 PM
Robine Hofman, MSc, Esther W. de Bekker-Grob, PhD, Hein Raat, MD, PhD, Theo J.M. Helmerhorst, MD, PhD, Marjolein van Ballegooijen, MD, PhD and Ida J. Korfage, PhD, Erasmus MC - University Medical Center Rotterdam, Rotterdam, Netherlands

Purpose: In 2009 the Human Papillomavirus (HPV) vaccine against cervical cancer has been added to the Dutch National Immunization Program for (pre)adolescent girls. Parental consent is not obligatory. In practice parents and girls tend to make shared decisions and preferences for characteristics of the HPV vaccine of both parties will influence uptake. To date similarities and differences between girls’ and parental preferences are unknown. We assessed how characteristics of the HPV vaccine determined uptake intentions in girls and parents, and compared preferences.   

Method: A discrete choice experiment (DCE) was conducted among a sample of girls aged 11-16 and among another sample of parents with a daughter aged 10-12. Respondents were asked to choose between hypothetical HPV vaccination programs comprising different levels of four characteristics: degree of protection, protection duration, risk of serious side-effects, and age at which vaccination is offered. A multinomial logit regression model was used to analyze the results.   

Result: Response rates were 325/359 (91%) for girls and 307/983 (31%) for parents. In both groups all vaccine characteristics proved to influence preferences for HPV vaccination (p<0.05). Serious side-effects were evaluated more negatively by parents than by girls (p<0.01), while parents had a more positive attitude towards HPV vaccinations at a later age than girls (p<0.05). Girls were willing to trade-off 12% and parents 10% of the protection against cervical cancer to obtain life-time protection instead of 25 years. To get an HPV vaccination at age 12 years instead of 9 years, girls were willing to trade-off 6% and parents 10% of the protection against cervical cancer.   

Conclusion: This study showed that parents and even girls were able to make trade-offs using numerical rates. Their trade-offs between the degree of protection against cervical cancer and other characteristics of HPV vaccinations were similar. However, girls evaluated serious side-effects less negatively than parents. Girls may possibly be less able to imagine the seriousness of such side-effects. Parents favored vaccination at a later age than girls did, possibly because parents underestimate their daughter’s age of sexual initiation. Since prevalence of HPV is highest shortly after sexual debut and protection is reduced after HPV infection, parents should be made aware that vaccination is less useful once their daughter has been infected with HPV.

2:00 PM
Nick Bansback1, John Brazier, PhD2, Aki Tsuchiya, PhD1 and Aslam H. Anis, PhD3, (1)University of Sheffield, Sheffield, United Kingdom, (2)School of Health and Related Research, Sheffield, United Kingdom, (3)University of British Columbia, Vancouver, BC, Canada

Purpose: There is concern that the tasks involved in the standard gamble and time trade off (TTO) are too complex for certain populations. The objective of this study is to explore a novel application of the Discrete Choice Experiment (DCE) that resembles the Time Trade Off (TTO) task to estimate values on the full health – dead scale for the EQ-5D. 

Method: The DCE is tested in a web survey alongside the TTO by inviting 4189 members of a market research panel in Canada to participate in an internet based survey. The DCE was designed to closely resemble the TTO task describing two health states each with a given survival. The DCE requires respondents to answer which health state they prefer, not the degree they prefer a health state as is required in the TTO. Experimental design theory was used to develop 144 choice sets which enable the interaction between survival and each EQ-5D attribute to be estimated.  Conditional and mixed logit models were used to analyse DCE results, and values were converted to the full health – dead scale. The TTO design and analysis followed existing EQ-5D valuation studies. Values from the two techniques are compared along with other issues such as rates of completion, failure to engage and understand the task and self reported ease of understanding and answering.

Results: Of the 1400 (33%) that responded to the survey, their socio-demographic profiles were largely representative of the Canadian general population. Over 13% of respondents failed to complete the 5 TTO tasks compared to 4% for the 8 DCE tasks. Over a third of the completers of the TTO appeared to fail to understand or engage with the task. The inclusion of these values in final models dramatically influenced average values. In contrast, the DCE results were robust to all completed values.

Conclusion: The study finds that the new DCE is able to derive logical and consistent values for health states valued on the full health – dead scale. The DCE overcame some issues identified in the version of TTO currently used to value EQ-5D, notably whether to exclude respondents who fail to understand the task and incorporating values considered worse than dead without transformation. This has important implications for providing values that represent the preferences of all respondents.

2:15 PM
Rebecca L. Hancock-Howard, PhD1, Deborah Marshall, PhD2, Wendy J. Ungar, PhD3, Adrienne Einarson, RN4, Michael Goodstadt, PhD5 and Gideon Koren, MD4, (1)i3 Innovus, Toronto, ON, Canada, (2)University of Calgary, Calgary, AB, Canada, (3)The Hospital for Sick Children, Toronto, ON, Canada, (4)Hospital for Sick Children, Toronto, ON, Canada, (5)University of Toronto, Toronto, ON, Canada

Purpose: Making decisions about using medications during pregnancy, particularly anti-depressants, requires appropriate information. It is not known what format is preferred for receiving counseling and information services on medication use during pregnancy. Two common sources are family doctors and teratology information services (TIS, which operate specialized telephone hotlines).The purpose was to determine public preferences for attributes of teratology counseling on anti-depressant use during pregnancy.

Method: We designed a survey to assess preferences for counseling and administered it to a community sample. Attributes and levels were generated in focus groups. Based on attribute ratings and analysis of discussion themes, five attributes (training of information provider (IP), method of counseling and waiting time, knowing the IP, confidence in the IP and helpfulness of information), with two levels each and a cost attribute with three levels were selected. Sawtooth® was used to test and select a fractional factorial design that maximized D-efficiency and achieved orthogonality, minimal level overlap, and level balance. Three versions were generated, each with 12 random choice pairs and three fixed choice pairs (for testing rationality and consistency). Pilot testing showed that the survey was robust. Logistic regression was conducted in SAS to estimate utilities and significance.

Result: 175/203 volunteers completed the survey (85% women). All beta estimates were statistically significant. The most important attribute was receiving very helpful information (beta 0.72, 95% CI 0.52-0.92). Features of TIS were generally preferred over features of an encounter with a doctor, i.e. counseled via telephone (beta 0.23, 95% CI 0.07-0.39) by a non-physician specialist in medication use during pregnancy (beta 0.13, 95% CI 0.01-0.26). It was preferred, however, to speak with a provider who was known to the user (beta 0.53, 95% CI 0.41-0.65). Willingness to pay was affected by demographic factors, including education level and level of anxiety about being exposed to anti-depressants.  

Conclusion: Given the preferred features of information service delivery found in this survey, counseling by a TIS likely offers the most benefit. The public valued high quality information and preferred receiving information in non-traditional formats (i.e. over the telephone with a non-physician), however, some sort of relationship with the HCP is still important. Preference information can be used to inform service design, and WTP values may be used in a future cost-benefit analysis.