Is Increased Screening and Early Antiretroviral Treatment for HIV-1 Worth the Investment?

An Analysis of the Public Health and Economic Impact of Improvement in the UK

AJ Brogan; SE Talbird; AE Davis; L Wild; D Flanagan

Disclosures

HIV Medicine. 2019;20(10):668-680. 

In This Article

Methods

Model Structure

A Markov model was developed in Microsoft Excel to follow theoretical cohorts of MSM, heterosexuals, and PWID with initially undiagnosed HIV-1 infection in the UK over their remaining lifetimes. The model took a payer perspective and had two arms to assess the impact of increased versus current screening efforts, both with early initiation of antiretroviral treatment following diagnosis. The base-case analysis focused on screening in SHS, where most new HIV diagnoses occur, and sensitivity analysis examined other screening settings. Each cohort entered the model with individuals distributed across three viral load ranges (HIV-1 RNA < 10 000, 10 000–99 999, and ≥ 100 000 copies/mL) and six health states defined by CD4 cell count ranges (0–50, 51–100, 101–200, 201–350, 351–500, and > 500 cells/μL). In each 3-month cycle, individuals could remain in or transition between these health states. Individuals could also move along the care pathway by receiving an HIV diagnosis or initiating treatment (Fig. 1). In each cycle, individuals were also at risk of death from causes related or unrelated to HIV-1 infection.

Figure 1.

Model structure diagram. In each 3-month cycle, individuals with HIV-1 infection in the model could remain in or transition to the adjacent CD4 health state. Individuals could also move along the care pathway by receiving an HIV diagnosis or initiating treatment. In each cycle, individuals were also at risk of transmitting HIV-1 to their partners and were at risk of death from causes related or unrelated to HIV-1 infection. ART, antiretroviral therapy. CD4 counts are in cells/μL.

Prior to treatment initiation, all individuals with diagnosed or undiagnosed HIV-1 infection experienced steadily declining CD4 cell count, consistent with natural disease progression patterns. Following treatment initiation, individuals could progress through up to four lines of antiretroviral treatment with durations and CD4 cell count changes appropriate for each line. In the fourth, nonsuppressive line of therapy, individuals experienced a steady decline in CD4 cell count until death.

As individuals progressed along the care pathway, they incurred costs for HIV screening and HIV-related clinical management. Utility values, which quantified patient preferences for remaining in each of the health states on a scale from 0 (worst possible health or death) to 1 (perfect health), were used to convert time spent in these health states into estimates of quality-adjusted life-years (QALYs).

Individuals in the model, in addition to experiencing progression in their own health, were at risk of transmitting HIV-1 infection to their partners (Fig. 1). For each subgroup (i.e. MSM, heterosexuals, and PWID), a simple Bernoulli transmission model[12] estimated the average number of secondary infections per 3-month cycle. Parameters of the Bernoulli model included the proportion of partners without HIV-1 infection, number of long-term and one-time partners, number of condomless interactions (or needle-sharing events) per partner, infectivity (i.e. probability of HIV transmission per condomless interaction or needle-sharing event), duration of long-term partnerships, and total duration of risk activity. This Bernoulli model allowed for a simple assessment of the impact of increased screening and early treatment on secondary infections by accounting for potential changes in behavioural patterns upon diagnosis and potential reductions in infectivity after initiation of suppressive antiretroviral therapy. The model also applied estimated lifetime costs and QALY losses to each secondary infection, further allowing for a health-economic assessment of the impact of averted onward infections. The model is not a full dynamic epidemiological model, however, and does not account for tertiary or further onward transmissions.

Input Parameters

Modelled Population. The model followed theoretical cohorts of MSM, heterosexuals, and PWID with initially undiagnosed HIV-1 infection in the UK over their remaining lifetimes. Because diagnosis is the first opportunity at which to collect data for this population, characteristics of the cohorts entering the model were based on subgroup-specific surveillance data for individuals with new diagnoses in the UK.[13] The mean age, the percentage male, and the viral load and CD4 cell count distributions of each subgroup upon entry to the model are shown in Table 1. A scenario with a healthier starting population was tested in sensitivity analysis to account for potential better health among individuals with undiagnosed HIV infection than at diagnosis. A scenario using data specific to black African heterosexuals was also examined in sensitivity analysis because of differences in several key parameters for this subgroup.

Transition Probabilities Along the Care Pathway and Between Health States. To progress along the care pathway, individuals with undiagnosed HIV-1 infection had the possibility every 3 months to be tested and become diagnosed. Subsequently, individuals with diagnosed HIV-1 infection could initiate treatment. In the first arm of the model, which represented the current state of HIV screening in the UK, individuals received diagnoses according to subgroup-specific observed annual diagnosis rates in the UK (57.2% for MSM, 33.2% for heterosexuals, and 66.7% for PWID),[10] converted to 3-month probabilities. The second arm of the model investigated increased screening efforts with 10%, 30%, or 50% relative improvement in diagnosis rates. To assess the impact of these increased screening efforts, given the recommendations regarding early treatment initiation, individuals in both arms of the model were linked to care and initiated antiretroviral treatment within 3 months of receiving a diagnosis. The current rate of linkage to care and treatment initiation in the UK (72% within 3 months) was tested in sensitivity analysis.[8] Following treatment initiation, individuals could progress through up to four lines of antiretroviral treatment, with the possibility of switching from one therapy line to the next every 3 months. These treatment-switching probabilities were estimated from expected durations of each therapy line as observed in the UK (Table 2).[14,15]

While continuing along the care pathway, individuals also experienced disease progression. Prior to treatment initiation, all individuals with diagnosed or undiagnosed HIV-1 infection experienced steadily declining CD4 cell count at rates calculated from the starting viral load distribution of each subgroup (Table 1)[13] and UK-specific data on natural CD4 cell count decline by viral load (Table 2).[16] Following treatment initiation, individuals experienced increasing CD4 cell count, representing treatment efficacy consistent with their line of antiretroviral therapy. CD4 cell count increases for typical first-, second-, and third-line antiretroviral regimens in the UK were obtained from published clinical trials and observational studies (Table 2). In the fourth, nonsuppressive line of therapy, individuals experienced a steady decline in CD4 cell count until death. This annual decline (22 cells/μL) was taken from a cohort of individuals with HIV-1 infection who remained on antiretroviral therapy despite three-class virological failure[17] and was notably slower than the CD4 cell count decline experienced by individuals prior to treatment initiation. All annual CD4 cell count changes were appropriately converted to 3-month transition probabilities in the model.

Mortality. Annual HIV-related mortality rates, stratified by CD4 cell count, were obtained from a UK Collaborative HIV Cohort study[18] and converted to 3-month probabilities (Table 2). Modelled mortality rates from a more recent study were tested in sensitivity analysis.[19] Non-HIV-related mortality was based on UK general population mortality by age and sex,[20] where risk of death in the model increased appropriately as individuals aged. A relative risk value of 1.5 was applied to these values based on evidence of higher non-HIV-related mortality risk among individuals with HIV-1 infection compared with age- and sex-matched general population controls (Table 2).[21,22] A scenario excluding this additional factor was tested in sensitivity analysis to account for the potential continued decline of this additional risk over time.

HIV Transmission. Table 1 presents the key parameter values for the Bernoulli transmission model component for each subgroup. For MSM and heterosexuals, estimates of long-term and one-time partners and condomless interactions before diagnosis were obtained primarily from UK-based survey studies of sexual behaviour.[23–27] PWID were assumed to have only one-time partners, with number of needle-sharing events before diagnosis obtained from an HIV transmission modelling study focused on PWID.[28]

MSM receiving an HIV diagnosis were expected to have a 21% reduction in mean number of long-term partners,[23] while no change in long-term partners was assumed for heterosexuals. Further, MSM and heterosexuals were expected to have a 53% reduction in condomless interactions with partners overall, based on a meta-analysis of studies of sexual behaviour before and after diagnosis.[29] Because of lack of data, this reduction was also applied to needle-sharing events for PWID (Table 1). To account for potential trends over time (e.g. more condomless interactions as a result of increased awareness of protection conferred by antiretroviral therapy), the 53% reduction was varied in sensitivity analysis.

Infectivity, or the probability of HIV transmission per condomless interaction or needle-sharing event, prior to treatment initiation differed based on the mode of interaction (Table 1).[30,31] For MSM and heterosexuals, infectivity was assumed to fall to 0% for all individuals on suppressive antiretroviral treatment based on recent results from the very large PARTNER and PARTNER2 studies among MSM and heterosexuals[5,6] and on similar earlier results primarily in heterosexuals.[7] Infectivity for PWID was assumed to decrease by 50% for all individuals on suppressive antiretroviral treatment.[28] Scenarios using alternate values for several of the transmission parameters were tested in sensitivity analysis.

Resource use and Costs. The model included costs for HIV screening, HIV-related clinical management, and management of secondary infections. The cost of HIV screening for each new diagnosis was estimated from the cost per individual test and the estimated proportion of positive tests among all tests. The cost per individual test was assumed to be £10 based on ranges reported in two UK studies examining HIV testing and staff costs.[32,33] Higher costs were examined in sensitivity analysis. The cost of screening per new diagnosis also depends on the proportion of positive tests among all tests. Within SHS, the positivity rate is 0.819% among MSM and 0.091% among heterosexuals,[8] yielding a cost per positive test of £1120 for MSM and £11 011 for heterosexuals. Positivity rates for PWID are not available (in SHS or any care setting). Although the prevalence of HIV-1 infection among PWID is 0.87%, two-thirds of PWID who accessed clinical care in 2017 were not tested.[8] Thus, because of lack of data, the model assumes a cost of £5000 per positive test for PWID (Table 1). For all groups, increased screening may lead to lower positivity rates (and hence higher costs per positive test); this possibility was examined in sensitivity analysis.

While most new HIV diagnoses occur in SHS, testing in general practice, secondary care, community, self-sampling, and home settings may also provide opportunities for increased screening. Where data on positivity rates were available,[8] scenarios were included in sensitivity analysis to examine the cost-effectiveness of increased screening and early treatment initiation in these settings. A £15 cost per test was used to reflect potentially higher test and staff costs compared with SHS. An exception was for self-testing, where the current retail price of about £30 was examined. Because positivity rates among self-tests are unknown, two hypothetical rates were examined in sensitivity analysis: 0.1% to represent a likely positivity rate given that individuals self-select for testing based on perceived risk, and 0.01% to represent the estimated prevalence of undiagnosed HIV-1 infection in the UK.[8]

HIV-related clinical management included antiretroviral therapy, hospital stays, emergency department visits, out-patient physician visits, laboratory tests, and other drugs including prophylaxis for opportunistic infections. Costs for antiretroviral therapy were based on a mix of recommended regimens typically used in each therapy line in the UK, with individual drug list prices taken from MIMS.[34,35] These costs were varied in sensitivity analysis to account for potential future treatments, increasing generic availability, evolving regimen selection along the treatment pathway, and negotiated prices. All other costs were obtained from a UK-specific cost study in which annual costs for the management of HIV infection were reported separately for the time periods before and after treatment initiation and stratified by CD4 cell count.[36] To capture costs incurred by individuals with undiagnosed HIV infection, the model applied costs equally before and after diagnosis for treatment-naïve individuals. All HIV management costs were varied in sensitivity analysis to account for the potential evolution of care over time.

For partners who acquired a secondary HIV-1 infection, the model applied an estimated lifetime cost of HIV management at the time of transmission. The discounted value for this cost was obtained from a lifetime HIV modelling study.[37] To ensure an appropriate calculation of the present cost of secondary infections, our analysis applied the discounted value at the time of transmission and further discounted that value to the present. To account for rising overall HIV management costs over time, a scenario with higher lifetime costs was tested in sensitivity analysis. All costs used in the model were inflated to 2017 UK pounds as needed[38] and are shown in Table 2.

Utility Values and QALYs. Utility values by CD4 cell count were taken from an analysis of EQ-5D responses from a large sample of HIV clinical trial participants (Table 2).[39] These utility values are commonly used in HIV economic analyses despite being somewhat dated. To account for potential shifts in quality of life by CD4 cell count over time, an alternate set of newer utility values was examined in sensitivity analysis (Table 2).[40] In both cases, utility values were applied equally along the care pathway and multiplied by the time spent in each health state to estimate total QALYs accrued by the modelled cohort.

To estimate QALY losses incurred by partners with a secondary HIV-1 infection, the model applied an estimated lifetime QALY loss at the time of transmission. Values for this QALY loss were obtained from the same modelling study used to estimate the lifetime cost of HIV management (Table 2).[37] Discounting was handled as described for costs. To account for life-expectancy and quality-of-life improvements over time for people with HIV-1 infection, a scenario with lower QALY losses was tested in sensitivity analysis.

Analyses

Base-case Analysis. The base-case analysis assessed the cost-effectiveness of increased SHS screening efforts (with 10%, 30%, or 50% relative improvement in diagnosis rates) compared with current SHS screening among MSM, heterosexuals, and PWID with initially undiagnosed HIV-1 infection from a payer perspective in the UK. To assess the impact of these increased screening efforts given recommendations regarding early treatment initiation, the analysis assumed that all individuals initiated antiretroviral treatment regardless of CD4 cell count within 3 months of receiving a diagnosis. For each population subgroup, model outcomes included lifetime per-person average total costs, QALYs, number of onward transmissions, and the incremental cost per QALY gained. Costs and QALYs included those accrued by individuals in the initial model cohort as well as by any of their partners with a secondary HIV-1 infection. All costs and health outcomes were discounted at an annual rate of 3.5% to represent current year values.[41] Undiscounted cost-effectiveness results were also estimated to give additional context regarding the value of the public health benefits expected to arise from additional screening.

Sensitivity Analysis. A deterministic one-way sensitivity analysis was conducted to assess the impact of changes to the model parameters most likely to influence the cost-effectiveness of HIV screening. Various alternate assumptions were tested about the population, the probability of early treatment initiation, disease progression, costs, utilities, behavioural patterns with partners, HIV infectivity, and HIV test positivity. Screening settings other than SHS were also examined in sensitivity analysis. For each scenario, discounted and undiscounted incremental cost-effectiveness ratios were estimated for MSM, heterosexuals, and PWID.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....