Data were obtained from the Israeli Central Bureau of Statistics (Demographic, Mortality, Employment Costs), Israeli Ministry of Health (Cost Data), Israeli Ministry of Health Department of Tuberculosis and AIDS (HIV Care Protocols, Prevalence and Incidence), WHO/CDC (Utility Weights). The above data were supplemented by data from a search of the world literature from 1990 to mid-2018 on PubMED, using keywords: (HIV or AIDS) and (prevention or prophylaxis or PrEP), supplemented by unpublished conference proceedings (PrEP Effectiveness, Side-Effects, transitions from HIV to AIDS and from AIDS to death).
Age-specific data for HIV-positive MSM (ie: infected) persons for 2015 were obtained from the National HIV/AIDS registry maintained by the Department of Tuberculosis and AIDS of the Ministry of Health. The age-specific distribution for HIV-negative MSM was based on an estimated 78,013 MSM in Israel, representing 3% of the male population [8] and an adjustment for these “susceptibles” being one year younger than HIV-positive persons on average [9].
Cost-utility analysis
Model development
An Excel-based spread-sheet model was constructed, incorporating intervention costs, treatment costs, disability weights, epidemiology and health service utilization, mortality rates, PrEP efficacy, and indirect costs. (Parameter values listed in Appendix 1).
Cost utility calculation
The model calculated the effect of PrEP on the incidence and mortality from HIV/AIDS in Israel in the MSM risk group as a basis of the calculation of the gold-standard “cost-utility ratio” (CUR) to estimate the cost-effectiveness of providing for ten years:-
- (a)
a continuous PrEP regimen.
- (b)
a PrEP regimen “on demand” countries, such as that used in France [10].
The cost utility ratio (CUR) calculated the net costs per averted Disability Adjusted Life Year (DALY) added as a result of using PrEP, using the standard formula:
$$ \mathrm{Net}\ \mathrm{Costs}\ \mathrm{per}\ \mathrm{a}\mathrm{verted}\ \mathrm{DALY}=\frac{\left(\mathrm{Costs}\ \mathrm{of}\ \mathrm{PrEP}-\mathrm{Savings}\ \mathrm{in}\ \mathrm{treating}\ \mathrm{HIV}\&\mathrm{AIDS}\right)}{\mathrm{DALYs}\ \mathrm{averted}\ \mathrm{a}\mathrm{s}\ \mathrm{a}\ \mathrm{result}\ \mathrm{of}\ \mathrm{decreased}\ \mathrm{mortality}\&\mathrm{morbidity}} $$
Costs were viewed from a societal perspective at mid-2018 price levels and therefore included costs not only incurred by the health and welfare services but also included work absences, transport costs to receive treatment, and premature burial costs. All future costs and averted DALYs were discounted at an annual rate of 3%.
We valued the societal costs of premature mortality by using a methodology that calculated discounted premature burial costs alongside a “zero friction cost” [11]. Zero friction costs assume that a worker, upon death, will be replaced by another unemployed worker with a similar skill set. We refrained from using discounted future productivity losses (whether or not adjusted by discounted future consumption losses) as these do not represent real resource costs, even though they are invariably used by advocates of disease specific interventions for advocacy purposes to inflate the monetary impact of their disease.
Intervention costs
Intervention costs were based on PrEP costs according to a daily regimen plus the costs of the treatment protocols recommended by the Ministry of Health, which consisted of visits to the prescribing physician as well as numerous laboratory tests. Co-payment costs were excluded from the analysis as these are basically a transfer payment from the individual to the pharmaceutical company via the individuals’ HMO.
A sensitivity analysis was performed using unit costs of monitoring tests provided by one of the HMOs for the research evaluation study, since these more closely represent the true resource costs as opposed to the Ministry of Healths price schedule whose costs represent the maximum authorized cost.
We made no provision for any costs associated with PrEP toxicity, since most of the PrEP trials did not report any differences in the rate of serious adverse events between the study and control groups [2, 3, 12,13,14]. Furthermore, no severe adverse events were reported from a safety and tolerability study [15].
Generic forms of PrEP pharmaceuticals have recently become available, but from experience with other pharmaceuticals, we only expect that there will be a further 10% fall in the current maximum price authorized by the MOH. We used this 10% decrease as our baseline scenario. In our sensitivity analyses we explored scenarios where the price decrease might decrease further as a result of each HMO negotiatiating separately with the pharmaceutical suppliers.
Treatment costs
Treatment costs were based on the global annual sum of $26,359 USD per HIV/AIDS person that is currently paid to the Israeli HMOs to cover all out-of-hospital costs incurred by their members with HIV/AIDS (Dr. Daniel Chemtob, personal communication, 2018).
In addition, monitoring costs were added based on the Ministry of Health protocols and price lists for the 81.8% of HIV-positive patients who are Israeli citizens, living in Israel in 2016.
Costs of AIDS were based on the discounted ART and monitoring costs applied to the respective survival rates of the estimated 96.5% (ie: 65% plus an estimated 90% of the original 35% who did not originally receive ART) who will receive ART and the 3.5% who did not take ART. HIV-positive persons used 4.26 general hospital days annually (Personal Communication. Ziona Haklaii, Ministry of Health Statistical Department), compared with 0.62 days use by HIV-negative persons. These 3.64 additional general hospital days a year used by HIV-positive persons were costed at an average cost of 617 USD per day [16].
However, due to lack of data on differential utilisation (by HIV-positives and negatives) of pharmaceuticals, ambulatory, emergency room and out-patient visits for diagnoses not relating to HIV, we were unable to estimate increased utilisation costs on account of HIV-positivity. No hospice costs were included as this care modality is nowadays not used any more.
Disability weights (quality of life)
No reductions in quality of life were assumed on account of taking PrEP since clinical trials have indicated minimal side effects [2, 17]. Age-specific health utilities were multiplied by the following utilities in order to calculate utilities (or disability weights) for the following health or disease states: 1.00 for non-symptomatic HIV, 0.87 for HIV-positive taking ART [18], 0.80 for HIV-positive not taking ART [18], 0.85 AIDS case taking ART [17, 19], 0.71 AIDS case not taking ART [20, 21]. Resultant DALYs were discounted at 3% per annum.
Epidemiology
Treatment impact according to different periods
From the period 2010–2013 to the period 2014–2017, there was a decrease in HIV incidence among MSM in Israel from 2.16 to 1.74 per 1000 MSM (1, MOH Department of TB and AIDS), resulting in an annual decrease of 5.20%.
This was caused by lower transmission probabilities as a result of the gradual adoption of improved preventive strategies and of improved ART drugs and protocols which lowered the viral load thresholds prescribing ART.
Due to this decline, our baseline model assumed that in the event that PrEP would not be made available (i.e. the non-intervention scenario), HIV incidence would continue to decrease at a rate of 5.20% per annum for a further four years and then remain constant. Two additional sensitivitity analyses were carried out under assumptions that there will be no further decrease and that the decrease will last for eight years more.
Natural history of HIV and the length of time before reaching AIDS or death
In building our model, we had to consider the natural history of HIV/AIDS in order to estimate the length of time between two different clinical phases- “asymptomatic” (HIV) and symptomatic (AIDS). By definition, the natural history time estimates have to be based on a period when no HIV treatment was available.
The median time for asymptomatic HIV to “progress” to symptomatic HIV is around ten years [22]. Transition and mortality rates were based on sources both from actual trials and from modelling studies based on infectivity and frequency of sexual relations.
Relative annual age-specific progression rates from HIV to AIDS for persons not taking ART [23, 24] were applied to non age-specific Israeli data from 1981 to 96 which showed the average progression time to be 15.5 years [1]. Progression rates in persons taking ART were assumed to be 23.6% those of persons not taking ART [23,24,25].
Similarly relative annual age-specific mortality rates from AIDS to death for persons not taking ART [23, 24, 26, 27] were applied to (non age-specific) Israeli mortality data from 1981 to 96 which reported an average time of 5.1 years until death without treatment [1]. Mortality rates in persons taking ART were assumed to be 11.5% of persons not taking ART [23, 24, 28].
Annual age-specific mortality rates in HIV-positive persons (without an AIDS diagnosis) were calculated, assuming the rates to be 11.1 and 16.7% of those for AIDS diagnosed patients who did [23, 26, 27] and did not receive ART respectively [23, 28, 29]. Finally, an adjustment was made for gender-specific mortality rates for diagnoses not attributable to HIV or AIDS [23].
PrEP efficacy
The essential drivers of the model included PrEP efficacy as obtained from the literature [2,3,4,5] and optimal expected compliance rates described in the Israel Ministry of Health’s PrEP evaluation proposal for the continuous regimen, and from the literature for the “on-demand” regimen.
The continuous PrEP regimen requires daily adherence to a fixed drug combination of Tenofovir Dixoproxil Fumarate with Emtricitabine (TDF/FTC). Users take one tablet once every 24 h [5]. The on-demand regimen includes the same fixed-dose pill (TDF/FTC), and the user is instructed to take a “loading dose” of two pills 2–24 h before sex, an additional pill 24 h after the first dose, and a final pill 48 h after the first dose. In instances of multiple consecutive intercourse, users are instructed to continue one pill every 24 h during the period, plus an additional two days [5].
Our definition of a high-risk group targetted for PrEP, were MSM who engage in unprotected anal intercourse (UAI), who account for around 21.3% of MSM [9].
Lack of accurate data precluded us from building a model based on the product of the risk of transmission, which is influenced by UAI, circumcision [30], sexually transmitted disease status [31], age-specific frequency of sexual contacts, and the probability of being an HIV carrier.
Instead, we used the following formulae to calculate (by aggregating over each age group) the expected number of incident HIV cases in MSM:
$$ \mathrm{Number}\ \mathrm{of}\ \mathrm{HIV}\ \mathrm{incident}\ \mathrm{cases}=\mathrm{Incidence}\ \mathrm{Rate}\times \mathrm{Number}\ \mathrm{of}\ \mathrm{Susceptibles} $$
$$ \mathrm{Number}\ \mathrm{of}\ \mathrm{susceptibles}=\mathrm{Number}\ \mathrm{of}\ \mathrm{MSM}-\mathrm{Number}\ \mathrm{of}\ \mathrm{MSM}\ \mathrm{HIV}-\mathrm{positive} $$
where the incidence rate for the period 2014–2017 was 174.2 per 100,000. Based on the HIV prevalence rate of 2.4% amongst MSM in Israel (MOH Department of TB and AIDS), the incidence rate among susceptibles is178.4 per 100,000.
Next, we assumed that being high-risk increases the relative risk (RR) of HIV approximately fivefold since condom use reduces heterosexual HIV transmission by 80% [32, 33]. This enabled us to calculate the number of incident cases that would occur in the high-risk and low risk (i.e.: practicing protected anal intercourse) MSM populations.
We then applied the protective efficacy of PrEP of 86% found in the clinical trials [4, 5] to estimate the reduction in HIV cases that would occur if PrEP were offered to the MSM high risk group, under the assumption that 80% of the high risk MSMs in Israel would take the opportunity to try PrEP.
In addition, we assumed in the baseline scenario that 25% of the low risk MSMs would be interested in taking PrEP, and subsequently around 75% of this group would cease to use condoms (ie: 6.25% of the low risk group would transition to high risk) [34, 35].
Indirect costs
Our model asumed there were no differences in employment rates between general male and MSM populations, based on a Canadian study [36]. Based on the average of 40.5 h during a five-day work week for Israeli males [37] we estimated time off work to visit medical services to be 8.06 h per emergency room visit, 4.03 h for all other visits [38, 39] and 5.78 h per hospitalized day (this takes into account persons who are in hospital over weekends). Visiting frequencies were based on the Ministry of Health care protocols for ART and non-ART patients (taking into account initial confirmatory visits). Total age-specific work losses were estimated by the product of: frequency of visits, time off work per visit, the average employment cost per hour of 24.20 USD [39], and the male labour force participation rate adjusted by the unemployment rate [39]. Further indirect costs were added to take into account premature burial costs which are defined as the discounted value of burial costs of the person dying from AIDS less the discounted burial costs of dying in the future from causes other than AIDS.
Figure 1 illustrates the structure of the model where each path has associated specific probabilities leading to differential path -specific outcomes, costs and utilities.
Sensitivity analyses
These were carried out in combinations of the following scenarios:
- i.
The major sensitivity analysis was based on examining a range of 0–25%-50–75% (baseline) for the percentage of low risk MSM who on receiving PrEP abandon using condoms due to their enhanced feelings of protection against HIV.
- ii.
Assuming absolutely no low-risk MSM take up PrEP.
- iii.
Various efficacies of PrEP ranging from the lower 90%/95% bound from trials 56.3% [2, 4, 5] to reported efficacy of 97.1% from an observational study [10].
- iv.
Based on the 10% (baseline), and possible 50% & 90% [40] future declines in current pharmaceutical prices of PrEP.
- v.
Scenarios of zero, four (baseline) and eight years decrease in HIV incidence rates in a non-intervention scenario.
- vi.
The effect of altering the efficacy of PrEP was also examined for the baseline scenario..
- vii.
A lower cost “on-demand” regimen, assuming a 40% decrease in PrEP utilization without loss of efficacy [28, 29].
- viii.
The effect of adding an arbitrary average friction cost of 3000 USD representing costs to cover instances where persons have to be retrained to replace the deceased person and their specific employment skill set.
- ix.
Using real resource annual costs of monitoring PrEP (503 USD for first year, then subsequently 365 USD) instead of the costs based on the maximal MOH pricing lists (1179 USD for first year then subsequently 1018 USD).
- x.
Since some of the high-risk MSM who used PrEP, especially those who are discovered to be HIV-positive during the screening process, might start to use condoms in order to reduce their HIV transmission risk due to their greater awareness of its benefits [2, 10] we ran a sensitivity analysis that assumed between 10 and 25% high risk receivers of PrEP started using condoms as extra protection.
While we recognize that in examining the feasibility of government-funded PrEP, there is a need to consider the benefits and costs that consequently fall on other ministries or private individuals, this cost-utility calculation was based only on a health services perspective for the purpose of the sensitivity analysis.
Decision rules
In the absence of Israel specific cost-effectiveness guidelines, decision rules were based on the WHO criteria that take into account the resources available for investment in health services in a country [41]. The PrEP intervention will be defined as being very cost-effective and cost-effective if the cost per averted DALY is less than the 40,439 USD per capita GDP of Israel [42,43,44] or between 1 and 3 times the per capita GDP respectively (40,439–121,316 USD). If the cost per averted DALY is more than three times the GDP (121,316USD) per capita then the intervention will be regarded as not being cost-effective. In the event that treatment is effective and treatment savings exceed intervention costs then the PrEP intervention will be cost-saving, a win-win situation [45].
As a form of sensitivity analysis we also considered a second novel stricter decision rule [46] with a cost-effectiveness threshold range of (19,463–22,208 USD at 2013 prices updated to 2018) that takes into account the opportunity costs (of the effect on health) involved in resource allocation decisions.