A cost-effectiveness analysis has found that injectable, long-lasting antiretroviral therapy (ART) would need to cost no more than $ 131 a year in order to be cost-effective if used for patients who are not fully virally suppressed in lower-income settings in regions like sub-Saharan Africa.
The analysis found that injectable cabotegravir/rilpivirine would only reach the threshold of cost-effectiveness, defined as $ 404 per disability-adjusted life year (DALY) averted, in people who had viral loads over 1000 on standard oral ART. The price of oral ART was considered to be $ 78, in line with prices cited by the Clinton Health Access Initiative in 2019 for generic tenofovir/emtricitabine/efavirenz and tenofovir/emtricitabine/dolutegravir.
Injectable cabotegravir/rilpivirine (CAB/RPV) has been approved by European regulators for injection once every one or two months, while licensing in North America is currently only for use once a month. In Europe, the two drugs are branded as Vocabria and Rekambys, respectively. In North America, the two drugs are packaged together, with the brand name of Cabenuva.
At present, the injectable therapies are very expensive. A US list price of over $ 48,000 a year was quoted by the New York Times the day Cabenuva was licensed, while a recent cost-effectiveness analysis of injectable PrEP used Cabuneva‘s Canadian list price of $ 25,800 (since we do not yet have a price for Vocabria as PrEP). These prices are likely to fall as discounts are negotiated.
When used as ART, studies found that CAB/RPV was as effective as comparison oral therapies but not more so. The results for PrEP are different, with considerably improved efficacy for injectable versus oral PrEP in two recent studies, so cost-effectiveness calculations may come to different conclusions.
The study was conducted by a team from University College London along with the Bill & Melinda Gates Foundation and four other universities.
The model used is the so-called HIV Synthesis Model, which in effect sets up a ‘virtual epidemic’, in this case based on the epidemic in sub-Saharan Africa, which is able to predict outcomes in viral suppression, health, mortality and cost when numerous different starting conditions are varied. Among examples of those possible input variables are the rate of HIV testing, the percentage of HIV-positive people on ART, the percentage with a viral load over 1000, the extent of viral load monitoring, the rates of linkage to and retention in care, the rates of adherence and of complete ART interruption, the degree of emergent drug resistance, and the proportion who have switched to second-line therapy.
The inputs included three scenarios:
- one in which everyone on ART switched to injectable cabotegravir/rilpivirine,
- one on which only those with a viral load over 1000 did, and
- one that looked only at the cost of switching people with a viral load under 1000 to the injectables.
Input values were based on observed data, but adjusted somewhat for the time since the studies took place. Examples include an overall regional HIV prevalence of 17% and an annual incidence of 1.1% (both at the higher end of observed data); a rate of 88% both for the proportion of HIV-positive people diagnosed and the proportion of those diagnosed who are on ART (again at the upper end of recently observed rates); 64% for the proportion of all HIV-positive people with viral loads below 1000, and 90% for the same proportion in people on ART; 14% for the proportion of people on oral second-line (usually protease-inhibitor-based) ART; and 13% for the proportion of people with resistance to NNRTI drugs.
The latter is an important consideration, as NNRTI resistance in people who have taken drugs of that class, such as efavirenz, will also have significant resistance to rilpivirine. The model also took into account the fact that in comparison studies, cabotegravir proved to be somewhat more ‘fragile’ to emergent integrase-inhibitor resistance mutations than dolutegravir, the main oral therapy of this class, and even more so than bictegravir. People on cabotegravir with those mutations are roughly twice as likely to develop an unsuppressed viral load than people on dolutegravir.
The model found that in all three scenarios the introduction of injectables would increase the proportion of people with a viral load under 1000: by 5.3% if injectables were given to everyone, and by 4.1% in people with unsuppressed HIV (partly because the model included an assumption that they were likely to be less adherent). The resultant reduction in AIDS-related deaths was 0.19% per 100 patient-years, i.e. that if you took 1000 patients, two more out of those 1000 would survive who would otherwise have died. Mortality only reduced by one in 2000 patients per year in the group with suppressed viral load.
Integrase inhibitor resistance would increase, but only slightly, by 0.8%, and less so in people with viral loads over 1000, even though they are more likely to already have resistance, because it was assumed poorer adherence would mean they would actually have less exposure to the drugs. For the same reason fewer people starting injectables with viral loads over 1000 would develop new NNRTI resistance (1.5%) than if everyone started injectables (4.3%).
“For injectables to be cost-effective, they would have to be carefully targeted towards people with unsuppressed HIV viral loads and/or poor adherence.”
In total, introducing injectables would avert the loss of 30,400 disability-adjusted life years (DALYs) per year to the million-strong ‘virtual cohort’ modelled. It would avert the loss of 17,900 DALYs per year to the 35% of people with viral loads over 1000.
What this means is that each year during the ten years of the model studied, the use of injectables would stop that many people dying or developing incapacitating illness during the year.
The extra cost of achieving this would be $ 42.9 million over ten years if everyone switched to injectables, but only $ 5.4 million if only those with an unsuppressed viral load did.
This leads to the ‘bottom line’ – the incremental cost-effectiveness ratio (ICER), i.e. the cost per DALY averted. This would be $ 404 per DALY averted in the group starting with a viral load over 1000. This was just within the threshold of $ 500 per DALY averted which, from previous studies, the researchers calculated might be cost-effective in sub-Saharan Africa (it was $ 750 in the upper-middle-income setting of South Africa). The cost per DALY averted would be $ 1638 if all people switched to the injectables and $ 2808 in people who were already virally suppressed – neither would be cost-effective.
The proportion of different model scenarios in which injectable cabotegravir/rilpivirine was cost-effective was 59% in people with viral loads over 1000, but only 7% in all people taking ART and 2% in people with viral loads below 1000. The maximum cost of the injectables, including costs such as cold-chain storage and more clinic appointments, would be $ 131 per person per year in order for the $ 404 cost per DALY averted to be achieved, or $ 53 more than the minimum quoted price for TDF/FTC/dolutegravir oral ART. Cost-effectiveness would be somewhat improved if injections were every two months rather than monthly, or if more people were on second-line therapy, which is more expensive.
There are many uncertainties the model could not include. Among these are the degree to which the injectable therapies might improve adherence and improve linkage to care and retention in people reluctant to take oral ART.
The model does however imply that for injectable cabotegravir/rilpivirine to be cost-effective, it would have to be carefully targeted towards people with unsuppressed HIV viral loads and/or poor adherence (this also depends on viral load monitoring becoming more general). Young people under 20, people without regular housing, and migrant workers who regularly change location were among the possible populations suggested.
Because of the uncertainties in the model, the researchers say that they will continue to introduce new findings from trials and implementation studies to improve their model’s accuracy. The same sort of analysis would need to be done for injectable PrEP, where the superiority of injectables shown in studies, and the fact that PrEP does not need to be delivered continuously, might produce a different result.