Our analyses were performed to assess the potential benefits and costs in the population of men having sex with men(MSM) related to immediate starting treatment at the time of HIV diagnosis compared to standard treatment path.
For this paper a new simple computational model was built to assess the cost-effectiveness of rapid cART treatment for newly diagnosedHIV infected patients, from the public payer’s perspective. Due to the analyzed disease and compared treatment paths, it were decided to take into account the effects and costs only in the period from diagnosis to start cART (possible gains resulting from earlier treatment at later time were omitted). Because of characteristics of HIV treatment and short period of delaying treatment, the presented simplification should not cause significant differences in patient's state of health and precisely illustrates the incremental effect of compared paths of treatment (Figure 1). The above assumption in case of previously published papers suggesting that delaying to cART is associated with additional costs, could be considered as a conservative approach.6
In our calculations, we also used the lifetime Markov model built during previous study (Kowalska 2017), which allowed to perform cost-utility analysis and determinate quality-adjusted lost years of life and additional costs for the payer for newly HIV infected patients.6
The model has one-month cycles and takes into account 33 events or illnesses divided into 18 health states and 8 additional events or diseases affecting estimated costs and the length of life. The baseline state of the model is an asymptomatic HIV, that was the people with HIV who did not experience additional comorbidities. In each cycle of analysis, patients were distributed between health states with assigned corresponding probabilities. We made the assumption, that after changing baseline health state it was not possible for the person to change their health state, except for death incidence, there was no possibility of the occurrence of the same event repeatedly and no possibility of having several diseases at the same time(Figure 2).Detailed information about used Markov model was described in the previous study Kowalska 2017.6
For the purposes of this analysis, the previously developed Markov model were also updated in case of baseline characteristics of patientswho registered in HIV specialist care between 1st January 2016 and 31st December 2017: mean CD4+ cells count, median HIV RNA and others (Table 1) and costs.
Risk of HIV transmission per sexual act
In the first stage, based on the unsystematic literature review, the number of potentially avoided new HIV infections were estimated due to rapid treatment implementation.
To find the necessary data on the risk of HIV transmission due to a sexual acts among MSMs, a research of the Medline medical database (via Pubmed) were performed. During the search, attempts were made to narrow down to the most reliable studies, i.e. meta-analysis that would be best suitable to our analysis population namely MSM patients not yet treated with cART. As part of the search, three publications were finally included to the analysis, Lasry 2014,7 Patel 20148 and Baggaley 20189 in which the risk of transmission per sexual act were found(Table 2).Due to the fact that the Baggaley 2018 study was published quite recently in 2018 and its significant extent of scope of study is also included in other reviews, it was decided to use this data in the base case scenario of analysis.In addition, the results from the remaining reviews were decided to be tested in a sensitivity analysis.
Time from diagnosis to start cART treatment
Based on the data collected for the MSM who registered in HIV specialist care, such as time of conducting the HIV-test, time of HIV diagnosis and the start of cART treatment, the average and median time of delay in access to therapy were determined. The statistical analysis of survival curves for the time to start treatment. was conducted. In the final calculations, curves based on the generalized gamma distribution (main scenario) and Weibull distribution (sensitivity analysis) were selected cause by the best fit according to the AIC and BIC criteria (Figure 3).
Viral load and risk transmission
Literature studies clearly show that plasma viral load is directly associated with risk of sexual transmission of HIV. Hence, the new simple computational model built for this analysis allows to perform calculations for two different variants: 1. Excluding impact of viral load at the risk of HIV transmission and 2. Including the level of viremia.
The baseline risk of HIV transmission adopted based on the data found in the reviewwas adjusted for HIV RNA viral load according to the Quinn 200010 study. In that paper, it was estimated thar each log increase in viral load was associated with an increase by a factor of 2·45 in the risk of transmission.
In the first stage, the data for the cohort was stratified into five risk groups depending on the viral load levels, similarly to the Quinn 200010 publication. Based on this data, the average level of HIV RNA viral load in each of the five groups was also determined.Finally, due to the small size of the group with level of viral load between 1·70 to 3·54 log10 HIV RNA copies/ml (only 11 patients from 344 patients from the cohort, 3%) it was decided to include them into group of patients under 3·54 log10 copies HIV RNA copies. Then, the probability of HIV transmission was determined for each group based on the difference between mean viral load in each group compared to the mean value on level of HIV viral load in the whole cohort and using data from the Quinn 2000 study.
For example, in the group of patients with 1·70- 3·54 log10HIV RNA copies/ml, the average level of viremia was 2·77 log10copies and was about 1·86 log10copies/ml lower than the average level of viral load in the entire cohort. According to the data in the Quinn study, this difference is associated with more than 5-fold reduction risk of sexual transmission to 19% of the baseline risk from Baggaley 20189 and others.
Ultimately, when impact of viral load at the risk of HIV transmission was included into our assesment , the propability of HIV transmission per insertive sexual act for patients with less than 3·54 log10 HIV RNA copies/ml was established at level 0·01% compared to 0·17% reported in the Baggaley 20189 study. Detailed information about probabilites of new HIV infections per sexual act for all stratifield groups was preseneted in Table 2.
Risk profiles of sexual behavior
In our analysis, HIV transmission was assumed to occur only through sexual contacts.
The probability of infection was based on data found in published meta-analyzes. Our calculation also takes into account the impact of condom uses for final risk of transmission.
Due to the methodology of the analysis and showing the incremental effect of immediate starting treatment at the time of HIV diagnosis, the estimated number of new HIV infections relates to the period in whichpatient is not receiving cART. Profiles of the risk of transmission were adopted in a similar way as in the previous study Kowalska 20176 i.e based on the average number of sexual partners, number of sexual acts, % frequency of condom use per act.
Additionally, we assumed that each patient from the analysed cohort had the same number of intercourses and had the same number of intercourses with each sexual partner. For the medium risk scenario, which was considered a baseline model, the rate of transmission was estimated assuming that an average HIV positive person has 10 partners per year, 10 monthly sex acts and 50% frequency of condom use per act. For the low and high-risk scenarios we assumed a person to have 3 and 50 partners per year, 10 and 20 sex acts per month and 90% and 0% coverage with condom use, respectively.
In our analysis also assumed that 28% of MSM patients have HIV+ partners (this assumption reduces the total number of new potential infections)6
Costs and other data
For the purposes of this analysis, previously Markov model was updated in terms of each health state representing different AIDS and Non-AIDS defining illness and the costs of cART treatment. The costs of cART treatment were adopted based on actual data from the National Program of cART Treatment and expert opinion, which was adopted at EUR 482 per month (461EUR drugs and 22 EUR monitoring treatment)while the cost of treating health states were adjusted based on inflation rate between 2015 and 2019 published by the Central Statistical Office.11Additionally, data about patient mortality used in Markov model, i.e. life expectancy tables was updated.
Additional costs related to the implementation of rapid cART were not included due to the method of financing of Centers of HIV Treatment in Poland as flat-fee. The simplification approach should not have a significant impact on inference from the analysis and results.