Table 1.
Characteristics of the contributing models
| Goals | Optima HIV | HIV Synthesis | Imperial College London model | EMOD | |
|---|---|---|---|---|---|
| Structure | Compartment model with disaggregation by risk group | Population-based compartment model with sex, age, and risk group disaggregation; 1-month time step | Individual-based stochastic model; 3-month time step | Compartmental model with sex, age, and risk structure | Individual-based stochastic simulation; daily time step; network model |
| Approach to calibration of data and estimates for specific settings | Epidemiological parameters (probability of transmission per sex act; variation by stage of infection, presence of other STIs, and effectiveness of condoms and ART) are varied to fit the model to prevalence estimates from surveillance and surveys | Epidemiological parameters (probability of transmission per sex act, variation by stage of infection [informed by CD4 cell counts and viral load monitoring], HIV testing rate, mortality rate, presence of other ulcerative STIs or tuberculosis, or both, and effectiveness of condoms, circumcision, and unsuppressive or suppressive ART) are varied to fit the model to prevalence estimates from surveillance and surveys | Parameters relating to population characteristics, sexual behaviour (condomless sex), age-gender mixing (ie, distribution of ages of male sexual partners of women of a given age and vice versa), HIV acquisition, HIV testing, natural history (CD4 cell count and viral load), ART, and risk of AIDS and death varied within plausible bounds to create a range of setting scenarios; country situations are mimicked by selecting setting scenarios with epidemic and programmatic features consistent with country data | Epidemiological and HIV intervention parameters (probability of transmission per sex act, proportion of the population in each risk group, sex-specific and risk-specific sexual contact rates, risk-specific condom use, amount of mixing of at-risk groups, ART and VMMC uptake) are varied to fit the model to country-specific prevalence, incidence, ART and VMMC coverage, and mortality estimates, reported by UNAIDS | Parameterised with epidemiological data including population size, fertility, mortality, VMMC coverage, and health-seeking and sexual behaviour; data from South Africa on age-specific and sex-specific HIV prevalence, ART coverage, population size, and HIV incidence were used to calibrate the model; calibration was done using a parallel simultaneous perturbation optimisation algorithm; roulette resampling in proportion to the likelihood of each simulation was used to select 250 model parameter sets |
| Sexual behaviour | Behaviours (number of partners, sex acts per partner, condom use, needle sharing) differ by risk group: female sex workers, male clients of sex workers, men and women with non-regular partners, monogamous couples, MSM, and PWID | Behaviours (type of partners [regular, casual, or commercial sexual; injecting], sex acts per partner, condom use, needle sharing) differ by age and risk group (female sex workers, clients of sex workers, MSM, and PWID) | Number of short-term condomless sex partners in a 3-month period, and potentially one long-term condomless sex partner in a 3-month period | Sexual contact rates differ by risk group (low, medium, high) and by age; condoms are assumed to be used differentially by each risk group | Four types of sexual partnerships (marital, informal, transitory, and commercial) are remembered over time and formed according to specifiable partner age patterns |
| HIV acquisition determinants (including prevention interventions included) | Acquisition depends on characteristics of the individual (number of partners, circumcision status, PrEP use), the partner population (HIV prevalence, ART use, stage of infection), and the partnerships (sex acts per partner, prevalence of other STIs, type of sex, condom use) | Acquisition depends on characteristics (type of act [regular, casual, or commercial sexual; injecting], number of acts, circumcision and PMTCT status, and PrEP and PEP use), and population status (HIV testing, diagnosis, HIV prevalence, unsuppressive or suppressive ART use, stage of infection), in specified partnerships | Acquisition risk for each condomless sex partner depends on viral loads of people of (age-mixing relevant) opposite sex; circumcision and PrEP are modelled | Acquisition risk for each sex act depends on stage of HIV infection, ART status, circumcision status, PrEP use, and condom use | Acquisition risk depends on characteristics of an individual that include number of partners, circumcision status (males), condom use, other STIs, and coital acts; additionally, characteristics of the partner (including ART use if HIV positive and stage of infection) also determine an individual's acquisition risk |
| HIV natural history | Rate of decrease in CD4 cell count off of ART depends on current CD4 count and age; mortality off of ART depends on sex, age, and CD4 cell count; mortality on ART depends on CD4 cell count at initiation, age, sex, and duration on ART | Rate of decrease in CD4 cell count and viral load off of ART depends on current CD4 cell count and viral load; change in CD4 cell count on ART depends on current CD4 cell count and viral suppression of treatment; mortality both on and off of ART depends on current CD4 cell count and ART status (unsuppressive or suppressive) | Level and trend in viral load is dependent on sex and age; CD4 cell count decrease is dependent on viral load; risk of AIDS and death is dependent on current CD4 cell count, current viral load, and age | Rate of progression through each stage of infection is modelled on the basis of the mean duration of each stage; mortality on ART depends on whether ART was initiated at a CD4 count of ≥350 cells per μL or <350 cells per μL | HIV prognosis is calculated using a Weibull distribution where the parameters of the distribution are derived from CD4 cell count and age at the time of infection; CD4 cell count decreases from the time of infection; CD4 cell count increases if an individual initiates ART; new prognosis is calculated for ART dropouts |
| HIV testing and diagnosis | Testing is by type of test and population group and determines knowledge of status, but is not linked to transmission since ART coverage is a direct input | Testing is by type of test, population group, and year, which determines knowledge of status and allows linkage to care and initiation of ART on the basis of coverage level | A person without a diagnosis of HIV can be tested or not in each 3-month period; testing is indicated in antenatal clinics and for symptoms potentially of HIV, and general testing with various degrees of targeting at people with higher probability of infection | HIV testing and diagnosis is not explicitly represented but is considered a prerequisite for any initiation of treatment | HIV testing and diagnosis occurs voluntarily, at antenatal visits, or once symptomatic |
| ART | Risk of mortality while on ART is determined by age, sex, CD4 cell count at treatment initiation, and duration of treatment | Risk of mortality while on ART is determined by dynamically changing CD4 cell count over the course of treatment | Specific drugs and their current level of activity given drug resistance, and current ART adherence; being currently on ART has a small independent effect on risk of AIDS and HIV death over and above these factors; ART interruptions for reasons apart from disruptions are modelled; transmission of drug resistance | Risk of mortality while on ART is specific to whether ART was initiated early (CD4 count of ≥350 cells per μL) or late (CD4 count of <350 cells per μL) | Risk of mortality while on ART is determined by age, sex, CD4 cell count at treatment initiation, and duration of treatment |
| ART interruption | Immediate return to CD4 cell count at time of treatment initiation; survival progression is identical to those who are treatment naive | Substantial initial decrease in CD4 cell count towards pre-ART nadir (consistent with Grund et al15) and gradual return after interruption; viral load return is dependent on viral load monitoring | Immediate viral load return to pre-ART levels, substantial initial decrease in CD4 cell count towards pre-ART nadir (consistent with Grund et al)15 | Mean survival time of 14 years after stopping ART, exponentially distributed, based on the survival time for HIV-positive individuals who have never been on ART (consistent with Todd et al16); a sensitivity analysis is also modelled using mean survival time of 32 years (Grund et al15) and 8 years after stopping ART (a hypothetical worst-case scenario); immediate return of death rate to pre-interruption levels on resumption of services with no effect on survival, except if individuals have already progressed to AIDS | CD4 cell count decreases after drop out from ART; prognosis is recalculated on the basis of age and CD4 cell count at the time of interruption of ART; calculation of ART prognosis after re-enrolling is identical to the initial enrolment with prognosis parameters corresponding to the age and CD4 cell count at re-enrolment |
| MTCT | Depends on the duration of breastfeeding and the CD4 cell count of the mother if no prophylaxis, or if on prophylaxis regimen, the type of regimen, retention at delivery, and interruption of ART during breastfeeding; retention on prophylaxis at delivery, duration of breastfeeding, and drop out from prophylaxis during breastfeeding | Depends on CD4 cell count of the mother, PMTCT coverage, and duration and breastfeeding practice | Dependent on viral load of the mother at birth | MTCT and PMTCT are not explicitly modelled | Depends on CD4 cell count of the mother if no prophylaxis, or prophylaxis regimen; retention on prophylaxis at delivery |
Data are for adults and children, unless otherwise stated. ART=antiretroviral therapy. EMOD=Epidemiological MODeling software. MSM=men who have sex with men. MTCT=mother-to-child transmission. PEP=post-exposure prophylaxis. PMTCT=prevention of mother-to-child transmission. PrEP=pre-exposure prophylaxis. PWID=people who inject drugs. STI=sexually transmissible infection. VMMC=voluntary medical male circumcision.