PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Lancet. Author manuscript; available in PMC Feb 7, 2012.
Published in final edited form as:
PMCID: PMC3274419
NIHMSID: NIHMS312757
The Role of Acute and Early HIV Infection in the Spread of HIV-1 in Lilongwe, Malawi: Implications for “Test and Treat” and Other Transmission Prevention Strategies
Kimberly A. Powers, Azra C. Ghani, William C. Miller, Irving F. Hoffman, Audrey E. Pettifor, Gift Kamanga, Francis E.A. Martinson, and Myron S. Cohen
Department of Medicine (KA Powers PhD; WC Miller MD, PhD, MPH; IF Hoffman, PA, MPH; MS Cohen, MD), Department of Epidemiology (KA Powers; WC Miller; AE Pettifor, PhD; MS Cohen), and Department of Microbiology and Immunology (MS Cohen), University of North Carolina, Chapel Hill, NC, USA; Department of Infectious Disease Epidemiology (AC Ghani, PhD), Imperial College, London, UK; and UNC Project Malawi, Lilongwe, Malawi (G Kamanga, CO; FEA Martinson, MD, PhD)
Correspondence to: Kimberly A. Powers, PhD, Center for Infectious Diseases, CB 7030, Bioinformatics 2nd Floor, Chapel Hill, NC 27599-7030, USA, powersk/at/email.unc.edu
Background
HIV transmission risk during acute and early HIV infection (EHI) is sharply elevated, but the contribution of EHI to ongoing HIV transmission is controversial. However, in settings where EHI contributes substantially to secondary transmissions, early diagnosis and intervention may be critical for HIV prevention. We estimated the contribution of EHI to HIV incidence in Lilongwe, Malawi and predicted the future impact of hypothetical prevention interventions affecting EHI only, chronic HIV infection (CHI) only, or both stages.
Methods
We developed a deterministic mathematical model describing heterosexual HIV transmission, informed by detailed behavioural and viral load data collected in Lilongwe. We included sexual contact within and outside steady pairs and divided the infectious period into multiple intervals to allow for changes in transmissibility by infection stage. We used a Bayesian melding approach to fit the model to HIV prevalence data collected over time at Lilongwe antenatal clinics. We evaluated interventions that reduced the per-contact transmission probability to 0·00003 in those receiving them and varied the proportion of individuals receiving the intervention in each stage.
Findings
We estimated that 38·4% (95% credible interval: 18·6%-52·3%) of ongoing HIV transmissions in Lilongwe are attributable to sexual contact with EHI index cases. Interventions acting only during EHI substantially reduced HIV prevalence, but did not lead to elimination, even with 100% coverage. Interventions acting only during CHI also reduced HIV prevalence, but coverage levels of 95%-99% were required to move the epidemic toward elimination. In scenarios with <95% CHI coverage, additional interventions reaching 25%-75% of EHI cases reduced HIV prevalence substantially.
Interpretation
Our results suggest that EHI plays an important role in HIV transmission in this sub-Saharan African setting. Without near-perfect coverage, interventions during CHI will likely have incomplete effectiveness unless complemented by strategies targeting the heightened transmission risk of EHI.
Acute HIV infection (AHI) is the period between HIV acquisition and the development of detectable antibodies against the virus. Early HIV infection (EHI), including AHI, is characterized by rapid viral replication, intense immune response and immune destruction, and viral diversification.1 Phenotypic factors unique to the founder virus(es) causing infection,2 along with exceptionally high viral loads,3, 4 result in greater transmission risk during EHI.5-8
The population-level effect of transmission prevention efforts during EHI will vary across settings, depending on the contribution of EHI cases to epidemic spread. Estimates of this contribution have varied widely,6, 9-18 depending on site-specific factors such as risk behaviour patterns and the local epidemic stage. Recently, a strategy to use mass antiretroviral treatment to stop the spread of HIV-1 has gained considerable attention.19 The “test and treat” strategy advocates regular, widespread HIV testing and immediate initiation of antiretroviral therapy (ART) for infected persons, based on the premise that ART-induced reductions in viral load will decrease infectiousness. However, the effectiveness of this approach will depend, in part, on the role of EHI in ongoing transmission. Where EHI has a major role, the effectiveness of test-and-treat strategies will likely be limited unless patients with EHI can be detected and included. A recent mathematical modelling exercise19 that has sparked extensive controversy20-29 concluded that an annual test-and-treat strategy could eliminate HIV, but EHI was not rigorously addressed in this model.22, 24, 25
In this study, we estimate the contribution of EHI to HIV spread in Lilongwe, Malawi, where HIV-1 is hyperendemic and transmission is almost entirely through heterosexual contact. We also examine the population-level impact of prevention interventions affecting only EHI, only chronic (post-“early”) HIV infection (CHI), or both EHI and CHI, with a particular emphasis on estimating the impact of an annual test-and-treat strategy.19 To address these aims, we used empirical data from our studies in Lilongwe3, 30 to develop a mathematical model describing heterosexual HIV transmission. The results suggest that EHI plays an important role in the ongoing spread of HIV in Malawi, and that involvement of people with EHI will be required for maximal HIV prevention.
Modelling analyses proceeded in five steps (Figure 1): defining the model structure and equations (Step 1); conducting a Bayesian melding procedure to identify model parameter values most compatible with observed epidemic dynamics, given prior estimates for biological and behavioural parameters (Steps 2-4); and estimating the contribution of EHI and predicted intervention effects (Step 5). We describe our methods briefly below, with additional details in the Web Appendix.
Figure 1
Figure 1
Flow diagram of modelling methods
Model structure
We constructed a compartmental, deterministic model explicitly describing heterosexual partnership formation and dissolution (Figure 2). Following previous modifications15 on the classic pair-formation model,31 sexual contact was assumed to occur: 1) at a constant frequency within steady partnerships, 2) as casual, one-off contacts by paired individuals outside of steady partnerships, and 3) as casual, one-off contacts by unpaired individuals (singles). This structure captures phenomena that are important in the context of time-varying HIV infectivity: HIV is “trapped” within a pair while each infected member is monogamous, but HIV can “escape” upon partnership dissolution or sexual contact outside the pair. Additionally, pairs of uninfected individuals are “sheltered” from HIV while each partner is monogamous, but HIV can “enter” through outside contacts. Long-term concurrency is not captured, but the model allows consecutive partnerships and sporadic concurrency. Our data from Lilongwe suggest that long-term monogamy predominates in this population, with a minority engaging in narrowly spaced consecutive partnerships, sporadic concurrency, or long-term concurrency.30 Individuals entered the model as singles, and exited after an average sexual lifespan of 35 years, with additional AIDS-related mortality in the final infection stage.
Figure 2
Figure 2
Simplified diagram of model structure
To capture variation in transmission probabilities over time, we divided HIV into EHI, asymptomatic HIV, early AIDS, and late AIDS. We defined EHI as the initial one- to six-month period of elevated infectivity, based on the best available estimates of transmission rates by infection stage,6 calculated among HIV-serodiscordant couples in Rakai, Uganda.5 We divided this period into five intervals to allow changes in transmission probabilities related to evolving viral loads. Intervals 1-4 were each one week in duration to capture the initial viral load changes.3 We sampled interval 5 from a uniform distribution of 1·4 weeks to 5 months, corresponding to the total assumed EHI duration of ~1 to ~6 months.6 We represented the asymptomatic period as three equal intervals (intervals 6-8) of 1·8 to 3·2 years each to approximate observed survival time distributions in untreated cases.32 We based the durations of “early AIDS” (interval 9) and “late AIDS” (interval 10) on the Rakai analyses, specifying normal distributions with mean 0·75 year and 0·83 year, respectively.6
As an additional extension, we stratified the model population into two groups to accommodate sexual behaviour heterogeneity. The prior distributions that we specified corresponded to the “lower-risk group” having longer partnerships and lower rates of sexual contact by singles than the “higher-risk group.” Individuals remained in one risk group and steady pairs were formed within groups. However, one-off contacts with casual partners were chosen without risk-group restrictions. The model allowed for increased HIV transmission probabilities for contacts with higher-risk partners, representing an assumed greater likelihood of transmission-amplifying cofactors, such as ulcerative sexually transmitted infections (STIs) or anal intercourse. Sexual behaviour parameters were held constant across infection status/stage categories. The model is readily adaptable to other settings with predominately heterosexual or homosexual transmission; adaptation to settings where parenteral transmission plays a large role is more complex.
Statistical analyses
We used a Bayesian melding approach32, 33 to fit the model to empirical HIV prevalence data and to account for uncertainty in model inputs and outputs (Figure 1). This approach combines prior information about inputs (e.g., sexual behaviour, transmission probabilities) with data about a primary output (HIV prevalence). Sources of prior information on inputs are described below. For data on model output, we used HIV prevalence estimates from Lilongwe antenatal clinics (ANC) over the period 1987-2005.34 We implemented a sample-importance-resample algorithm to identify the input parameter values producing epidemic curves most closely matching ANC data. Briefly, we ran 100,000 model simulations, sampling randomly from the prior distributions of all input parameters in each simulation. Next, we weighted each simulation according to its likelihood-based compatibility with ANC data, and then we resampled (with replacement) from the simulations, with probability of selection proportional to the assigned weight. Under this approach, the simulation resampled most frequently (the mode) is the best-fitting simulation (i.e., the most compatible with empirical HIV prevalence data). The 2·5th and 97·5th percentiles were used to obtain 95% credible intervals (CrI) for model parameters and output.
We used the Runge-Kutta 4 algorithm in Berkeley Madonna 8·0·1 (Berkeley, California, USA) to solve the model numerically. We performed statistical and graphical analyses with SAS 9·1 (SAS Institute, Cary, North Carolina, USA), Stata 9 (StataCorp, College Station, Texas, USA), and R.
Parameter values
We based the initial size of the adult (ages 15-49) population on census data35 (Table 1), and sampled uniformly in the range 1960-1985 for the year of the first HIV case in Lilongwe, reflecting the uncertainty around this time.
Table 1
Table 1
Input parameter definitions, prior distributions, and posterior distributions
We based the per-contact transmission probability for asymptomatic HIV, along with transmission co-factor effects, on meta-analysis estimates.7, 8 We allowed the transmission co-factor effect to range from 1 (no transmission amplification) to 6 (per-contact transmission probability 6 times as high in the higher-risk group than in the lower-risk group). The upper bound was based on estimates for the co-factor effect of concomitant STIs,7, 8 but could also broadly represent other co-factors, such as the possibility of some contacts being anal rather than vaginal. To calculate per-contact transmission probabilities for EHI and early AIDS, we multiplied the asymptomatic-period estimate by the transmission rate ratios comparing EHI and early AIDS to asymptomatic infection in Rakai (Table 1).6 We then used longitudinal viral load data from AHI cases in Lilongwe,3 along with transmission rates according to viral load,36 to estimate separate transmission probabilities within each EHI interval, subject to the constraint that the weighted average transmission probability across intervals equaled the overall EHI-period average. To match the lack of transmission events observed during late AIDS in the Rakai data (likely due to ceasing of sexual activity due to illness),6 we set the per-contact transmission probability for this period to 0.
We estimated sexual behaviour parameters from data we collected at Kamuzu Central Hospital STI Clinic in Lilongwe.3, 30 These data included detailed information about partnership durations and contact frequency by marital status and partner type.
Estimating proportion new infections from EHI
We calculated the annual proportion of new infections attributable to EHI transmitters from model equations tracking cumulative infections by calendar time and index infection period.
Predicting intervention effects by stage of initiation
We explored the potential effects of an HIV prevention intervention that was assumed to dramatically decrease the per-contact transmission probability in all contacts affected by the intervention. Such an intervention could be strictly behavioural (e.g., 100% condom use by the male partner), strictly biological (e.g., ART use by infected cases), or a combination of the two. The per-contact transmission probability in such contacts was reduced to 0·000033. This value is the midpoint of male-to-female and female-to-male transmission probabilities estimated under ART-induced viral suppression,37 but can also approximate effective condom use or other highly effective interventions. The intervention was not assumed to affect pair formation or dissolution rates, nor the frequency of sexual contact within or outside of steady partnerships. We varied the intervention coverage in each period, i.e., the proportion of cases (0%, 25%, 50%, 75%, 85%, 90%, 95%, 100%) in whom transmission probabilities were successfully reduced, to explore scenarios with intervention effects during EHI only (as a benchmark), CHI only, or both EHI and CHI.
To compare the maximum possible benefits of interventions in each period, we assumed that interventions began early in a given period. Interventions during EHI were assumed to start in week 3 and to end at the start of CHI. This assumption, which is based on our experience with AHI detection in Lilongwe,3 allows for blood collection in the second week of infection and an additional week to report positive HIV RNA or p24 results. Interventions acting in CHI were assumed to start at the beginning of the earliest CHI interval (interval 6) and continue through AIDS. Although this assumption reflects HIV diagnosis earlier than most current diagnoses,38 it approximates the time at which a highly effective test-and-treat program with annual HIV antibody tests would detect cases.19 Under such a strategy, infected individuals would be detected (on average) ~6 months into infection, approximately the time at which CHI starts in our model.
Interventions were assumed to begin in 2010 (at a mature epidemic phase). Our main intervention effect measures were the predicted HIV prevalence and incidence over the years 2010-2040. We also calculated the levels of coverage required for “elimination” of HIV during this time frame, using two separate definitions: 1) the “Granich definition” (reduction in annual incidence to less than 1 case per 1000 persons);19 and 2) the “Dahlem definition” (reduction in annual incidence to 0 cases), developed at the WHO-sponsored Dahlem Workshop on the Eradication of Infectious Diseases.39 As a complementary analysis, we predicted the percentage of new infections averted between 2010 and 2015.
As intervention effects on index case life expectancy could range from minimal (with strictly behavioural interventions) to substantial (with ART), we modelled two extremes of CHI interventions: a) no effect on infection duration, and b) increased average infection duration by ~10 to ~15 years.19, 40 The life expectancy increase in the latter scenario is based on survival-time estimates for individuals starting ART at the beginning of CHI.19, 40 Therefore, the CHI-only intervention with increased life expectancy approximates an annual test-and-treat strategy.
For base-case analyses, we used the modal input parameters. To explore intervention effects in situations with a greater or lesser importance of EHI, we used the input parameters producing the upper and lower 95% credible limits of the proportion of new cases attributable to EHI in 2010.
Additional sensitivity and influence analyses (described in more detail in Web Appendix)
To consider intervention effects under a variety of alternate conditions, we conducted analyses with greater assumed life expectancy increases among those receiving ART, as well as later intervention starts within EHI and/or CHI. We also explored the individual influence of each input parameter on results, examined epidemic dynamics within the high-risk and low-risk groups separately, and considered a model in which sexual behaviour parameters could change over the course of the epidemic.
Role of the funding source
The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.
Model Fit to HIV Prevalence Data
The Bayesian melding procedure yielded a posterior distribution of 380 unique epidemic curves in good agreement with the ANC data used to define HIV prevalence in Lilongwe (Figure 3). From the modal simulation, we estimated that HIV prevalence in Lilongwe’s general adult population peaked at 24·7% in 1996, declined to 17·3% in 2005, and is currently 14·3%. ANC prevalence estimates peaked at 27·0% (95% confidence interval: 22·7%, 31·6%) in 1996 and declined to 18·6% (95% confidence interval: 16·0%, 21·3%) at the most recent survey in 2005.
Figure 3
Figure 3
HIV ANC prevalence data and posterior distribution of output prevalence curves
The parameters producing the best fit to ANC data included an estimate that 40% of the population was in the initial “high-risk” group (Table 1); this proportion decreased to ~20% as the epidemic progressed. The best-fitting partnership durations were 2·5 years and 1·3 months in the low-risk and high-risk groups, respectively, with gaps of 1·5 years and 11 days between partners. Per-contact transmission probabilities were five times as great in the higher-risk versus lower-risk group (Table 1), representing transmission co-factor effects.
For most input parameters (e.g., viral load, transmission probabilities), the specified prior distributions were similar to the posterior distributions resulting from the fit to empirical data (see Web Appendix). The model fit was most informative for the size of the high-risk group, transmission co-factor effects, and high-risk pair formation and dissolution rates.
Estimated EHI contribution and related parameters
Based on the model results, 38·4% (95% CrI: 18·6%, 52·3%) of incident HIV infections resulted from contact with a partner with EHI in 2010 (Figure 4). The best-fitting EHI duration was 4·8 months (95% CrI: 1·6, 5·8) (Table 1). The best-fitting transmission rate ratio comparing EHI to asymptomatic chronic infection was 30 (95% CrI: 14, 47); that is, the weighted average per-contact transmission probability during EHI was 30 times as high as during asymptomatic infection (Table 1). The modal transmission rate ratio, in conjunction with other modal parameters, corresponds to per-contact transmission probabilities during EHI ranging from 3/1000 to 4/100 in the low-risk group (Table 1), and a cumulative transmission probability during EHI of ~26% within low-risk, HIV-discordant pairs (result not shown in Table 1).
Figure 4
Figure 4
Estimated proportion of incident HIV infections attributable to contact with EHI index case
Intervention effects
Biological or behavioural interventions assumed to sharply reduce HIV transmission probabilities during EHI (with no residual effect thereafter) were predicted to substantially reduce HIV prevalence in Lilongwe, but not to result in HIV elimination, even when 100% of EHI cases received the intervention (Figure 5A, solid green line). As the EHI-only intervention was not assumed to affect life expectancy, changes in HIV prevalence (Figure 5A) and incidence (Figure 5B) were similar.
Figure 5
Figure 5
Figure 5
Predicted effects of interventions during EHI only, CHI only, or both periods on HIV prevalence and incidence
In several scenarios, lifelong CHI-only interventions initiated ~6 months after infection were predicted to have greater effects on HIV prevalence and incidence than EHI-only interventions, but results were sensitive to assumptions about life expectancy (Figures 5C-5F). HIV elimination within 30 years under the “Dahlem definition” (reduction in annual incidence to 0 cases39) was not possible at any CHI-only coverage level, and elimination under the “Granich definition” (reduction in annual incidence to less than 1 case per 1000 persons19) was only possible with 95% or 99% CHI-only coverage, assuming no life expectancy increase or a 10-15-year increase, respectively (Figure 6). If the intervention was assumed to increase life expectancy (e.g., treatment as prevention19, 41), CHI-only coverage levels of 75% or less were predicted to increase HIV prevalence above the level that would be expected with no intervention (Figure 5E).
Figure 6
Figure 6
Intervention Coverage Levels Required for HIV Elimination in Lilongwe
The combination of EHI and CHI interventions reduced HIV prevalence and incidence substantially, even if CHI intervention coverage was only 75%-85% (Figures 5G-5N, dotted lines). Though the intervention coverage in each stage required for elimination depends on the definition of elimination and the intervention effects on life expectancy (Figure 6), the combination of results shown in Figures 5 and and66 suggests that if very high levels of coverage (i.e., ≥95%) cannot be attained with a CHI-only intervention, then the addition of EHI interventions will be necessary to sustainably reduce HIV prevalence and incidence in Lilongwe.
We observed these same patterns in sensitivity analyses where EHI was assumed to contribute 18·6% or 52·3% (our lower and upper 95% credible limits) of ongoing transmissions. Complementary analyses of infections averted showed similar trends. Sensitivity analyses in which CHI interventions increased life expectancy by 10-20 additional years provided further support for inclusion of EHI interventions, as did analyses where CHI interventions were assumed to start ~2 years – rather than ~6 months – into infection. (See Web Appendix for results of sensitivity analyses and other complementary analyses.)
The magnitude of the HIV epidemic in sub-Saharan Africa has been difficult to explain. One explanation for the extensive spread of HIV emphasizes the importance of persons with acute and early HIV infection, who are highly infectious but rarely aware of their infection status.42 Amplified transmission in this stage of infection has been ascribed to high viral loads3, 4 and an apparent increase in viral infectivity.43 In rhesus macaques, the ratio of infectious virions to total virions is up to 750 times as high during AHI as during CHI.43 In Ugandan HIV-serodiscordant couples, transmission rates were approximately 25 times as high during early versus asymptomatic infection.6 Phylogenetically-defined infection clusters16-18, 44 and documented acute-to-acute transmission events45, 46 further support the potential importance of early infection.
We undertook the current study to understand the contribution of EHI to the HIV epidemic in Lilongwe, Malawi, where we have conducted studies to identify patients with AHI,47-49 characterize their sexual behaviours,50 and measure viral load changes during EHI.3 We estimated that EHI index cases are responsible for 19%-52% of HIV transmissions in Lilongwe, with a mode of 38%. Our results suggest that the initial period of elevated transmissibility may be fairly long (~5 months), and that transmissibility during EHI is 30 times as great as during CHI.
Mathematical modelling estimates of the importance of EHI have varied widely, 6, 9, 10, 12-15, 31, 51 due in part to differing assumptions and a paucity of data for parameter definitions. Endemic-phase estimates of the proportion of new infections due to EHI in sub-Saharan Africa have ranged from 7% to 31% in modelling studies.6, 13 Phylogenetic studies conducted among MSM or mixed populations in western settings have estimated that 25%-49% of incident infections are due to EHI.16-18, 44
A strength of our model is the extensive use of local data for model parameterization. We used sexual behaviour data from Lilongwe30 to define contact patterns specific to the setting of interest and viral load data from AHI patients in Lilongwe3 to provide high resolution in the time-course of transmission probabilities during EHI. We fitted the model to local HIV prevalence data using a Bayesian melding approach. The close parameterization of the model based on data from the setting of interest allows us to expect results with greater reliability and applicability – at least in Lilongwe and similar settings – than could be expected from a model using a combination of parameter values derived from disparate populations.
Our model was enhanced in several other ways in comparison to previous models addressing AHI/EHI. To model contact patterns relevant to HIV transmission, we explicitly incorporated both steady pairs and casual contacts, with consideration of both high-risk and low-risk groups. The Bayesian melding approach allowed us to account for input and result uncertainty. Finally, we conducted sensitivity analyses to assess intervention effects under a range of predicted EHI contributions.
The idea of using ART as a transmission prevention strategy has gained remarkable attention.19, 41 This idea emerges from the reduction of HIV replication in the genital tracts of persons on ART,52, 53 and apparent suppression of HIV transmission in serodiscordant couples when the index case receives ART.54-56 A widely cited mathematical model has concluded that annual test-and-treat strategies could virtually eliminate the epidemic in South Africa,19 but the importance of EHI appears to have been underestimated in that model,22, 24, 25 and other modelling studies examining the potential benefits of ART have been less optimistic about the test-and-treat approach.57, 58
Accordingly, we examined the effects of behavioural and/or biomedical interventions that might drastically reduce sexual transmission at different stages of HIV. One such intervention was assumed to improve survival and reduce transmissibility from the onset of CHI, approximating an annual test-and-treat strategy.19 Our results suggest that even highly effective behavioural and/or biological interventions – including “test-and-treat” – are unlikely to eliminate HIV in Lilongwe and similar settings unless people with EHI are included. Even if the contribution of EHI to ongoing transmission is as low as ~20% (the lower credible limit in our analysis), intervention only during CHI is unlikely to eliminate HIV unless nearly all CHI cases experience life-long transmission suppression. If CHI-only intervention coverage is imperfect, additional EHI interventions can lead to dramatic improvement. Our results suggest that strategies preventing transmission from both CHI and EHI cases provide the greatest chance for marked, durable reductions in HIV incidence and prevalence. Sensitivity analyses in which CHI interventions were assumed to provide even greater survival benefits, or to start at times more typical of current clinical practice, provided even greater support for inclusion of interventions during EHI.
Interventions directed toward patients with EHI have unique challenges. While antibody tests may detect some post-acute EHI cases, a “test-and-treat” approach for specifically identifying EHI cases would require a very brief interval for repeat testing (~3-6 months), and reliance on antibody tests would result in missed AHI cases. Large-scale programs of quarterly or semi-annual HIV testing would be difficult to implement and sustain. Both biological and behavioural interventions intended for EHI may require more targeted approaches, such as partner notification or campaigns aimed at encouraging HIV testing among individuals with recent risky behaviour and acute retroviral symptoms.59 These case-finding strategies, in combination with pooling of blood samples,47 targeted HIV RNA screening,48, 60 and/or newer HIV detection tests1, 61 could increase the numbers of EHI cases detected, even in resource-limited settings. Encouragingly, our best-fitting value of 4·8 months for the period of elevated transmissibility suggests that interventions provided during the first few months of infection, rather than the first few weeks, may have substantial public health benefit. In earlier work we demonstrated that the HIV concentration in seminal plasma remained elevated for more than two months after infection, consistent with this idea.3
Interventions initiated during EHI may also have unique benefits that are not explicitly captured in our model. Adherence to biological or behavioural interventions initiated during EHI may remain high at least through the most infectious period,50 potentially maximizing cost-effectiveness and minimizing the detrimental effect of waning adherence observed with some interventions.62, 63 Additionally, at least some studies suggest a clinical benefit from initiation of treatment during EHI.64
Mathematical models of HIV transmission depend heavily on assumptions about sexual behaviour. The input values for sexual behaviour parameters in our model were based on a cross-sectional study conducted among STD clinic patients in Lilongwe30 that found long-term monogamy to be common, with only 14% reporting long-term concurrency, sporadic concurrency, or consecutive, monogamous partnerships in rapid succession. Although based on a relatively small sample (n=186) with generalizability likely limited to similar settings, these data represent some of the most detailed information available on partnership durations and gaps in sub-Saharan Africa. In our best-fitting model simulation, “steady” partnerships in the higher-risk group were 1.3 months in duration, with intervening gaps of 11 days; by contrast, steady partnership durations and gaps in the lower-risk group were 2.5 years and 1.5 years, respectively. Therefore, the difference between “steady” and “casual” partners appears to be considerably less distinct for higher-risk than lower-risk individuals, potentially explaining why some behavioural parameters resulting from the model fit vary in somewhat unexpected ways across groups. For example, despite the higher casual contact rate that we initially posited for higher-risk singles, the best-fitting parameter set suggests a lower rate, possibly because the shorter gap between “steady” partnerships translates to less time as a single in the high-risk group. We also note that the partner change rate in our higher-risk group was slower than in the highest-risk groups of several previous HIV epidemic models,13, 65 and that none of our behavioural parameters was particularly extreme, despite being collected in an STI clinic population. The latter result may be due to the high prevalence of STIs in Malawi,66, 67 which likely results in considerable overlap between STI clinic populations and the “general” population.
The importance of sexual partner concurrency in the HIV epidemic of sub-Saharan Africa has been emphasized and debated.68-72 Intuitively, concurrency seems potentially important for transmission during EHI, because long-term monogamy would limit the high transmissibility of newly infected persons to a single partnership. Our model captured only a simple form of concurrency – one-off encounters outside of pairs; it did not include long-term concurrency. However, our best-fitting parameter set included short gaps between partners (11 days) in a relatively sizeable higher-risk group. These short gap lengths are consistent with (and based on) our data from Malawi,30 and provide an alternate explanation for rapid HIV spread and the corresponding importance of EHI; however, the potential contribution of concurrency cannot be excluded.
All mathematical models have limitations. In our model, individuals and pairs were restricted to a given risk group, and only a very simple form of concurrency was captured, as noted above. Additionally, our model did not incorporate population age structure or male-female behavioural asymmetry. The inclusion of behavioural heterogeneity across two separate risk groups may capture some age-related behavioural variation, but the results give an average picture for the sexually active population overall. Nevertheless, our division of EHI into numerous intervals, our inclusion of more than one risk group, and our incorporation of both steady and casual contacts likely reflect transmission dynamics more accurately than previous models that have assumed only one-off contacts occurring at random within populations. Additional considerations, such as drug resistance, side effects, behavioural disinhibition, and cost, must also be carefully appraised before implementing specific treatment-based interventions.
In summary, our analyses suggest that EHI remains a critical factor in the ongoing HIV epidemic in Lilongwe. This result suggests that acute and early HIV infection can be important not only in the earliest phases of HIV epidemics, but also in more mature epidemics. Consequently, prevention approaches directed at all stages of HIV will likely be necessary to ensure a durable effect on the epidemic in Lilongwe and similar settings. As plans for “treatment as prevention” are developed, our results suggest that strategies for detection and management of patients with acute and early HIV must be included.
Systematic review
In a recent review of mathematical models estimating the contribution of AHI/EHI to ongoing HIV transmission,73 we noted that estimates have varied widely, due to differences in epidemic stage, populations studied, model structure, parameter values, and assumptions in the absence of site-specific data. None of these modelling studies compared the potential impact of transmission prevention interventions initiated in acute/early HIV versus chronic infection. Previous modelling studies examining “treatment as prevention” strategies for established infection19,57,58,65 have had mixed results related to the control of HIV.74
Interpretation
The events that transpire during acute HIV infection are critical to the health of the infected individual and the health of the public.1 Our study suggests that transmission prevention interventions achieving intermediate levels of coverage during both early and chronic HIV will have a far greater impact on the spread of HIV in Malawi than interventions focused on chronic HIV alone. As such, the population-level impact of “test-and-treat” strategies is likely to be optimized only if individuals with acute/early HIV are included or if complementary interventions targeting the earliest phases of infection are incorporated into the overall prevention approach.
Supplementary Material
01
Acknowledgments
We thank Christophe Fraser for suggestions on the mathematical modelling methods, and Ward Cates, Jim Shelton, and David Burns for providing helpful comments on an earlier draft of this manuscript.
This research was supported by the National Institutes of Health (R01 MH068686, R37 DK049381, R01 AI083059, 5T32AI070114) and the University of North Carolina at Chapel Hill Center for AIDS Research (CFAR), an NIH funded program (#P30 AI50410).
Footnotes
Contributors KAP, ACG, WCM, IFH, and MSC participated in study conception and design. KAP, ACG, and WCM participated in model development and analysis. KAP, ACG, WCM, IFH, AEP, FEAM, GK, and MSC contributed to data collection or interpretation. KAP drafted the manuscript, with critical revision by ACG, WCM, IFH, AEP, and MSC. All authors saw and approved the final version of the manuscript.
Conflicts of interest We declare that we have no conflict of interest.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1. Cohen MS, Shaw GM, McMichael AJ, Haynes BF. Acute HIV-1 Infection -- Basic, Clinical and Public Health Perspectives. New Engl J Med. 2011 In press.
2. Vaidya NK, Ribeiro RM, Miller CJ, Perelson AS. Viral dynamics during primary simian immunodeficiency virus infection: effect of time-dependent virus infectivity. J Virol. 2010;84:4302–4310. [PMC free article] [PubMed]
3. Pilcher CD, Joaki G, Hoffman IF, et al. Amplified transmission of HIV-1: comparison of HIV-1 concentrations in semen and blood during acute and chronic infection. AIDS. 2007;21:1723–1730. [PMC free article] [PubMed]
4. Fiebig EW, Wright DJ, Rawal BD, et al. Dynamics of HIV viremia and antibody seroconversion in plasma donors: implications for diagnosis and staging of primary HIV infection. AIDS. 2003;17:1871–1879. [PubMed]
5. Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis. 2005;191:1403–1409. [PubMed]
6. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 Transmission, by Stage of Infection. J Infect Dis. 2008;198:687–693. [PubMed]
7. Powers KA, Poole C, Pettifor AE, Cohen MS. Rethinking the heterosexual infectivity of HIV-1: a systematic review and meta-analysis. Lancet Infect Dis. 2008;8:553–563. [PMC free article] [PubMed]
8. Boily MC, Baggaley RF, Wang L, et al. Heterosexual risk of HIV-1 infection per sexual act: systematic review and meta-analysis of observational studies. Lancet Infect Dis. 2009;9:118–129. [PubMed]
9. Jacquez JA, Koopman JS, Simon CP, Longini IM., Jr Role of the primary infection in epidemics of HIV infection in gay cohorts. J Acquir Immune Defic Syndr. 1994;7:1169–1184. [PubMed]
10. Koopman JS, Jacquez JA, Welch GW, et al. The role of early HIV infection in the spread of HIV through populations. J Acquir Immune Defic Syndr Hum Retrovirol. 1997;14:249–258. [PubMed]
11. Pinkerton SD. Probability of HIV Transmission During Acute Infection in Rakai, Uganda. AIDS Behav. 2007;7:7. [PMC free article] [PubMed]
12. Pinkerton SD. How many sexually-acquired HIV infections in the USA are due to acute-phase HIV transmission? AIDS. 2007;21:1625–1629. [PMC free article] [PubMed]
13. Abu-Raddad LJ, Longini IM., Jr No HIV stage is dominant in driving the HIV epidemic in sub-Saharan Africa. AIDS. 2008;22:1055–1061. [PubMed]
14. Prabhu VS, Hutchinson AB, Farnham PG, Sansom SL. Sexually acquired HIV infections in the United States due to acute-phase HIV transmission: an update. AIDS. 2009;23:1792–1794. [PubMed]
15. Xiridou M, Geskus R, de Wit J, Coutinho R, Kretzschmar M. Primary HIV infection as source of HIV transmission within steady and casual partnerships among homosexual men. AIDS. 2004;18:1311–1320. [PubMed]
16. Brenner BG, Roger M, Routy JP, et al. High rates of forward transmission events after acute/early HIV-1 infection. J Infect Dis. 2007;195:951–959. [PubMed]
17. Pao D, Fisher M, Hue S, et al. Transmission of HIV-1 during primary infection: relationship to sexual risk and sexually transmitted infections. AIDS. 2005;19:85–90. [PubMed]
18. Yerly S, Vora S, Rizzardi P, et al. Acute HIV infection: impact on the spread of HIV and transmission of drug resistance. AIDS. 2001;15:2287–2292. [PubMed]
19. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with immediate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet. 2009;373:48–57. [PubMed]
20. Garnett GP, Baggaley RF. Treating our way out of the HIV pandemic: could we, would we, should we? Lancet. 2009;373:9–11. [PubMed]
21. Holtgrave DR. Potential and limitations of a ‘test and treat’ strategy as HIV prevention in the United States. Int J Clin Pract. 2010;64:678–681. [PubMed]
22. Ruark A, Shelton JD, Halperin DT, Wawer MJ, Gray RH. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1078. [PubMed]
23. Assefa Y, Lera M. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1080. [PubMed]
24. Cohen MS, Mastro TD, Cates W., Jr Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1077. [PubMed]
25. Epstein H. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1078–1079. [PubMed]
26. Jaffe H, Smith A, Hope T. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1080. [PubMed]
27. Jurgens R, Cohen J, Tarantola D, Heywood M, Carr R. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1079. [PubMed]
28. Hsieh YH, de Arazoza H. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1079–1080. [PubMed]
29. Wilson DP. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet. 2009;373:1077–1078. [PubMed]
30. Powers KA, Miller WC, Hoffman IF, et al. Sexual partnership patterns in Malawi: implications for HIV/STI transmission. Sex Transm Dis. 2011 e-published ahead of print. [PMC free article] [PubMed]
31. Kretzschmar M, Dietz K. The effect of pair formation and variable infectivity on the spread of an infection without recovery. Math Biosci. 1998;148:83–113. [PubMed]
32. Hallett TB, Gregson S, Mugurungi O, Gonese E, Garnett GP. Assessing evidence for behaviour change affecting the course of HIV epidemics: A new mathematical modelling approach and application to data from Zimbabwe. Epidemics. 2009;1:108–117. [PubMed]
33. Alkema L, Raftery AE, Brown T. Bayesian melding for estimating uncertainty in national HIV prevalence estimates. Sex Transm Infect. 2008;84:i11–i16. [PMC free article] [PubMed]
34. USCensusBureau. HIV/AIDS Surveillance Database. http://hivaidssurveillancedb.org/hivdb/
35. Malawi Population and Housing Census. National Statistics Office; Zomba, Malawi: 2008.
36. Quinn TC, Wawer MJ, Sewankambo N, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med. 2000;342:921–929. [PubMed]
37. Wilson DP, Law MG, Grulich AE, Cooper DA, Kaldor JM. Relation between HIV viral load and infectiousness: a model-based analysis. Lancet. 2008;372:314–320. [PubMed]
38. Hosseinipour M, Neuhann F, Kanyama C, Namarika D, Phiri S. Mortality is the leading cause of loss to follow-up among antiretroviral patients in Malawi. 11th Conference on Retroviruses and Opportunistic Infections. 2004
39. Dowdle WR. The principles of disease elimination and eradication. Bull World Health Organ. 1998;76(Suppl 2):22–25. [PubMed]
40. Hallett TB, Gregson S, Dube S, Garnett GP. The impact of monitoring HIV patients prior to treatment in resource-poor settings: insights from mathematical modelling. PLoS Med. 2008;5:e53. [PMC free article] [PubMed]
41. Cohen MS, Gay C, Kashuba AD, Blower S, Paxton L. Narrative review: antiretroviral therapy to prevent the sexual transmission of HIV-1. Ann Intern Med. 2007;146:591–601. [PubMed]
42. Cates W, Jr, Chesney MA, Cohen MS. Primary HIV infection--a public health opportunity. Am J Public Health. 1997;87:1928–1930. [PubMed]
43. Ma ZM, Stone M, Piatak M, Jr, et al. High specific infectivity of plasma virus from the pre-ramp-up and ramp-up stages of acute simian immunodeficiency virus infection. J Virol. 2009;83:3288–3297. [PMC free article] [PubMed]
44. Lewis F, Hughes GJ, Rambaut A, Pozniak A, Leigh Brown AJ. Episodic sexual transmission of HIV revealed by molecular phylodynamics. PLoS Med. 2008;5:e50. [PMC free article] [PubMed]
45. Pilcher CD, Eron JJ, Jr, Vemazza PL, et al. Sexual transmission during the incubation period of primary HIV infection. JAMA. 2001;286:1713–1714. [PubMed]
46. Taylor MM, Rotblatt H, Brooks JT, et al. Epidemiologic investigation of a cluster of workplace HIV infections in the adult film industry: Los Angeles, California, 2004. Clin Infect Dis. 2007;44:301–305. [PubMed]
47. Fiscus SA, Pilcher CD, Miller WC, et al. Rapid, real-time detection of acute HIV infection in patients in Africa. J Infect Dis. 2007;195:416–424. [PubMed]
48. Powers KA, Miller WC, Pilcher CD, et al. Improved detection of acute HIV-1 infection in sub-Saharan Africa: development of a risk score algorithm. AIDS. 2007;21:2237–2242. [PMC free article] [PubMed]
49. Pilcher CD, Price MA, Hoffman IF, et al. Frequent detection of acute primary HIV infection in men in Malawi. AIDS. 2004;18:517–524. [PubMed]
50. Pettifor A, Macphail C, Corneli A, et al. Continued High Risk Sexual Behavior Following Diagnosis with Acute HIV Infection in South Africa and Malawi: Implications for Prevention. AIDS Behav. 2010 Oct 27; epub ahead of print. [PMC free article] [PubMed]
51. Pinkerton SD, Abramson PR. Implications of increased infectivity in early-stage HIV infection. Application of a Bernoulli-process model of HIV transmission. Eval Rev. 1996;20:516–540. [PubMed]
52. Graham SM, Holte SE, Peshu NM, et al. Initiation of antiretroviral therapy leads to a rapid decline in cervical and vaginal HIV-1 shedding. AIDS. 2007;21:501–507. [PubMed]
53. Vernazza PL, Troiani L, Flepp MJ, et al. Potent antiretroviral treatment of HIV-infection results in suppression of the seminal shedding of HIV. The Swiss HIV Cohort Study. AIDS. 2000;14:117–121. [PubMed]
54. Bunnell R, Ekwaru JP, Solberg P, et al. Changes in sexual behavior and risk of HIV transmission after antiretroviral therapy and prevention interventions in rural Uganda. AIDS. 2006;20:85–92. [PubMed]
55. Del Romero J, Castilla J, Hernando V, Rodriguez C, Garcia S. Combined antiretroviral treatment and heterosexual transmission of HIV-1: cross sectional and prospective cohort study. BMJ. 2010;340:c2205. [PMC free article] [PubMed]
56. Donnell D, Baeten JM, Kiarie J, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet. 2010;375:2092–2098. [PMC free article] [PubMed]
57. Smith RJ, Okano JT, Kahn JS, Bodine EN, Blower S. Evolutionary dynamics of complex networks of HIV drug-resistant strains: the case of San Francisco. Science. 2010;327:697–701. [PubMed]
58. Dodd PJ, Garnett GP, Hallett TB. Examining the promise of HIV elimination by ‘test and treat’ in hyperendemic settings. AIDS. 2010;24:729–735. [PMC free article] [PubMed]
59. Schacker TW, Hughes JP, Shea T, Coombs RW, Corey L. Biological and virologic characteristics of primary HIV infection. Ann Intern Med. 1998;128:613–620. [PubMed]
60. Miller WC, Leone PA, McCoy S, Nguyen TQ, Williams DE, Pilcher CD. Targeted testing for acute HIV infection in North Carolina. AIDS. 2009;23:835–843. [PMC free article] [PubMed]
61. Cohen MS, Gay CL, Busch MP, Hecht FM. The detection of acute HIV infection. J Infect Dis. 2010;202(Suppl 2):S270–277. [PubMed]
62. Coates TJ, Richter L, Caceres C. Behavioural strategies to reduce HIV transmission: how to make them work better. Lancet. 2008;372:669–684. [PMC free article] [PubMed]
63. Wakabi W. New strategies sought in Uganda as HIV infections rise. Lancet Infect Dis. 2008;8:285. [PubMed]
64. Bell SK, Little SJ, Rosenberg ES. Clinical management of acute HIV infection: best practice remains unknown. J Infect Dis. 2010;202(Suppl 2):S278–288. [PubMed]
65. Baggaley RF, Garnett GP, Ferguson NM. Modelling the impact of antiretroviral use in resource-poor settings. PLoS Med. 2006;3:e124. [PMC free article] [PubMed]
66. Malawi Demographic and Health Survey. Calverton, Maryland: 2004.
67. Glynn JR, Crampin AC, Ngwira BM, Ndhlovu R, Mwanyongo O, Fine PE. Herpes simplex virus type 2 trends in relation to the HIV epidemic in northern Malawi. Sex Transm Infect. 2008;84:356–360. [PMC free article] [PubMed]
68. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS. 1997;11:641–648. [PubMed]
69. Eaton JW, Hallett TB, Garnett GP. Concurrent Sexual Partnerships and Primary HIV Infection: A Critical Interaction. AIDS Behav. 2010 Oct 2; epub ahead of print.
70. Lurie M, Rosenthal S, Williams B. Concurrency driving the African HIV epidemics: where is the evidence? Lancet. 2009;374:1420. [PubMed]
71. Kim JH, Riolo RL, Koopman JS. HIV transmission by stage of infection and pattern of sexual partnerships. Epidemiology. 2010;21:676–684. [PMC free article] [PubMed]
72. Epstein H. The mathematics of concurrent partnerships and HIV: a commentary on Lurie and Rosenthal. AIDS Behav. 2009;14:29–30. [PubMed]
73. Miller WC, Rosenberg NE, Rutstein SE, Powers KA. The role of acute and early HIV infection in the sexual transmission of HIV. Current Opinion in HIV and AIDS. 2010;5:277–82. [PMC free article] [PubMed]
74. Cohen MS, Gay CL. Treatment to prevention transmission of HIV-1. Clin Infect Dis. 2010;50(S3):S85–S95. [PubMed]