Article
Reference
How to achieve the global 90-90-90 target by 2020 in sub-Saharan Africa? A mathematical modelling study
ESTILL, Janne Anton Markus, et al.
Abstract
The 90-90-90 target states that by 2020, 90% of people living with HIV should be diagnosed, 90% of those diagnosed treated, and 90% of those treated virally suppressed. We assessed the actions needed in each country of sub-Saharan Africa to achieve the 90-90-90 target.
ESTILL, Janne Anton Markus, et al. How to achieve the global 90-90-90 target by 2020 in sub-Saharan Africa? A mathematical modelling study. Tropical Medicine & International Health, 2018, vol. 23, no. 11, p. 1223-1230
DOI : 10.1111/tmi.13145 PMID : 30156355
Available at:
http://archive-ouverte.unige.ch/unige:115312
Disclaimer: layout of this document may differ from the published version.
1 / 1
How to achieve the global 90-90-90 target by 2020 in sub-Saharan
1
Africa? A mathematical modelling study
2
Running title: Achieving the 90-90-90 target 3
4
Janne ESTILL
1,2, Kimberly MARSH
3, Christine AUTENRIETH
3, Nathan FORD
45
1
Institute of Global Health, University of Geneva, Geneva, Switzerland
62
Institute of Mathematical Statistics and Actuarial Science (IMSV), University of Bern, Bern,
7Switzerland
83
Joint United Nations Programme on HIV/AIDS (UNAIDS), Geneva, Switzerland
94
Department of HIV/AIDS, World Health Organization (WHO), Geneva, Switzerland
1011
The study received funding from the Swiss National Science Foundation and the World
12Health Organization.
13 14
Correspondence to:
15
Janne Estill
16University of Bern, IMSV
17Alpeneggstrasse 22
183012 Bern, Switzerland
19Email: janne.estill@unige.ch
20Tel. +41 31 631 88 18
2122
Abstract 23
Objectives: The 90-90-90 target states that by 2020, 90% of people living with HIV should be 24
diagnosed, 90% of those diagnosed treated, and 90% of those treated virally suppressed. We 25
assessed the actions needed in each country of sub-Saharan Africa to achieve the 90-90-90 target.
26
Methods: We developed a mathematical model to assess the number of patients needing to start 27
ART between 2017 and 2020 to achieve 81% coverage by 2020 in each country, and the proportion 28
of treated patients who are virally suppressed in four scenarios, combining two scenarios of 29
retention (current-level or perfect), and routine viral load monitoring (current or universal 30
coverage). We performed two separate simulations, one using observed failure rates from cohort 31
studies, and one with considerably lower failure rates to set a theoretical lower limit.
32
Results: Our model projected that 2.9 million people started ART in 2017 in sub-Saharan Africa. If, 33
depending on scenario, at least 2.2 to 2.7 million patients continue to start ART annually, 81% ART 34
coverage will be reached in 2020 in sub-Saharan Africa on average. In 37% of the countries, a 35
multiple-fold increase in annual number of patients starting ART is needed. Virological 36
suppression >90% in 2020 could be reached only in the best-case scenario assuming low probability 37
of treatment failure, elimination of treatment interruptions, and universal routine viral load 38
monitoring.
39
Conclusion: The 90-90-90 target is realistic in sub-Saharan Africa on average, but not necessarily in 40
all individual countries. Each country should identify and focus on the specific gaps needing 41
attention.
42
Keywords: Africa south of the Sahara; mathematical model; cascade of care; HIV diagnosis;
43
antiretroviral therapy; viral load 44
Introduction 45
In 2014, UNAIDS established the global 90-90-90 treatment target: by 2020, 90% of all people living 46
with HIV (PLHIV) should be aware of their status, 90% of those testing HIV positive should be on 47
antiretroviral therapy (ART), and 90% of those on ART should be virally suppressed(1). The ultimate 48
aim of this ambitious target is to end the AIDS epidemic by 2030. Modelling studies have projected a 49
reduction of nearly 90% in both new HIV infections and AIDS related deaths until 2030 could be 50
possible if the target is reached on time(1).
51
Sub-Saharan Africa (SSA) bears the heaviest burden of the global HIV epidemic: in 2016, 69% of all 52
PLHIV worldwide resided in the region(2). ART was very limited in SSA until 2003, when the World 53
Health Organization (WHO) declared the epidemic a global emergency and a global programme was 54
established to scale up ART free of charge in the most affected countries. Since then, the number of 55
people on ART has increased rapidly, with 14.3 of the 20.9 million people on ART in 2017 living in 56
SSA(3).
57
Despite this remarkable progress, there are a number of challenges to reaching the 90-90-90 target, 58
notably limited HIV testing coverage, delays in starting treatment despite recent policy changes to 59
starting treatment as soon as HIV infection is confirmed, stock-outs of antiretroviral drugs (ARVs), 60
treatment interruptions, disengagement from care, and limited availability of viral load 61
monitoring(4–7). These challenges are common to many countries across SSA, but there are also 62
differences. Each country needs to adopt strategies that focus on overcoming the most important 63
challenges. Building on previous analyses by UNAIDS on HIV testing coverage(8) and mathematical 64
modelling of progression of HIV in patients receiving ART(9–12), we estimated what actions are 65
needed in each country of SSA to reach the 90-90-90 target by 2020.
66 67
Methods 68
To identify the largest gaps in the cascade from HIV infection to viral suppression, we assessed the 69
following for all SSA countries: 1) How many people need to start ART every year between 2018 and 70
2020 to achieve 81% coverage (90% coverage among the 90% of PLHIV aware of their status) among 71
all PLHIV by 2020? and 2) What needs to be done in terms of patient monitoring and retention to 72
achieve 90% suppression among those on ART by 2020?
73
We included a total of 41 countries, containing all member countries of the UNAIDS East and 74
Southern Africa and West and Central Africa regions, with the exceptions of five small countries with 75
very limited data (Cape Verde, Comoros, Mauritius, São Tomé and Prìncipe, and Seychelles). We first 76
reviewed the current situation of the epidemic and treatment cascade in each country of SSA. We 77
estimated the current cascade from the following data sources: programme data on diagnosed and 78
treated patients submitted to UNAIDS(2); modelling estimates from UNAIDS on the total infected 79
population(2); estimates from Demographic and Health Surveys (DHS) and AIDS Indicator Surveys 80
(AIS) on the proportion of infected people ever tested(13); and cascade estimates from the 81
Population-based HIV Impact Assessments (PHIA)(14). To estimate the total number of PLHIV in 82
2020, we collected the estimated number of PLHIV in 2016 and the average annual change over the 83
past 5 years. We assumed that the annual change would continue the same, and using this predicted 84
how many people would be living with HIV in 2020.
85
Simulation model 86
After the review, we applied a previously published ART forecasting model to each country(9). The 87
model simulates the progression of HIV from ART initiation until death in individual patients. Disease 88
progression is represented as a sequence of health states that reflect the patient’s current retention 89
status, treatment regimen, and virological and immunological treatment responses. The transition 90
rates between the health states have been parameterised primarily with data from routine ART 91
programmes in South Africa, Malawi, and Zambia, which are part of the International Epidemiology 92
Databases to Evaluate AIDS (IeDEA) collaboration’s Southern African region(15). The model is 93
implemented using the R package gems(16,17). This parameterisation, assuming a cumulative 94
probability of treatment failure of 6% at one year and 13% at 5 years from ART start, is likely to 95
overestimate the true suppression level in some settings. Therefore we also ran an alternative 96
simulation using higher virological failure rates, corresponding to 26% cumulative failure probability 97
at one and 51% at 5 years from ART initiation (supplementary Text S1).
98
We fitted the number of patients starting ART in the period 2003 to 2017 to the observed number of 99
patients on ART, using an Excel tool available online(9). For the years 2018-2020, the number of 100
patients starting ART was chosen so that 81% coverage among all PLHIV would be reached by the 101
end of 2020, and was assumed to be the same every year. This analysis was conducted for four 102
scenarios, combining two assumptions about retention on ART and switching to a second-line 103
regimen, and two assumptions about viral load monitoring (Table 1). With current retention and 104
switching, rates of drop-out and a delay between failure detection and switch to second-line were 105
assumed as in our previous modelling studies(11,18); with optimal retention and switching, we 106
assumed that there would be no treatment interruptions in the future and patients would switch to 107
second-line immediately after detection of failure. With targeted viral load monitoring, CD4 cell 108
counts were monitored annually for patients on ART and viral load test was used to confirm 109
treatment failure; with routine viral load monitoring, all patients were tested for viral load annually.
110
In Botswana, Malawi and South Africa routine viral load monitoring was assumed in all scenarios as 111
national guidelines have already adopted this approach. The analysis was based on estimates of all 112
PLHIV, but children were not explicitly considered.
113
Analysis of results 114
First, we compared the number of patients starting ART in years 2018-2020 to the 2017 estimate, to 115
determine the increase needed in ART scale-up. Second, we analysed the raw outputs of our 116
simulation model and calculated the proportion of patients in each simulated ART cohort who were 117
virally suppressed. We considered the following patients as viraemic: patients who started ART, 118
switched regimen or returned after a treatment interruption within the previous three months;
119
patients failing first-line ART before switching to second-line; or patients who failed second-line ART.
120
We performed the analysis using the main simulation for all countries, and in addition using the 121
simulation with higher virological failure rate for the countries with latest estimated suppression 122
<70% among treated patients.
123
We present the results concerning the needed scale-up for all countries separately. Since the 124
internal parameters of the model do not differ between countries, we only present the pooled 125
results on viral suppression, for the main analysis and the analysis using a higher failure rate.
126 127
Results 128
Review of current situation 129
Information on number of people living with HIV was available in the UNAIDS database for the year 130
2016 for all 41 reviewed countries. Complete cascades could be built for 24 countries. Proportion of 131
PLHIV aware of their status ranged between 7% and 87%, proportion of diagnosed patients on ART 132
between 56% and 100%, and proportion of treated patients who are virally suppressed between 133
11% and 94%. The largest gap in most countries was in awareness, although with some exceptions.
134
Moreover, the proportion who were diagnosed tended to be clearly higher in East and Southern 135
Africa. Complete results of the review are shown in the appendix (supplementary Table S1, 136
supplementary Figure S1).
137
Achieving the first two “90s” by 2020 138
According to our model, about 2.9 million people started ART in SSA in 2017, representing a 20%
139
increase from 2016. To achieve 81% coverage among all PLHIV by 2020, it is sufficient to keep the 140
average number of patients starting ART every year between 2018 and 2020 at 2.7 million, close the 141
2017 level (Table 2). In the optimal scenario, it would be sufficient to start 2.2 million patients 142
annually for ART to reach the target.
143
A comparison of individual countries showed substantial differences in the needed scale-up (Figure 144
1, Supplementary Table S2). Of the 41 assessed countries, 19 (46%) will reach the two first targets by 145
keeping the number of patients who start ART on the 2017 level until 2020. In seven countries 146
(17%), the ART initiation rate needs to be moderately scaled up, up to doubling the number of 147
patients starting ART annually. In the remaining 15 (37%) countries, a more substantial increase is 148
needed. Most of these countries are located in West and Central Africa. The situation is most critical 149
in Liberia, Madagascar, Mauritania and South Sudan, requiring more than a 10-fold increase.
150
Viral suppression among patients receiving ART 151
According to our model, about 78% of all patients on ART were virally suppressed in 2016. Predicted 152
viral suppression in 2020 was sensitive to assumptions about switching delay and treatment 153
interruptions and availability of routine viral load monitoring (Figure 2A). In the optimal scenario, the 154
predicted proportion of virally suppressed patients was 94.2% (routine viral load monitoring) or 155
89.8% (targeted viral load monitoring). In the current scenario, the corresponding proportions were 156
83.7% and 81.0%.
157
We applied the alternative analysis with higher failure rate into the following countries with current 158
viral suppression below 70%: Benin, Cameroon, Gabon, Lesotho, Liberia, Madagascar, Mali, Niger, 159
and Uganda. The model estimated viral suppression at about 54% in 2016. The proportion of 160
suppressed patients in 2020 was 51.6% in the current scenario and targeted viral load monitoring, 161
60.0% in the optimal scenario with targeted viral load monitoring, 64.0% in the current scenario with 162
routine viral load monitoring, and 76.9% in the optimal scenario with routine viral load monitoring 163
(Figure 2B; supplementary Table S2).
164 165
Discussion 166
Achieving the 90-90-90 target in SSA will continue to require investments across the cascade of care 167
also in the final two years, with priorities differing by country and region. Overall, keeping up the 168
current speed of ART scale-up should be sufficient to reach 81% ART coverage among all PLHIV by 169
2020 in SSA. However, there are substantial differences between countries. In some countries, the 170
target is already close and it is more important to focus on reducing drop-out and improving viral 171
suppression. In some countries, however, a substantial increase in the scale-up ART is needed.
172
In many East and Southern African countries, ART coverage is already close to the target. The 173
situation has improved during the last years. The cascade estimates we found in our review were 174
systematically higher that those in two recently published reviews, showing that several countries 175
are making progress in reducing the gap(19,20). Our model suggests that the target could be 176
reached in some countries even with a considerable decrease in the ART initiation rate. Although 177
this is not desirable, it may be a realistic scenario: if ART coverage is already high, it will become 178
more challenging to identify and treat the remaining PLHIV. These are likely to be among the harder- 179
to-reach populations, or population groups with generally low HIV prevalence. Screening for HIV in 180
such populations may be inefficient. It may be more efficient in these settings to focus on improving 181
retention and viral suppression among patients already in care, at the same time assuring universal 182
access to voluntary HIV testing and providing ART for all diagnosed patients.
183
The situation on ART coverage was critical especially in West and Central Africa. In the majority of 184
countries in this region, a multiple-fold increase in number of patients enrolling on ART annually is 185
needed. Based on the UNAIDS database, in most of these countries the main gap is from infection to 186
diagnosis(2). This suggests that massive investments are needed to identify the undiagnosed PLHIV, 187
at the same time assuring that all diagnosed patients can be treated in a timely manner. One major 188
reason for the differences between countries is likely the differing nature of the epidemic(21,22). In 189
East and Southern Africa, the epidemic is generalized, patients are predominantly young women 190
who are easy to reach through antenatal care, and due to the high prevalence HIV has been a 191
national priority. In West and Central Africa, overall prevalence is lower, and PLHIV are more 192
concentrated among population groups who are marginalized and may face stigma and whose 193
behaviour may be criminalized.
194
Achievement of 90% viral suppression among people on ART will be challenging. According to our 195
estimates, about 80% of the patients on ART were virally suppressed in 2016. A review of the 196
available data showed on average similar results, and our findings should thus be applicable to most 197
of the region, with the exception of some individual countries where notably lower suppression 198
rates were observed or predicted. Several countries may also have already reached the target 199
(Botswana, Malawi, Swaziland) or are close to it. In a secondary analysis we made new simulations 200
with a considerably higher failure rate, equivalent to 26% instead of 6% risk of failure one year after 201
treatment initiation. Since the failure rate is essentially a fixed parameter in our model, none of the 202
interventions could lift the suppression close to the target by 2020 in this analysis. The reasons for 203
the low suppression levels in some countries should be studied in detail to implement interventions 204
that can efficiently improve treatment response in the context of the particular setting. On the other 205
hand, failure rates lower than in our model have also been reported(23), showing that with an 206
adequate regimen and regular monitoring and adherence support, the viraemia can be kept well 207
under control.
208
We found that without improvements in monitoring, switching and retention, the proportion of 209
patients virally suppressed will stay around the current level. Routine viral load monitoring, 210
interventions to prevent treatment interruption and loss to care, and prompt switching to second- 211
line therapy when indicated would increase this proportion, but the 90% level would be met only by 212
combining multiple interventions. As the number of patients receiving and failing second-line ART 213
increases, the availability of treatment options beyond second-line could help to ensure continued 214
viral suppression(24). The incorporation of more effective and better tolerated ARVs in first- or 215
second-line regimens could also offer opportunities to further improve virological response, 216
although it will likely take several years before newer drugs are implemented at scale in SSA and 217
whether these drugs will show superior effectiveness in routine programme settings remains to be 218
seen(25).
219
Limitations 220
Our study has several limitations. Our model is not a transmission model, and it does not predict the 221
development of the HIV epidemic. Coverage estimates for 2020 were based on the assumption that 222
the number of PLHIV will stay approximately stable in the future. According to the UNAIDS, the 223
number of PLHIV in sub-Saharan Africa has increased slightly over the last few years, from 23.7 224
(20.9-26.7) million in 2010 to 25.5 (22.7-28.7) million in 2016. It is unlikely that this trend would 225
substantially change in the near future, and the number of PLHIV we calculated for 2020 (26.5 226
million) is well in line with these estimates. This is probably a reasonable assumption, but the 227
uncertainty around the UNAIDS estimates should be acknowledged: if the current coverage is 228
overestimated, the necessary increase in the annual number of patients is also higher, and vice 229
versa. In some countries, the UNAIDS upper limit of number of PLHIV in 2016 was up to 50% higher, 230
and the lower limit 35% lower, than the point estimate. When other sources of uncertainty over the 231
next years are taken into account, the margin in 2020 could be even wider. As an example, in a 232
country where a three-fold increase in patients starting ART was needed to achieve the target using 233
the point estimate for number of PLHIV, a 50% larger infected population in 2020 would lead to a 234
five-fold increase needed instead, whereas with a 35% smaller infected population a 1.5-fold 235
increase would be sufficient. Therefore, our results should be seen as an estimate of the magnitude, 236
rather than accurate projections of the need of resources.
237
As the model does not take into account transmission, it also cannot consider the potential influence 238
of achieving the 90-90-90 targets on incidence, mortality, and prevalence. Assessing the 239
epidemiological impact of the 90-90-90 target was not within the scope of this study, but a sudden 240
increase in ART and suppression coverage in the next years could also influence the size of the HIV 241
infected population. It is however unlikely that the number of PLHIV would considerably change 242
within the short period from 2016 to 2020, and the expected increase through prolonged life 243
expectancy would be offset by the reduction in new infections.
244
Our model used the same parameter values for all countries with the exception of the ART initiation 245
rate. We cannot therefore catch all differences between countries. Virological treatment failure was 246
modelled using fixed parameters, based on analyses of mainly South African data. The model is 247
therefore likely to be well applicable to many settings in southern Africa, but not to settings facing 248
serious problems in keeping viral suppression. Third, we estimated viral suppression only based on 249
events that have long-term influence on viral load. We did not model viral load trajectories explicitly, 250
nor did we take into account some potential causes of viraemia such as temporary poor adherence.
251
Conclusions 252
Our study has shown that the ambitious 90-90-90 target, ultimately meaning that 73% of all PLHIV 253
would be virally suppressed by 2020, is realistically reachable in sub-Saharan Africa overall, but not 254
in all individual countries. High-prevalence countries in East and Southern Africa have already 255
reached a high level of ART coverage. The key challenges in these countries will therefore be to 256
maximise viral suppression and prevent patients dropping out of care by improving laboratory 257
monitoring and counselling during treatment, and by establishing access to further treatment 258
options. On the other hand, countries with lower HIV prevalence will need to improve the pace of 259
scale-up of ART that in practice will also require intensive testing campaigns. The gaps in the cascade 260
from HIV infection through testing, diagnosis, linkage to care, ART initiation, and adequate retention 261
and adherence, to sustained and virally suppressed ART, differ between the countries. It is critical 262
that each country review in detail the status of their HIV epidemic and response and use this 263
information to prioritise the interventions targeting the different steps on the cascade.
264 265
References 266
1. 90-90-90. An ambitious treatment target to help end the AIDS epidemic. [Internet]. Joint 267 United Nations Programme on HIV/AIDS (UNAIDS); Available from:
268 http://www.unaids.org/sites/default/files/media_asset/90-90-90_en_0.pdf 269
2. Joint United Nations Programme on HIV/AIDS (UNAIDS). AIDSinfo [Internet]. UNAIDS; 2018.
270 Available from: aidsinfo.unaids.org 271
3. Progress Report 2016. Prevent HIV, Test and Treat All. WHO Support, for Country Impact.
272 [Internet]. Geneva, Switzerland: World Health Organization; 2016. Available from:
273 http://apps.who.int/iris/bitstream/10665/251713/1/WHO-HIV-2016.24-eng.pdf?ua=1 274
4. Lahuerta M, Ue F, Hoffman S, Elul B, Kulkarni SG, Wu Y, et al. The problem of late ART initiation 275 in Sub-Saharan Africa: a transient aspect of scale-up or a long-term phenomenon? J Health 276 Care Poor Underserved. 2013 Feb;24(1):359–83.
277
5. Mori AT, Owenya J. Stock-outs of antiretroviral drugs and coping strategies used to prevent 278 changes in treatment regimens in Kinondoni District, Tanzania: a cross-sectional study. J Pharm 279 Policy Pract. 2014;7:3.
280
6. Mugglin C, Estill J, Wandeler G, Bender N, Egger M, Gsponer T, et al. Loss to programme 281 between HIV diagnosis and initiation of antiretroviral therapy in sub-Saharan Africa: systematic 282 review and meta-analysis. Trop Med Int Health TM IH. 2012 Dec;17(12):1509–20.
283
7. Tenthani L, Haas AD, Tweya H, Jahn A, van Oosterhout JJ, Chimbwandira F, et al. Retention in 284 care under universal antiretroviral therapy for HIV-infected pregnant and breastfeeding 285 women (’Option B+’) in Malawi. AIDS Lond Engl. 2014 Feb 20;28(4):589–98.
286
8. Prevention Gap Report [Internet]. Geneva, Switzerland: Joint United Nations Programme on 287 HIV/AIDS (UNAIDS); 2016. Available from:
288 http://www.unaids.org/sites/default/files/media_asset/2016-prevention-gap-report_en.pdf 289
9. Estill J, Ford N, Salazar-Vizcaya L, Haas AD, Blaser N, Habiyambere V, et al. The need for 290 second-line antiretroviral therapy in adults in sub-Saharan Africa up to 2030: a mathematical 291 modelling study. Lancet HIV. 2016 Mar;3(3):e132-139.
292
10. Estill J, Salazar-Vizcaya L, Blaser N, Egger M, Keiser O. The Cost-Effectiveness of Monitoring 293 294 Strategies for Antiretroviral Therapy of HIV Infected Patients in Resource-Limited Settings:
Software Tool. PloS One. 2015;10(3):e0119299.
295
11. Estill J, Tweya H, Egger M, Wandeler G, Feldacker C, Johnson LF, et al. Tracing of patients lost to 296 297 follow-up and HIV transmission: mathematical modeling study based on 2 large ART programs
in Malawi. J Acquir Immune Defic Syndr 1999. 2014 Apr 15;65(5):e179-186.
298
12. Estill J, Egger M, Blaser N, Vizcaya LS, Garone D, Wood R, et al. Cost-effectiveness of point-of- 299 care viral load monitoring of antiretroviral therapy in resource-limited settings: mathematical 300 modelling study. AIDS Lond Engl. 2013 Jun 1;27(9):1483–92.
301
13. The DHS Program. Demographic and Health Surveys. ICF International;
302
14. The PHIA Project [Internet]. New York, NY, United States: ICAP at Columbia University;
303 Available from: http://phia.icap.columbia.edu 304
15. Egger M, Ekouevi DK, Williams C, Lyamuya RE, Mukumbi H, Braitstein P, et al. Cohort Profile:
305 the international epidemiological databases to evaluate AIDS (IeDEA) in sub-Saharan Africa. Int 306 J Epidemiol. 2012 Oct;41(5):1256–64.
307
16. Salazar Vizcaya L, Blaser N, Gsponer T. gems: Generalized Multistate Simulation Model. Version 308 1.1.1. [Internet]. Available from: https://cran.r-project.org/web/packages/gems/index.html 309
17. Blaser N, Vizcaya LS, Estill J, Zahnd C, Kalesan B, Egger M, et al. gems: An R Package for 310 Simulating from Disease Progression Models. J Stat Softw. 2015 Mar;64(10):1–22.
311
18. Estill J, Egger M, Johnson LF, Gsponer T, Wandeler G, Davies M-A, et al. Monitoring of 312 antiretroviral therapy and mortality in HIV programmes in Malawi, South Africa and Zambia:
313 mathematical modelling study. PloS One. 2013;8(2):e57611.
314
19. Levi J, Raymond A, Pozniak A, Vernazza P, Kohler P, Hill A. Can the UNAIDS 90-90-90 target be 315 achieved? A systematic analysis of national HIV treatment cascades. BMJ Glob Health.
316 2016;1(2):e000010.
317
20. Granich R, Gupta S, Hall I, Aberle-Grasse J, Hader S, Mermin J. Status and methodology of 318 publicly available national HIV care continua and 90-90-90 targets: A systematic review. PLoS 319 Med. 2017 Apr;14(4):e1002253.
320
21. Hill A, Pozniak A, Dauncey T, Levi J, Heath K, Essajee S, et al. Countries with lower HIV 321 prevalence have lower ARV coverage: UNAIDS 2015 database. In Boston, MA, United States;
322 2016.
323
22. Papworth E, Ceesay N, An L, Thiam-Niangoin M, Ky-Zerbo O, Holland C, et al. Epidemiology of 324 HIV among female sex workers, their clients, men who have sex with men and people who 325 inject drugs in West and Central Africa. J Int AIDS Soc. 2013 Dec 2;16 Suppl 3:18751.
326
23. Bernaud C, Khatchatourian L, Rodallec A, Hall N, Perre P, Morrier M, et al. Optimizing the 327 virological success of tenofovir DF/FTC/rilpivirine in HIV-infected naive and virologically 328 suppressed patients through strict clinical and virological selection. Infect Dis Lond Engl. 2016 329 Oct;48(10):754–9.
330
24. Meintjes G, Dunn L, Coetsee M, Hislop M, Leisegang R, Regensberg L, et al. Third-line
331 antiretroviral therapy in Africa: effectiveness in a Southern African retrospective cohort study.
332 AIDS Res Ther. 2015;12:39.
333
25. Vitoria M, Hill AM, Ford NP, Doherty M, Khoo SH, Pozniak AL. Choice of antiretroviral drugs for 334 continued treatment scale-up in a public health approach: what more do we need to know? J 335 Int AIDS Soc. 2016;19(1):20504.
336 337 338 339 340 341
Table 1. Definition of modelled scenarios for 2017-2020.
342
Viral load monitoring Targeted
• Annual testing of CD4 cell counts for all patients
• Referral to viral load test if CD4 cell count fulfils the WHO criteria for immunological failure
• Virological treatment failure confirmed with two tests fulfilling the WHO criteria
• In Botswana, Malawi and South Africa, routine viral load monitoring is applied also in this global scenario
Routine
• Annual testing of viral load for all patients
• Virological treatment failure confirmed with two tests fulfilling the WHO criteria
Retention Realistic
• Patients may stop/interrupt ART once during the course of treatment
• Hazard of ART interruption as in previous modelling analyses
• Patients who interrupt ART may come back to care (hazard of return as in previous analyses)
Perfect
• Patients who once start ART will remain on treatment until death
Switching Realistic
• Patients who fulfill the criteria for confirmed treatment failure will
become eligible to switch to second-line
• ART Eligible patients switch to second-line with constant rate (as in previous analyses)
Perfect
• Patients who fulfill the criteria for confirmed treatment failure switch immediately to second-line ART
343
Table 2. Number of patients who need to start ART annually between 2018 and 2020 to reach 344 universal 81% ART coverage among PLHIV in sub-Saharan Africa.
345
Scenario Number of patients
needed to start ART annually
Relative change to 2017 estimate Retention and switching Viral load
monitoring Treatment interruptions/drop-out*
Switching after random delay* Targeted only** 2,680,900 -6.6%
Routine 2,668,200 -7.1%
No treatment interruptions
Immediate switch after confirmed failure Targeted only** 2,211,800 -23.0%
Routine 2,202,800 -23.3%
Relative change is compared with the modelled estimate of 2,871,500 people starting ART in 2017.
346
*According to observed estimates (see previous modelling studies for parameterisation) 347
**Routine viral load monitoring was assumed to be continued in Botswana, Malawi and South 348
Africa.
349 350
Figure 1. Magnitude of increase in number of patients starting ART by country, needed to achieve 351 81% coverage by 2020. The increase refers to the number of patients that should start each year 352
between 2018 and 2020, compared with the modelled number of patients who started in 2017.
353
354
Figure 2. Percentage of patients receiving ART who are virally suppressed in sub-Saharan Africa 355 between 2016 and 2020 in the main analysis (panel A) and the alternative analysis with higher 356
failure rate (panel B). Light blue curves represent the scenario with optimal retention and 357
immediate switching; black curves the scenario where retention and switching remain unchanged.
358
Solid curves represent the scenario where routine viral load monitoring is implemented in all 359 countries; dashed curves the scenario where only targeted viral load monitoring is implemented.
360
361
362
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%
2016 2017 2018 2019 2020
Virally suppressed
Year
A
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%
2016 2017 2018 2019 2020
Virally suppressed
Year
Current retention, tVL Current retention, rVL Optimal retention, tVL Optimal retention, rVL