1. Impact of Expanding Treatment Eligibility on ART Initiation and Retention in Care in Patients with CD4>350 cells/L: Evidence from Zambia
Aaloke Mody, Izukanji Sikazwe, Carolyn Bolton, Kombatende Sikombe, Arianna Zanolini, Paul Somwe, Charles Holmes, Nancy Padian, Elvin Geng
IAS 2017
Paris, France
July 25, 2017
Thank you for this opportunity to present this work on behalf on my coauthors. My name is Aaloke Mody and I am from UCSF and also CIDRZ in Zambia. The title of this talk is the Impact of Expanding Treatment Eligibility on ART Initiation and Retention in Care in Patients with a CD4 greater than 350 in Zambia

2. Background

3. Background
The real world effects of adopting “treat all” are difficult to know
RCTs demonstrate efficacy in well-controlled environments making extrapolation uncertain
Positive effects of expanding eligibility may be both direct and indirect and should be quantified to sharpen our understanding of the benefits
Negative spillover effects—such as increased congestion and crowding out of patients—need to be identified to minimize unintended consequences
Real world program data is the most relevant for these questions, but causal inference is limited with observational data using traditional methods
The real world effects of expanding ART eligibility as we move into the era of “treat all” are difficult to know
RCTs demonstrate efficacy in well-controlled environments making extrapolation to the real world uncertain
click
The positive effects of expanding ART eligibility may be both direct and indirect and these should be quantified to sharpen our understanding of the benefits
Similarly negative spillover effects such as increased congestion and crowding out of patients should be identified to minimize unintended consequences.
Click
Real world program data is the most relevant for answering these questions, but causal inference is often limited with observational data when using traditional methods
WHO HIV Treatment Guidelines 2015; Abdool Karim NEJM 2015; Fox PLOS Med 2017

4. Background
Quasi-experimental designs offer an underused solution by approximating experimental conditions in real world data
Regression discontinuity used in South Africa to assess effects of ART eligibility in those with a CD4 below 350 cells/L
Zambia adopted universal treatment in December 2016, but previously changed its HIV treatment guidelines on April 1, 2014 to expand ART eligibility to include:
Patients with a CD4 between 350 and 500 cells/L
All pregnant and breastfeeding women (and their partners) under Option B+
We leverage this 2014 guideline change in a quasi-experimental study design to shed light onto what might be expected when expanding to “treat all”
Quasi-experimental study designs offer an underused strategy to address these challenges by approximating experimental conditions in real world data
An example of this is regression discontinuity which has been previously used in South Africa to assess the effects of ART eligibility in those with a CD4 below 350
Click
Zambia adopted universal treatment in December 2016, but prior to this, it changed its HIV treatment guidelines on April 1, 2014 to expand ART eligibility to include patients with a CD4 between 350 and 500 as well as pregnant and breastfeeding women and their partners under Option B+.
We leverage this 2014 change in Zambia’s HIV treatment guidelines in a quasi-experimental study design to shed light onto what might be expected when expanding to “treat all”
Bor Curr Opin HIV/AIDS 2015; Kluberg CROI 2016; Bor CROI 2016; Zambia HIV Guidelines 2010, 2014

5. Study Objectives
To estimate the real world effects of expanding treatment eligibility on ART initiation and retention in care
To quantify the effect of ART initiation on retention in care for HIV-infected individuals with high CD4 counts
Our study objectives are:
To estimate the real world effects of expanding treatment eligibility on ART initiation and retention in care.
To quantify the effect of ART initiation on retention in care for patients with high CD4 counts

6. Methods

7. Study Population Measurements Better Info for Health study
ART naïve patients greater than 15 years old
Newly enrolling 7 months before and after the change in Zambia’s treatment guidelines (September 1, 2013 to November 1, 2014)
Excluded patients enrolling within 1 month of the change to account for a transition period
64 clinics supported by the Centre for Infectious Disease Research in Zambia (CIDRZ) in Lusaka, Eastern, Western, and Southern Provinces
Measurements
Better Info for Health study
Sampling-based tracing to ascertain outcomes in LTFU
EMR system used in routine HIV care in Zambia (SmartCare)
Clinical forms manually filled by providers and entered into the electronic database along with lab data by data clerks
Our study population included ART naïve patients greater than 15 years old that newly enrolled in HIV care 7 months before and after this change in Zambia’s treatment guidelines
We excluded patients enrolling within 1 month of the change to account for a transition period as the new guidelines were being rolled out
We looked at patients enrolling at one of 64 clinics supported by the Centre for Infectious Disease Research in Zambia or CIDRZ
click
Our measurements were obtained from the Better Info for Health study, which is a Gates-funded study using sampling-based tracing to ascertain outcomes in patients LTFU.
And we also used data from the electronic medical record system used in routine HIV care in Zambia called SmartCare

8. Regression Discontinuity Design (RDD)
RDD to estimate the real world effects of implementing guidelines
Compare patients enrolling just before and after the change (local)
Assume similar on measured and unmeasured characteristics, thus, exposure is as random
Outcomes
ART Initiation by 3 months
Retention in Care at 6 months
Visit between 3 and 9 month post-enrollment
In care and on ART at 6 months
Composite Outcome
We used a regression discontinuity design to estimate the real world effects of implementing new guidelines
For this study design, we compare patients enrolling just before and just after the change in guidelines.
As the rollout date is arbitrary, we can assume that patients enrolling before and after will be similar on both measured and unmeasured characteristics, making assignment to exposure as random
Click
Our three main outcomes were:
1) ART initiation within 3 months of enrollment
2) Retention in care at 6 months which we defined as having made at least one visit between 3 and 9 months post-enrollment.
3) A composite outcome of having started ART at 3 months and still in care at 6 months.
We used a local linear regression adjusting for CD4, age, and gender and then also weighted patients proportional to how close they enrolled to the cutoff to strengthen the validity of the underlying RD assumptions
We also conducted sensitivity analyses varying the bandwidth around the cutoff and length of the transition period to ensure the robustness of our results which we do not show here.
Model
Local linear regression adjusted for CD4, age, and gender
Patients weighted proportional to how close they enrolled to the cutoff
Sensitivity analyses varying bandwidths around cutoff and length of transition period (results not shown)
Bor Epidemiology 2014

9. RD Instrumental Variable (IV) Analysis
We used implementation of new guidelines as an IV to estimate the effect of ART initiation on retention in care
Provides unbiased estimates even when there is unmeasured confounding between the treatment (ART initiation) and outcome (retention) under certain assumptions
IV is associated with the treatment
There are no common causes of the IV and the outcome
The IV only affects the outcome through the treatment
We also leveraged this RD design and used implementation of new guidelines as an instrumental variable to estimate the real world effect of ART initiation on retention in care.
The benefits of an IV analysis is that it can provide unbiased estimates even when there is unmeasured confounding between a treatment—in this case ART initiation—and the outcome—retention in care—under certain assumptions.
The instrument…in this case implementation of guidelines…is associated with the treatment
There are no common causes of the instrument and the outcome
The instrument affects the outcome only through the treatment of interest
In this situation, we estimate the effect of ART initiation on retention in care for those patients who initiated ART in response to the change in guidelines, also called the local average treatment effect
Swanson Epidemiology 2013

10. Patient Subgroups*
Always Eligible: Eligible by 2010 Guidelines at enrollment
CD4 less than or equal to 350 cells/L
WHO Stage 3 or 4
Active Tuberculosis
Unable to identify patients with serodiscordant partners or HBV coinfection and severe liver disease
Newly Eligible: Eligible by 2014 Guidelines at enrollment
CD4 between 350 and 500 cells/L
Pregnant or Breastfeeding Women
Unable to identify partners of pregnant or breastfeeding women
Not Yet Eligible: Not Eligible by 2010 or 2014 Guidelines at enrollment
CD4 greater than 500 cells/L
*Defined irrespective of the actual date of enrollment
For our analyses, we defined 3 patient subgroups which were defined irrespective of their actual date of enrollment
Our first subgroup were “Always Eligible” patients who would have been eligible by the 2010 guidelines at enrollment.
These were patients who had a CD4 below 350, WHO Stage 3 or 4, or had active TB
Data limitations prevented us from ascertaining patients with serodiscordant partners and patients with HBV coinfection and severe liver disease who also would have been in this category
Our second subgroup were newly eligible patients who would have been eligible by the 2014 guidelines
These were patients with a CD4 between 350 and 500 or pregnant and breastfeeding women
We were limited in our ability to identify partners of pregnant and breastfeeding women
Our last subgroup were not yet eligible patients who had a CD4 above 500 and would not have been eligible by either the 2010 or 2014 guidelines.

11. Results

12. Study Population Patient Characteristics at Enrollment, n=23,644
Gender, n (%)
Male
9,431 (39.9)
Female
9,410 (39.8)
Pregnant or Breastfeeding Female
4,803 (20.3)
Median Age, years (IQR)
34 (28, 41)
Median CD4 count, cells/L (IQR)
267 (134, 430)
WHO Stage, n (%) 1
11,092 (53.7) 2
3,740 (18.1) 3
5,345 (25.9) 4
469 (2.3)
Recent TB, n (%)
1,312 (5.5)
Patient Subgroup, n (%)
Always Eligible
16,567 (70.1)
Newly Eligible
4,455 (18.8)
Not Yet Eligible
2,622 (11.1)
Province, n (%)
Lusaka
4,277 (18.1)
Eastern
13,641 (57.7)
Southern
2,838 (12.0)
Western
2,888 (12.2)
Study Population
32,675 patients newly enrolled during our study period and 23,644 were included in our final analysis
Click
Newly enrolling patients had a median CD4 of 267 and about 70% would have been eligible by the 2010 guidelines at the time of enrollment.

13. Patient Characteristics at Enrollment, n=23,644
Always Eligible
n=16,567
Newly Eligible
n=4,455
Not Yet Eligible
n=2,622
Pre-Guidelines
n=8,711
Post-Guidelines n=7,856
n=2,141
Post-Guidelines n=2,314
n=1,276
Post-Guidelines n=1,346
Gender, n (%)
Male
3,876 (44.5)
3,560 (45.3)
520 (24.3)
567 (24.5)
442 (34.6)
466 (34.6)
Female
3,255 (37.4)
2,964 (37.7)
731 (34.1)
746 (32.2)
834 (65.4)
880 (65.4)
Pregnant or Breastfeeding Female
1,580 (18.1)
1,332 (17.0)
890 (41.6)
1001 (43.3)
Median Age, years (IQR) 35
(29, 42)
(30, 42) 31
(25, 37) 32
(27, 39) 33
(27, 41)
Median CD4 count, cells/L (IQR)
185
(92, 279)
181
(93, 272)
436
(389, 488)
439
(394, 492)
622
(556, 758)
635
(559, 769)
WHO Stage, n (%) 1
3416 (43.4)
3002 (43.3)
1409 (80.5)
1495 (79.2)
844 (78.3)
926 (82.3) 2
1321 (16.8)
1253 (18.1)
341 (19.5)
392 (20.8)
234 (21.7)
199 (17.7) 3
2876 (36.5)
2469 (35.6)
0 (0) 4
256 (3.3)
213 (3.1)
Recent TB, n (%)
675 (7.7)
637 (8.1)
This is a very dense slide but the main take away is that, within each patient subgroup, patients enrolling prior to the change in guidelines were very similar to those who enrolled afterwards, helping to verify the underlying assumptions of the RD analysis.

14. Here we also see that the number of new patient enrollments and the patient distribution remained essentially unchanged before and after the guidelines were rolled out, also helping to verify key assumptions of our analysis.

15. Always Eligible: ART Initiation by 3 months
These first set of slides shows results from our RD analysis in the Always Eligible group.
The x-axis shows the time since the guidelines were rolled out with the dotted line representing the date of implementation.
The y-axis shows the percent of patients with the outcome…in this case ART initiation at 3 months
As you can see, implementing new guidelines did not effect ART initiation in “always eligible” patients
Always Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
ART Initiation, % (95% CI)
75.8
(74.1, 77.4)
76.7
(75.1, 78.4)
+1.0
(-1.9, 3.9)
0.507

16. Always Eligible: Retention in Care at 6 months
It also did not effect retention in care
Always Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
Retention in Care, % (95% CI)
72.2
(70.5, 74.0)
72.7
(70.9, 74.5)
+0.5
(-2.6, 3.6)
0.753

17. Always Eligible: In Care and on ART at 6 months
Nor did it effect the percentage of patients who were in care and on ART at 6 months.
Always Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
In Care on ART, % (95% CI)
62.8
(60.9, 64.7)
63.9
(62.0, 65.9)
+1.2
(-2.2, 4.5)
0.493

18. Newly Eligible: ART Initiation by 3 months
Here are the results for the newly eligible patients. In this case, implementing new guidelines increased ART initiation at 3 months by 38%
Newly Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
ART Initiation, % (95% CI)
33.3
(30.0, 36.6)
71.2
(67.9, 74.4)
+37.9
(32.1, 43.7)
<0.001

19. Newly Eligible: Retention in Care at 6 months
It is also significantly increased retention in care at 6 months by 7%, p-value of
Newly Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
Retention in Care, % (95% CI)
65.6
(62.2, 69.0)
72.6
(69.5, 75.8)
+7.0
(1.2, 12.0)
0.018

20. Newly Eligible: In Care and on ART at 6 months
New guidelines also increased the percentage of newly eligible patients in care and on ART by 29%.
Always Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
In Care on ART, % (95% CI)
29.6
(26.4, 32.9)
58.5
(55.1, 62.0)
+28.9
(23.0, 34.9)
<0.001

21. Not Yet Eligible: ART Initiation by 3 months
For those with a CD4 above 500 and not yet eligible, implementation of new guidelines interestingly also increased ART initiation by about 14%, p-value <0.001
Not Yet Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
ART Initiation, % (95% CI)
13.9
(10.6, 17.1)
27.5
(23.6, 31.4)
+13.6
(7.4, 19.9)
<0.001

22. Not Yet Eligible: Retention in Care at 6 months
It did not significantly effect retention in care
Not Yet Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
Retention in Care, % (95% CI)
49.1
(44.5, 53.8)
53.5
(48.8, 58.1)
+4.3
(-3.9, 12.6)
0.302

23. Not Yet Eligible: In Care and on ART at 6 months
But it did increase the percent of patients in care and on ART by 9.4%, p=0.002
Not Yet Eligible
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
In Care on ART, % (95% CI)
12.8
(9.7, 15.8)
22.2
(18.6, 25.8)
+9.4
(3.5, 15.3)
0.002

24. All Patients: ART Initiation by 3 months
The overall effects of implementing new guidelines was a 9.7% increase in ART initiation
All Patients
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
ART Initiation, % (95% CI)
60.2
(58.8, 61.7)
69.9
(68.4, 71.4)
+9.7
(7.1, 12.2)
<0.001

25. All Patients: Retention in Care at 6 months
A minimal effect on retention
All Patients
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
Retention in Care, % (95% CI)
68.3
(66.8, 69.8)
70.5
(69.0, 71.9)
+2.2
(-0.4, 4.8)
0.102

26. All Patients: In Care and on ART at 6 months
And a 7.4% increase in the percentage of patients in care and on ART.
All Patients
Pre-Guidelines
Post-Guidelines
Risk Difference
p-value
In Care on ART, % (95% CI)
50.5
(49.0, 52.1)
58.0
(56.4, 59.5)
+7.4
(4.7, 10.1)
<0.001

27. Newly Eligible Patients
Instrumental Variable Analysis Effect of ART Initiation on Retention in Care
Newly Eligible Patients
26.9% increase in retention (95% CI 20.1, 33.8%, p<0.001)
Eligible by CD4: +29.3% (95% CI 20.7, 38.0%, p<0.001)
Eligible by Option B+: +18.2% (95% CI 17.5, 28.8%, p<0.001)
Not Yet Eligible Patients
29.1% increase in retention (95% CI 2.8, 55.5%, p=0.03)
Lastly, we have the results from our instrumental variable analysis.
In newly eligible patients, initiating ART increased retention in care by 26.9%.
This effect was more pronounced in patients newly eligible due to CD4 (29.3% increase) compared to those newly eligible due to Option B+ (+18% increase).
Additionally, we estimated that ART initiation also increased retention in care by 29% in not yet eligible patients with a CD4 above 500.

28. Limitations Unable to assess virologic suppression
Retention is an imperfect proxy for virologic outcomes
Limitations of primary data source
Potential for misclassification of subgroup/outcome
Potential bias from ”crossover” into post-guideline period
Likely a conservative bias and sensitivity analyses suggest bias is negligible
There are several limitations to our study
First, we were unable to assess virologic suppression, though recent results from ZAMPHIA suggest just under 90% of patients retained in care are virologically suppressed
Two, there are inherent limitations of our primary data source, and, though we did conduct chart reviews and sensitivity analyses to validate our categorization, there is still potential for bias from misclassification
Lastly, there is potential bias due to patients crossing over from pre-guidelines to post-guidelines, we anticipated this to be a conservative bias and our sensitivity analyses suggested any bias was negligible.

29. Conclusion

30. Conclusions Effects of expanding ART eligibility
Newly Eligible: Increased ART initiation by 37.9%, retention by 7.0%, and the percentage in care on ART at 6 months by 28.9%
Not Yet Eligible: Increased ART initiation by 13.6% and the percentage in care and on ART by 9.4%
Potentially due to informal loosening of eligibility criteria
Always Eligible: No decreases in ART initiation or retention in care
All Patients: 9.7% increase in ART initiation and 7.4% increase in those in care and on ART
Initiating ART increased retention by ~25 to 30% in patients with CD4 greater than 350 cells/L
70% of patients still present at a more advanced stage
25% of eligible patients are not initiated on ART by 3 months
40% are not on ART and in care at 6 months
In conclusion, we found that expanding ART eligibility led to overall positive effects in those with higher CD4 counts without any negative spillover effects in those with lower CD4 counts.
In Newly eligible patients, ART initiation within 3 months increased by 38%, retention in care at 6 months increased by 7%, and the percentage of patients in care and on ART at 6 months increased by 29%.
In patients not yet eligible for ART, the new guidelines also increased ART initiation by 14% and the percentage of patients in care and on ART by 9.4%. We posit that this is likely due to an informal loosening of eligibility criteria—a positive spillover effect.
There was no evidence of negative spillover effects in always eligible patients and, overall implementing new guidelines increased ART initiation by 9.7% and the percentage of patients in care and on ART by 7.4%.
Click
Additionally, we estimated that initiating ART in patients with a CD4 above 350 results in about a 25 to 30% increase in retention in care.
Lastly, about 70% of patients still present at a more advanced stage, 25% of eligible patients are still not initiated on ART by 3 months, and about 40% of eligible patients are not in care and on ART at 6 months

31. Implications
Adopting “treat all” will likely lead to similar improvements in the number of patients on ART and retained in care without crowding out sicker patients
Substantial efforts are still needed to diagnose and link patients to care earlier and keeping them retained in care
Overall effects of expanding ART eligibility alone may be modest
Further evidence still needed on real world effects of “treat all”
Its overall impact on virologic suppression
Better understanding of those that never start ART or quickly drop out
Targeted interventions for those that don’t benefit from eligibility alone
Quasi-experimental designs are a powerful and necessary tool for understanding the effects of real world implementation of evidence-based interventions
Based on these results, we believe we will likely see similar improvements in number of patients on ART and retained in care as we move into the era of ”treat all”, without associated negative spillover effects
click
Nevertheless, depending on the underlying population, the overall effects of expanding ART eligibility alone may be modest, and substantial efforts are still needed to improve earlier diagnosis and linkage to care and keeping those patients retained.
Further evidence is still needed into the real world effects of “treat all” specifically
Its effect on virologic suppression
A better understanding of those patients eligible for ART that still don’t start or drop out shortly after
And then targeted interventions for these patients that don’t benefit from eligibility alone
Click
Lastly, quasi experimental research designs are a powerful and necessary tool to understand the effects of real world implementation of evidence-based interventions

32. Thank you! UCSF Elvin Geng Monika Roy Megha Mehrotra
Nancy Czaicki (in memorium)
UC Berkeley
Nancy Padian
JHU
Charles Holmes
CIDRZ
Izukanji Sikazwe
Carolyn Bolton-Moore
Thea Savory
Mwanza wa Mwanza
Kombatende Sikombe
Arianna Zanolini
Paul Somwe
Jake Pry
Anjali Sharma
Zambian Ministry of Health
Bill and Melinda Gates Foundation