1. Hormonal contraceptive use and HIV progression: A systematic review
Sharon Phillips, MD MPH
Department of Reproductive Health and Research
World Health Organization
Kate Curtis, PhD
Division of Reproductive Health, CDC
Chelsea Polis, PhD
Office of Population and Reproductive Health
United States Agency for International Development
Good afternoon, my name is Sharon Phillips and I am pleased to be here today to discuss the results of a systematic review conducted with Drs Kate Curtis and Chelsea Polis on hormonal contraceptive use and HIV progression. This review was initially prepared for a WHO consultation on hormonal contraception in January of this year.

2. Need for comprehensive reproductive health services among women living with HIV
Women living with HIV who desire children should have support to safely conceive and deliver
Substantial unmet need for contraception and unintended pregnancy among women living with HIV
All women who wish to prevent pregnancy deserve access to voluntary family planning services
Many HIV-positive women are sexually active.  Those who wish to become pregnant should receive appropriate care to safely conceive and deliver a healthy infant.  However, many women wish to prevent pregnancy, and several studies have documented high unmet need for family planning and high rates of unintended pregnancy among women living with HIV.  Greater attention is needed to help all women, including those who are living with HIV, find safe, effective, and acceptable ways to prevent unintended pregnancy.  This is true from a human rights perspective, as well as from a public health perspective. 

3. Key Questions
Are women living with HIV who use hormonal contraception at increased risk of:
Death or progression to AIDS
Measured by CD4 <200, initiation of ART, or clinical AIDS
Change in CD4 or viral load
The key questions we aimed to address with this review are the following: Are women living with HIV who use hormonal contraception at increased risk of :1 - Death or progression to AIDS, as measured either by progression to clinical AIDS, CD4 under 200, or ART initiation, or 2- Disease progression as measured by change in CD4 or viral load

4. Methods
First I'll discuss the methods used for the systematic review.

5. Methods: Study selection
Primary reports of studies examining hormonal contraceptive use among women living with HIV
PUBMED and EMBASE searched for published articles in any language through December 15, 2011
634 unique references identified, 16 full-text articles assessed, 12 reports included
Excluded: studies with no comparison group; case control studies
Study information independently abstracted by 2 authors (SP & KC)
We designed a search strategy to identify primary reports of studies examining use of hormonal contraceptive among women living with HIV. We searched PUBMED and EMBASE for articles published or in press in any language through December 15, The search identified 634 unique references, of which 16 were assessed in full text and 12 were included in the review. We excluded studies that lacked a comparison group or case control studies. Study data was independently abstracted by 2 authors.

6. Methods: Quality criteria
Methodology used to minimize confounding
Accurate measurement and analysis of exposure
Composition of comparison group
Loss to follow-up
Length of follow-up
Additional considerations for RCTs
Adequate randomization
Allocation concealment
Distribution of potential confounders between groups
Maintenance of comparability of groups
We assessed quality of the studies based on the methodology used to minimize confounding, the accuracy of measurement and analysis of exposure, the composition of the comparison group, loss to follow-up and length of follow-up. One RCT was included, and additional quality considerations for this RCT included adequacy of randomization, allocation concealment, distribution of potential confounders between groups, and the maintenance of comparability of the groups.

7. Results
Next I will discuss the results of the review.

8. 12 reports (of 11 studies) met inclusion criteria
Results
12 reports (of 11 studies) met inclusion criteria
1 RCT (2 reports)
10 observational
Outcomes considered
Mortality or progression to AIDS
Change in CD4 or viral load
12 reports met inclusion criteria. These 12 reports are of 11 studies; the randomized controlled trial was reported in both 2007 and reanalyzed in of the studies were observational. As I stated before, we considered the following two outcomes: Mortality or progression to AIDS, or a change in CD4 or viral load

9. Outcome 1: Mortality or progression to AIDS
9 reports of 8 studies
1 RCT (2 reports), 7 observational studies
In the first outcome considered, we identified 9 reports of 8 studies. 1 was an RCT, which was reported in 2007 and reported as a reanalysis in 2009, and 7 were observational.

10. 1. Mortality or progression to AIDS
RCT: Stringer et al., 2007/2009 reanalysis
Designed to assess safety of IUD in women living with HIV
599 postpartum women living with HIV, Zambia
Randomized to either copper IUD or hormonal contraception (choice of OCs or DMPA)
2 year follow-up, 6 month visits
High loss to follow-up rates
31% of hormonal group, 23% of IUD group
High method discontinuation/switching rates
49% of IUD users discontinued, 76% of these switched to HC
13% of hormonal users discontinued, 16% switched to IUD
Within hormonal group, 34% switched between OC and DMPA
I'll talk briefly about the RCT. It was published initially in 2007 by Stringer et al, and a reanalysis was published in The purpose of the RCT was to assess the safety of the copper IUD in women living with HIV women living with HIV in Zambia were recruited post-partum and were randomized to either receive a copper intrauterine device (or IUD) or hormonal contraception. Those in the hormonal contraception group had a choice between depo medroxyprogesterone acetate (or DMPA) or oral contraceptives (OCs). Women were followed for 2 years every 6 months. There were high rates of loss to follow-up, with 31% of the hormonal group and 23% of the IUD group lost over the 2 years. There were also high and differential rates of method discontinuation and switching, with nearly half of IUD users discontinuing the method, 13% of hormonal users discontinuing, and 34% of hormonal users switching hormonal method.

11. 1. Mortality (all cause) or progression to AIDS
Stringer 2007/2009 RCT (continued): HR (95% CI)
ITT
Actual use
Mortality
OC vs IUD
1.06 ( )
1.24 ( )
DMPA vs IUD
1.39 ( )
1.83 ( )
CD4<200 or initiate ART
1.54 ( )
1.67 ( )
1.81 ( )
1.62 ( )
Composite outcome (mortality, CD4 <200, or initiate ART)
1.52 ( )
1.81 ( )
This slide shows the results from the Stringer RCT. The intent-to-treat analysis is an analysis based on the assigned contraceptive group; that is, everyone who was assigned to an IUD was analyzed as if she continued using the IUD throughout the study, regardless of whether or not she discontinued the method. The actual use analysis, which analyzes participants based on which method they actually used, is important given the frequency of method discontinuation and switching. Significant results are in red. As you can see, in the intent-to-treat analysis, there were significant results for the outcome of a CD4 under 200 or ART initiation for DMPA users, but not for OC users, and significant results for a composite outcome of either mortality, CD4 count of under 200, or ART initiation for both DMPA users and OC users. In actual use analysis, results were significant for both the outcome of CD4 under 200 or ART initiation for both DMPA and OC users, and for the composite outcome in both DMPA and OC users.

12. 1. Mortality or progression to AIDS (Cohort)
Author, year
Design
Population
Progression
Mortality
(all-cause*)
AIDS/ mortality
Quality
MRC Collaborative
1999
Prospective up to 4 yrs
505 HIV/GU clinic patients
Britain/Ireland
AIDS
OCs ↔
Poor
Kilmarx 2000
Prospective, median 81 mos
194 sex workers, Thailand
CD4< 200
DMPA ↔
Allen
2007
Prospective, 6 yrs
460 women, Rwanda
Fair
Stringer multi-country 2009
Prospective, median 1 yr
7846 women, 12 African countries
ART eligible
OCs ↔
DMPA/ETG↔
OCs 0 (0-)
ART, death
Polis
2010
Retrospective mean 4 years
625 newly seroconverted Uganda
AIDS, death
Good
Morrison 2011
Prospective median 58 months
306 newly seroconvertedUganda & Zimbabwe
AIDS, death, ART initiation
Heikinheimo 2011
Retrospective 5 years
40 women
Finland
ART initiation
LNG-IUD ↔
This is a summary table of the evidence for the impact of hormonal contraception on mortality or progression to AIDS. There are 7 studies, 5 of which are prospective, and 2 retrospective. Most studies had women with exception of the Stringer multi-country, with over 7800 women. 4 looked at HIV progression as measured by CD4 count under 200, ART eligibility, ART initiation, or clinical AIDS; 5 looked at mortality, which was all-cause mortality with the exception of the Allen study, and 3 looked at a composite outcome. Quality ranged from poor to good. The horizontal arrows are indicative of no significant association; as you can see, no significant associations are reported here. All focused on OCs or DMPA with exception of Heikenheimo.
*Except Allen 2007 (HIV-related mortality only)

13. Studies assessing injectables and progression to AIDS OR mortality (composite outcome) (adjusted hazard ratio)
Stringer RCT (2009)*
(DMPA vs IUD)
Stringer Multi-Country (2009)
(Inj/imp† vs no HC)
Morrison (2011)
(DMPA vs no HC)
Polis (2010)
The following slides are graphical representations of the hazard ratios in those studies that assessed the composite outcome of mortality, progression to AIDS, or initiation of ART, with diamonds representing point estimates of and horizontal lines representing 95% confidence intervals. Confidence intervals crossing 1, represented by the vertical line, indicate no significant association. The RCT published by Stringer in 2009 found a significant increase in the risk of the composite outcome with use of injectables, while 3 observational studies failed to find a significant association, with point estimates equal to or less than 1.
Injectables decrease risk of progression
Injectables increase risk of progression
*Actual use analysis
† DMPA, NET-EN, implants

14. Studies assessing OCs and progression to AIDS OR mortality (composite outcome) (Adj hazard ratio)
Stringer RCT (2009)*
(OCs vs IUD)
Morrison (2011)
(OCs vs no HC)
Stringer Multi-Country (2009)
Polis (2010)
Among studies that looked at oral contraceptive pills and the composite outcome of progression to AIDS or mortality, the Stringer RCT found a significant association, while 3 observational studies found no association, with point estimates falling on both sides of 1. 1
OCs increase risk of progression
OCs decrease risk of progression
*Actual use analysis

15. Results: Outcomes considered
Mortality or progression to AIDS
Change in CD4 or viral load
We just talked about the first outcome, mortality or progression to AIDS. Now I'll briefly discuss the second outcome, change in CD4 count or viral load. In this case we looked at studies that did not report progression to a specified CD4 count of under 200, but rather studies that compared CD4 between groups without a specific threshold, or that looked at change in viral load over time.

16. 2. Change in viral load, CD4 (5 observational studies)
Author, year,
study design
Population
Changes in CD4
Changes in HIV RNA
Quality
Kilmarx, 2000
Prospective cohort, median 81 months
194 sex workers
Thailand
Rapid decline
OCs ↔
DMPA ↔
Poor
Cejtin, 2003
Prospective cohort, 1-2 years
1721 women US
Hormonal
Contracept ↑
(no adverse effect)
Hormonal Contracept ↔
Fair
Lavreys, 2004
Prospective cohort, median 34 months
161 newly seroconverted sex workers Kenya
Hormonal Contracept ↔
Richardson, 2007
Prospective cohort, 24 mos
283 postpartum women
Kenya
Heikinheimo 2011
Retrospective cohort, 5 years
40 women
Finland
LNG-IUD ↔
Most of the studies had between 150 and 300 people, with one study having 1700 participants and one study that looked only at the LNG-IUD having 40 participants. All were observational studies. 4 were prospective studies, 1 was retrospective. Quality ranged from poor to fair. No significant difference was seen in the outcome assessed in any of the studies, with the exception of an increase in CD4 noted among hormonal contraceptive users as opposed to non-users of uncertain clinical significance.

17. Discussion: Outcome 1 Mortality or progression to AIDS
7 observational studies find no association between HC and HIV disease progression
1 RCT found increased rates of
time to CD4 count < 200 and
time to CD4 count < 200 and mortality
among HC users compared with IUD users (both OC and DMPA users)
In summary, going back to the two outcomes we looked at, in terms of mortality or progression to AIDS, overall 7 observational studies found no association between hormonal contraception and HIV disease progression. 1 RCT did find increased rates of progression to CD4 count less than 200 and the composite outcome of progression to CD4 <200 and mortality. This increase was seen among both OC and DMPA users when compared with IUD users in actual use analysis, and among DMPA users in intent-to-treat analysis.

18. Discussion: Outcome 1 Mortality or progression to AIDS
Strengths
Many observational studies with similar findings, 2 with very strong methodology
One very large study (n=7846)
Limitations
Some small sample sizes
Follow-up time
RCT:
loss to follow-up
method switching
comparison with IUD
Strengths in the studies reviewed include multiple observational studies with similar findings, 2 with very strong methodology that took into account many potential confounders, assessed contraceptive use over time, and assessed women at time of HIV infection. One of the studies had a very large sample size of 7800.
Limitations in the studies assessing this outcome include small sample sizes in most cases, short follow-up time, and limitations of the RCT include high levels of loss to follow-up, method switching, and comparing to the IUD rather than no contraception, as the use of the copper IUD itself among women living with HIV is not well studied and may have an effect we are unaware of.

19. Discussion: Outcome 2 Change in viral load, CD4
5 observational studies find no adverse association between HC and change in viral load or CD4
Limitations
Small sample sizes
Failure to separate HC methods in some studies
Lack of control for potential confounders
In terms of the second outcome, change in viral load or CD4, overall we have 5 observational studies that found no adverse association between use of hormonal contraception and these markers of HIV progression, as well as some on the safety of the levonorgestrel IUD for women living with HIV. 20

20. Discussion: Study Quality – Observational studies
Quality ranged from poor to good
"Good" studies
Incident HIV cases
Multivariate analysis
Time-varying analysis of use of hormonal contraception
Findings similar to those rated as "fair" and "poor"
"Fair" studies
Prevalent HIV cases
Control for baseline health characteristics in multivariate model
"Poor" studies
No separate analysis of different contraceptive methods
Inclusion of other HC users in comparison group
No multivariate analysis
I want to provide an overview of characteristics of the studies in the different categories. There was a range of quality. Those studies rated as good generally had incident cases of HIV (that is, women were followed from estimated time of seroconversion), employed multivariate analysis, and used time-varying analysis of hormonal contraception (that is, attribution of the participant to contraceptive category was updated throughout the study). The findings of those studies rated as good were similar to the findings of those rated fair or poor. Fair studies tended to use prevalent HIV cases, that is, women already had HIV when they were enrolled rather than being enrolled upon seroconversion, and controlled for baseline health characteristics in a multivariate model. Poor studies had one or more of the following characteristics: No separate analysis of different contraceptive methods, inclusion of users of other HC methods in the comparison group, or no multivariate analysis. 21

21. Discussion: Limitations in body of research
Minimal or no information on newer methods (LNG-IUD, patch, ring, implants)
Limited data for women with clinical AIDS
All studies observational, with the exception of 1 RCT
There remain some important limitations in the body of research. We still have minimal or no information on newer hormonal contraceptive methods, such as the levonorgestrel IUD, the patch, the contraceptive ring, and contraceptive implants. We have limited data regarding the safety of contraceptives for women with AIDS as most studies include primarily relatively healthy women living with HIV rather than women with AIDS. With the exception of 1 randomized controlled trial, all studies are observational and carry the inherent potential for confounding that all such studies have.

22. The preponderance of evidence thus indicates that use of OCs or of DMPA does not affect HIV disease progression among women with HIV
Conclusion

23. Acknowledgements
Members of the WHO Hormonal Contraception & HIV Advisory Group (manuscript review)
Andy Gray, Olav Meirik, & Catherine Hankins
Assistance with project development
Mary Lyn Gaffield, Nathalie Kapp, & Roger Chou
Assistance with EROS software
Agustin Ciapponi & Demián Glujovsky
Assistance with literature search
Nellie Kamau & LaToya Armstrong
Contact:

24. Studies assessing injectables and mortality (Adjusted hazard ratio)
Stringer RCT (2009)*
(DMPA vs IUD)
Kilmarx (2000)
(DMPA vs non-DMPA†)
Polis (2010)
(DMPA vs no HC method)
Stringer Multi-Country (2009)
(Imp/inj vs no HC method†)
Allen (2007)
(unspec inj. vs
never used injectables)
The following slides are graphical representations of the hazard ratios in each study, with 95% confidence intervals. The vertical line at 1 is indicative of no difference. The diamond represents the sample size. As you can see, when looking at injectables and the outcome of mortality, there are no significant findings for a difference, as all confidence intervals cross 1, and all point estimates fall at or below one
Injectables decrease risk of mortality
Injectables increase risk of mortality
*Actual use analysis
†Mostly OCs
†DMPA, NET-EN, implants

25. Studies assessing OCs and mortality (Adjusted hazard ratio)
Stringer RCT (2009)*
(OCs vs IUD)
Kilmarx (2000)
(OCs vs non-OCs†)
MRC (1999)
(OCs vs other/no contraception)
Polis (2010)
(OCs vs no hormonal method)
Allen (2007)
(OCs vs never used OCs)
Similarly, there was no association between oral contraceptives and mortality in any of the studies reviewed.
OCs decrease risk of mortality
OCs increase risk of mortality
*Actual use analysis
†Mostly DMPA

26. Studies assessing injectables and progression to AIDS (adjusted hazard ratio)
Stringer RCT (2009)*
(DMPA vs IUD)
Stringer Multi-Country (2009)
(Inj/imp vs no HC†)
Morrison (2011)
(DMPA vs no HC)
Kilmarx (2000)
(DMPA vs non-DMPA†)
Looking at studies assessing injectables and progression to AIDS, the Stringer RCT found a significant increase in risk of progression to AIDS, while 3 observational studies failed to find a significant difference between groups and point estimates are at or below 1. 1
Injectables decrease risk of progression
Injectables increase risk of progression
*Actual use analysis
†DMPA, NET-EN or implant
†Mostly OCs

27. Studies assessing OCs and progression to AIDS (Adjusted hazard ratio)
Stringer RCT (2009)*
(OCs vs IUD)
Kilmarx (2000)
(OCs vs non-OCs†)
Morrison (2011)
(Low dose OCs vs no HC)
Stringer Multi-Country (2009)
(OCs vs no HC)
MRC (1999)
(OCs vs other or no HC)
Similarly, looking at studies assessing the relationship between the use of oral contraceptive pills and progression to AIDS, only the Stringer RCT found a significant increase; 4 observational studies found no significant difference, with all 95% confidence intervals crossing 1. Point estimates fall on both sides of 1.
OCs decrease risk of progression
OCs increase risk of progression
*Actual use analysis
†Mostly DMPA

28. Study Flow