We included clinical trials (randomised and nonrandomised), cohort, case-control, and pharmacokinetic (PK) studies, and excluded abstracts, case reports, case series, cross-sectional studies, letters, editorials, review articles without primary data, and non-published results. We required that all included studies have a comparison group, and therefore also excluded prospective observational studies without a control group. We included studies of any method of HC in combination with any rifamycin, but excluded studies of non-contraceptive formulations of steroid hormones (IV estrogen). Our clinical outcomes included pregnancy, presumed ovulation by serum progesterone with or without ultrasound, TB disease progression or TB-related death, and adverse health effects (breakthrough bleeding, drug side effects or adverse events). We included PK studies of either the contraceptive steroid hormone or the rifamycin drug that reported area under the curve (AUC), maximum serum concentration (Cmax), minimum serum concentration (Cmin) or mean 24-hour drug levels, and excluded studies only reporting steroid hormone urinary excretion.

We included studies of women with or without TB.

We searched MEDLINE, Embase, Clinicaltrials.gov, and Cochrane libraries for articles in any language from database inception to May 2017 using search terms developed with a reference librarian (Supporting Information Appendix S1), and scanned references sections of included articles and relevant review articles to identify additional studies. Studies of non-rifamycin antibiotics are reported separately.¹⁰

One author (K.B.S.) performed the database search and screened titles and abstracts for initial exclusion based on study type, exposures and outcomes. Two authors independently reviewed the full text of all articles not excluded during abstract review to determine which met the criteria for inclusion. Differences were resolved by discussion with a third author. Articles were translated into English as needed.

One author independently extracted relevant study information into evidence tables and a second author independently reviewed tables for accuracy prior to study grading (Supporting Information Appendix S2). Information recorded for each article in standardised abstraction tables included study design, objectives, population, drug exposures (contraceptive and rifamycin drugs including duration of use, dose, timing and measures of adherence), outcomes including PK parameters and criteria for detecting ovulation, primary findings for each outcome, and assessment of confounders. For PK outcomes, we recorded AUC, Cmax, Cmin, and/or mean 24-hour drug levels of contraceptive and/or rifamycin drugs alone and in combination, including the results of statistical comparisons. For clinical outcomes we recorded any reported pregnancies, frequency of ovulation based on serum progesterone with or without ultrasound, adverse health effects including frequency of irregular bleeding, and TB disease outcomes. We did not contact authors to obtain additional non-published information.

We used the United States Preventative Services Task Force (USPSTF) grading scale (good, fair, poor) to evaluate the quality of evidence each study provided for its primary outcome. For cohort and case-control studies, we graded each of the following criteria: definition and assessment of exposures, definition and identification of outcomes, selection of controls, blinding, confounders, sample size, response rate/follow up, and internal validity.¹¹ As no standardised grading scale exists to assess the quality of PK studies, we utilised a rating system previously reported to evaluate studies with PK outcomes, which included assessment of study design, sample size, exposures, outcomes, timing, intersubject variability, population, steady state of perpetrator drug, and validation of assays.¹² We graded studies based on their primary outcome, but also report secondary outcomes. A study received the overall grade of good if every graded component received a score of good, and poor if one or more grading criteria was graded as poor (considered a fatal flaw that would invalidate results). All other studies were graded as fair. The quality of each study was assigned independently by two authors, and any differences were resolved through discussion with a third author.

We used evidence tables to synthesise data and quality evaluations. Findings are reported descriptively for each drug. We could not perform meta-analysis due to heterogeneity of exposures and outcomes, as well as limited data for certain drugs.

One study addressed COC PK with rifaximin, a drug approved by the US FDA for the treatment of travelers’ diarrhoea (Table 2). A single-sequence crossover study of 28 healthy women reported COC PK following a single COC pill, and again following a second single COC pill after 3 days of rifaximin (200 mg every 8 hours).²⁰ This study reported no difference in EE or norgestimate (NGM) AUC or Cmax GMRs before or after co-therapy.

One study addressed COC PK with the experimental drug rifalazil (Table 2). This single-sequence crossover study administered a COC to 14 postmenopausal women for 14 days, and measured EE PK before and after a single oral dose of rifalazil (25 mg) on day 8.²¹ EE AUC and Cmax remained within the pre-specified GMR range of bioequivalence (CI of 80–125%) following rifalazil.

Women with medical conditions that expose them to increased health risks in pregnancy, such as TB, may be particularly motivated to avoid an unintended pregnancy.⁷ The most relevant drug interaction outcomes are pregnancy (reflecting decreased contraceptive effectiveness) or TB disease progression (reflecting decreased anti-TB therapeutic effectiveness). Although our review of rifamycin antimicrobials and HC demonstrated some evidence of drug interactions when COCs are administered with rifampin, no studies directly evaluated the risk of pregnancy. Our review also noted varying degrees of drug interaction among other rifamycins. We found no studies directly evaluating rifamycins with other HC methods and very limited data on the effect of COCs on anti-TB therapy PK or disease progression.

Surrogate markers of contraceptive effectiveness in this review showed mixed effects. In two fair quality studies, up to 50% of subjects taking rifampin with COCs had presumed ovulation diagnosed by serum progesterone (combined with a single ultrasound in one study), whereas no women ovulated on COCs alone.^18,19 Two additional good quality studies found no evidence of ovulation by serum progesterone with the COC/rifampin combination.^13,14 Breakthrough bleeding with COCs (which does not indicate ovulation but may affect pill tolerance or compliance) was more common during rifampin administration than without in two of three studies addressing this outcome (fair to poor quality).^14,23,24

Differences in drug exposure were more consistent. Progestin exposure as measured by AUC was reduced 30–83% in five studies of good to fair quality when COCs were co-administered with rifampin compared with COCs alone.^{13–15,17,18} Statistically significant reductions in progestin half-life and C_max, and increases in drug clearance, were also observed in some of these studies.^13,15,17 EE exposure by AUC was reduced by 42–66% in four of these studies when COCs were administered with rifampin compared with alone,^13,14,16,17 with similar changes noted in other PK parameters. Minimum efficacy thresholds are not defined for EE or progestins, so it is not possible to predict the degree of contraceptive compromise based on PK changes alone.²⁵ However, progestin levels are critical for contraception, so these significant decreases combined with the observed alterations in ovulation suppression are concerning for a possible reduction in contraceptive effectiveness when adherence-dependent methods such as COCs are combined with rifampin.

Studies of other rifamycins demonstrated less consistent effects on ovulation and contraceptive PK. Rifabutin induces and is metabolised by CYP3A4, though its degree of enzyme induction is less than rifampin.⁸ No ovulation was detected by serum progesterone when COCs were combined with rifabutin, and observed PK changes were generally smaller with rifabutin than with rifampin in two studies of good quality.^13,14 Nevertheless, PK interaction was still present and a reduction in contraceptive effect remains possible with rifabutin.

No studies addressed surrogate markers of contraceptive effectiveness for rifaximin or rifalazil. In one fair quality study, rifaximin did not alter systemic progestin or EE exposure, as would be expected, as this drug is not orally absorbed and does not circulate systemically.²⁰ One poor quality study indicated no change to EE parameters when COCs were administered with rifalazil. Rifalazil was developed partially because it is not an inducer of CYP P450 enzymes and therefore was expected to have fewer drug interactions than rifampin. However, it was never brought into clinical use due to its side effect profile.⁸ We did not identify any studies that addressed the remaining FDA-approved rifamycin, rifapentine, with HC. This drug has an intermediate level of CYP 3A4 induction compared with rifampin and rifabutin, so a similar degree of interaction would be expected.⁸

Combined HCs also have the ability to affect metabolism of co-administered drugs, as EE is a known moderate inhibitor of several CYP P450 enzymes.⁶ One poor quality study did not observe significant changes in rifampin PK when administered with COCs,²⁶ and another poor quality study observed no difference in TB treatment response among COC users and non-users.²³

This systematic review has several strengths, including strict inclusion criteria requiring that all studies have a comparison group. COCs have a typical-use 1 year failure rate of 5–9%,^27,28 so it is inappropriate to conclude that COC failures in women using rifamycins are always due to drug interaction, as was proposed in older case series and uncontrolled observational studies.⁵ Likewise, due to our inclusion of multiple pertinent outcomes, we were able to evaluate consistency of findings between PK and clinical outcomes.

However, this review is limited by the quantity and quality of published evidence. Importantly, no studies measured the most pertinent outcome of pregnancy during concurrent COC/rifamycin use, so our conclusions are based on surrogate outcomes for pregnancy risk. Policy-makers and clinicians need to consider this important limitation. Studies that addressed ovulation are limited by small sample sizes, and in some cases infrequent or poorly timed measurements of serum progesterone, which may have led to inadequate detection of ovulation. Serum progesterone is a surrogate marker for ovulation, and no studies used daily ultrasound to confirm ovulation.¹⁹ PK studies had weaknesses including not randomising, not addressing drug adherence, small sample sizes, use of statistical comparisons that do not take into account therapeutic bioequivalence, and lack of attention to potential confounders such as body mass index. No studies addressed the combination of rifamycins with non-oral contraceptive formulations such as the transdermal patch, vaginal ring, injectables or progestin implants. Finally, data on newer rifamycins were limited to single PK studies, limiting the ability to extrapolate these findings clinically.

The WHO and US Medical Eligibility Criteria for Contraceptive Use consider the use of rifampin and rifabutin with oral, patch, and ring contraceptives to be category 3, meaning that theoretical or proven risks generally outweigh advantages, due to a presumed reduction in contraceptive effect.^7,29 However, the relative risk of pregnancy in HC users taking rifamycins compared with those not taking rifamycins is unknown. Women should be informed of the theoretical risk of drug interactions but should not be denied use of any method given this important knowledge gap. In the absence of data, recommendations for injectable and implantable contraceptive methods with rifamycins are category 1 and 2, respectively, meaning safe or generally safe to use, with the caveat that contraceptive effect may be reduced with contraceptive implants. The effectiveness of progestin-only injectables appears to be unaffected by enzyme-inducting medications, due to the higher systemic exposure to progestin.³⁰ Given the large number of women on rifamycin therapy worldwide, many of whom are of reproductive age, further research on interactions between rifamycins and injectable/implantable contraception is urgently needed. Intrauterine contraception does not rely on systemic drug levels and is also category 1 (safe) for women with non-pelvic TB.^7,29