Description of studies

Twelve randomized controlled trials met inclusion criteria. The 2014 update yielded 4 new studies (Legro 2008; Machado 2010a; Jensen 2012; Stephenson 2013), with a total of 1109 participants.

Cachrimanidou 1993: Multicentered trial conducted in three centers in Sweden, funded by Organon. It compared COC regimens containing 30 µg ethinyl estradiol/150 µg desogestrel administered in 28‐day (N=96) versus 70‐day (N=198) cycles for 1 year.

Coutinho 1995: Multicentered trial in Brazil, China, Egypt. This study compared vaginal dosing of COCs with 50 mg ethinyl estradiol/250 mg levonorgestrel administered in 28‐day (N=454) versus 1 year (N=446) cycles for 1 year.

Miller 2001: Multicentered trial conducted in four clinical sites in Seattle, Washington. Compared 30 µg ethinyl estradiol/300 µg norgestrel COCs administered in 28‐day (N=44) versus 49‐day (N=46) cycles for 1 year.

Anderson 2003: Multicentered trial from 47 U.S. sites, funded by Barr. Compared COCs with 30 µg ethinyl estradiol and 150 µg levonorgestrel administered in 28‐day (N=226) versus 91‐day (N=456) cycles for 1 year.

Kwiecien 2003: This study from one site in Portland, Oregon compared COC regimens containing 20 µg ethinyl estradiol and 100 µg levonorgestrel, administered in 28‐day (N=16) versus 168‐day (N=16) cycles for 6 months.

Miller 2003: Trial from one site in Seattle, Washington, funded by Wyeth. Analysis of 20 µg ethinyl estradiol and 100 µg levonorgestrel administered in 28‐day (N=40) versus 336‐day (N=39) cycles for 1 year.

Miller 2005: Multicentered trial (10 European and 10 U.S. sites) funded by Organon. This study compared four different regimens of the contraceptive vaginal ring (15 μg ethinyl estradiol/120 μg etonogestrel): 28‐day (N= 108), 49‐day (N = 107), 91‐day (N= 105) and 364‐day (N = 109) cycles for 1 year.

Stewart 2005: Multicentered trial conducted in 9 U.S. sites, funded by Ortho‐McNeil. Compared the transdermal patch (0.75 mg ethinyl estradiol/6 mg norelgestromin) administered in 28‐day cycles (n = 81) versus 91‐day cycles (n = 158).

Legro 2008: Double‐blinded trial from one center in Hershey, Pennsylvania of COC regimens containing 20 µg ethinyl estradiol/1 mg norethindrone acetate administered in 28‐day (N=31) versus 168‐day (N=31) cycles for 6 months.

Machado 2010a: Multicentered trial conducted at three health centers in Brazil, funded by Libbs. This study compared COC regimens containing 30 µg ethinyl estradiol/3 mg drospirenone administered in 28‐day (N=39) versus 168‐day (N=39) cycles for 6 months.

Jensen 2012: Nested randomized clinical trial conducted at 84 U.S. sites, funded by Bayer. This study evaluated 20 µg ethinyl estradiol/3 mg drospirenone COCs administered in a flexible regimen (continuous for at least 24 days, with 4‐day hormone free interval initiated anytime from cycle day 25‐120; N = 234) versus 28‐day cycles (24 days active hormone with 4‐day hormone‐free interval; N = 232) for 1 year.

Stephenson 2013). Multicentered trial conducted at 8 sites in the U.K. This study compared COCs containing 30 µg ethinyl estradiol and 150 µg levonorgestrel administered administered in a tailored regimen (continuous until 3 days of bleeding triggers 3‐day hormone free interval;
N = 251) versus 28‐day cycles (N = 252) for 1 year.

Two used the same COC formula of 20 μg ethinyl estradiol and 100 μg levonorgestrel for both their continuous and traditional regimens (Kwiecien 2003; Miller 2003), and two others used 30 μg ethinyl estradiol and 150 μg levonorgestrel (Anderson 2003; Stephenson 2013). The remainder used a variety of formulations and delivery systems.

All but one eligible study defined cyclic administration as a 28‐day cycle with 21 days of active pills and 7 days of placebo. The 28‐day cycle in Jensen 2012 had 24 active pills and 4 days of placebo. The length of the continuous administration varied greatly: 70 days for five cycles (Cachrimanidou 1993), 365 days (Coutinho 1995), 49 days for four cycles (Miller 2001), 91 days for four cycles (Anderson 2003), 168 days (Kwiecien 2003; Legro 2008; Machado 2010a), 336 days (Miller 2003), 91 days (Stewart 2005), and several different continuous cycle arms including 49, 91, and 364 days (Miller 2005). Jensen 2012 compared cyclic COC administration to a flexible regimen that allowed women to control the timing of withdrawal bleeding by initiating a 4‐day hormone‐free interval at any time during days 25‐120 of the cycle. Subjects were instructed to observe the 4‐day break at any time, no more frequently than once every 28 days, and at least once every 124 days. Stephenson 2013 randomized subjects to a traditional 28‐day cyclic COC, or a tailored regimen with daily use until 3 days of consecutive bleeding (requiring use of sanitary protection more than a panty liner) initiated a 3‐day hormone‐free interval (after at least 21 days of continuous use). CHCs were given orally in most of the studies except for Coutinho 1995 who dosed pills via the vaginal route, and in the studies of the contraceptive ring and patch (Miller 2005; Stewart 2005).

Analysis and reporting of bleeding patterns varied in each study. Cachrimanidou 1993 and Miller 2005 defined 'spotting' as requiring no or at most one sanitary napkin per day and 'bleeding' as requiring at least two sanitary pads per day. Most authors (Coutinho 1995; Miller 2001; Anderson 2003; Kwiecien 2003; Miller 2003; Stewart 2005; Machado 2010a; Jensen 2012) evaluated bleeding using definitions adapted from the World Health Organization (WHO) (Belsey 1986). The WHO bleeding definitions state that spotting is bloody vaginal discharge that does not require protection and bleeding requires protection. In Cachrimanidou 1993, diaries were analyzed in 70‐day blocks for continuous use and 84‐day blocks for cyclic use (withdrawal bleeding days were analysed separately from bleeding and spotting days). Coutinho 1995 and Jensen 2012 analysed diaries in 90‐day reference periods (withdrawal bleeding days were included) as recommended by the WHO. Miller 2001 and Miller 2003 analyzed diaries in 84‐day blocks or trimesters (withdrawal bleeding days were included). In Miller 2001, since 49 days does not divide equally into an 84‐day block, one withdrawal bleeding week was included in the first and fourth trimesters and two in the second and third trimesters whereas for the 28‐day cycle, this included three 28‐day cycles with three withdrawal weeks. Kwiecien 2003 analyzed diaries in 28‐day blocks but provided us with information that allowed analysis in 84‐day blocks (withdrawal bleeding days were included). Anderson 2003 analysed diaries in 364‐day blocks (withdrawal bleeding days included). Stewart 2005 analyzed bleeding in 84‐day blocks and also evaluated the duration of the withdrawal bleed. Miller 2005 used 91‐day reference periods (withdrawal bleed included) to analyze bleeding patterns. Legro 2008 and Machado 2010a analyzed bleeding in 28‐day blocks. In Legro 2008, number of bleeding days was only reported for the full 168‐day cycle. Machado 2010a reported rates of amenorrhea and spotting during the sixth 28‐day cycle. Stephenson 2013 reported bleeding episodes (those less than 3 days, and 3 days or more) and bleeding days for the entire one year study period, and calculated bleeding rates per month and incidence rate ratios.

A large majority of women enrolled were 'switchers' (those who were on COCs during the cycle prior to entering the study) versus 'new starts' (never taken COCs). Cachrimanidou 1993, Miller 2001, Anderson 2003, Miller 2003, and Stephenson 2013 reported a 60% or higher percentage of COC switchers in both the continuous and traditional groups. Stewart 2005 and Jensen 2012 reported 45% or higher of CHC switchers in both the continuous and cyclic dosing groups. Coutinho 1995 and Kwiecien 2003 did not specify the percentage of switchers per group but Coutinho 1995 reported that 91% of study participants were on some type of birth control prior to study entry and Kwiecien 2003 only reported prior use of COCs [traditional group 10/16 (62.5%); continuous group 11/16 (68.7%)]. Miller 2005, Legro 2008, and Machado 2010a only enrolled women not currently using hormonal contraception. In Miller 2005, women started the study protocol with a "run‐in" cycle or 3 weeks of contraceptive ring use followed by a 7‐day hormone‐free interval.

Endometrial safety of extended‐cycle CHCs was monitored in four studies (Kwiecien 2003; Miller 2003; Miller 2005; Legro 2008) by measuring endometrial thickness with a vaginal probe ultrasound and/or endometrial biopsy. Kwiecien 2003 and Miller 2003 defined a normal endometrial thickness to be less than 5 mm while on CHCs, as this is the cut‐off value for concern for abnormal pathology in postmenopausal women. Miller 2005 used an endometrial thickness of 10 mm or greater as the trigger to obtain an endometrial biopsy. They also performed endometrial biopsies on 10 subjects in each of the four trial arms. Legro 2008 measured endometrial thickness on ultrasound and performed endometrial biopsies, but did not specifically define normal values. All subjects underwent transvaginal ultrasound and endometrial biopsy prior to initiation of study medication, and at the end of the sixth 28‐day cycle.

Risk of bias in included studies

Information regarding randomization and allocation concealment obtained from the publications and written correspondence with the authors proved these two areas to be adequate in most of the included studies. Eleven studies reported the use of computer‐generated randomization schemes, while one failed to report the method of randomization (Cachrimanidou 1993). Allocation concealment was by sequential, sealed, opaque envelopes (Kwiecien 2003; Miller 2001; Machado 2010a); sealed opaque envelopes (Cachrimanidou 1994); sequential, sealed brown bags (Miller 2003). Allocated was by centralized automated telephone system in Miller 2005 Allocation concealment was not reported in Anderson 2003. In Jensen 2012, subjects were provided randomization numbers via an interactive voice response system. Allocation was produced online in Stephenson 2013. Three studies (Coutinho 1995; Stewart 2005; Legro 2008) provided no information regarding their concealment techniques. Once allocation to treatment groups had occurred, actual treatment was unblinded for both participants and investigators in all of the studies, except for Legro 2008. In Legro 2008, subjects and investigators were blinded by over‐encapsulating the pills during the fourth week of each 28‐day cycle and repackaging into 28‐day dispensing packs. Eight studies used an intent‐to‐treat analysis (Kwiecien 2003; Miller 2003; Miller 2005; Stewart 2005; Legro 2008; Machado 2010a; Jensen 2012; Stephenson 2013). One study excluded participants who did not complete full trimesters (84‐day blocks) of study participation (Miller 2001). Another study excluded patients from Pearl‐index calculations who were noncompliant with their assigned pill‐dosing regimen (Anderson 2003). Coutinho 1995 excluded data from participants in the continuous group who decided to have a withdrawal bleed. The analysis in Cachrimanidou 1993 was unclear regarding intent‐to‐treat.

Effects of interventions

Satisfaction: Participant satisfaction was reported in ten studies, but was inconsistently measured. Cachrimanidou 1993 obtained satisfaction data only from 'ever‐users' of contraception in the extended cycle arm (63% of whom reported a preference for an extended cycle). Coutinho 1995 did not report satisfaction outcomes except for non‐medical reasons for study discontinuation (e.g., dislikes method). Miller 2001 and Miller 2003 used a five‐point Likert scale to determine participant satisfaction, but neither trial found a significant difference in satisfaction between study arms. Anderson 2003 reported that participants from both study groups expressed a preference for fewer menstrual periods, but there was no analysis comparing the groups. Kwiecien 2003 used 10‐cm visual analog scales (VAS) with the anchors unsatisfied and very satisfied to determine participant satisfaction. Participants in both groups were very satisfied, and there was no difference between groups (p = 1.0). Miller 2005 used a questionnaire to determine overall satisfaction with the contraceptive ring and with the bleeding profile. Overall satisfaction with the ring in all 4 groups was high but shorter cycles (28 and 49 days) demonstrated a trend towards higher levels of satisfaction. Stewart 2005 had participants evaluate overall satisfaction using a 5‐point Likert scale and then a final overall assessment of the drug regimen using a 4‐point Likert scale. Overall satisfaction was not significantly different between groups with the majority (>86%) being somewhat or very satisfied. The majority of both groups rated the overall assessment of the drug regimen as good or excellent. Jensen 2012 reported that satisfaction was similar between groups, with most women either much satisfied or very much satisfied. In Stephenson 2013, participants were queried regarding their bleeding satisfaction and overall satisfaction with the regimen, selecting either satisfied, dissatisfied, or indifferent. Satisfaction was lower in the continuous (tailored) group for both bleeding (p=0.03) and overall (p=0.007). Legro 2008 and Machado 2010a did not measure participant satisfaction.

Discontinuation: Overall study discontinuation (including lost to follow‐up) was higher in extended or continuous cycles as compared to traditional 28‐day cycles in two studies [(Anderson 2003, OR 1.6; 95% CI 1.2 to 2.3), (Miller 2005, 91‐day cycle OR 2.02; 95% CI 1.13‐3.61; 364‐day cycle OR 2.28; 95% CI 1.29‐4.03)]. Otherwise, the extended versus cyclic groups did not differ in overall study discontinuation [(Cachrimanidou 1993, OR 1.4; 95% CI 0.9 to 2.4), (Coutinho 1995, OR 1.0; 95% CI 0.7 to 1.5), (Miller 2001, OR 0.7; 95% CI 0.3 to 1.6), (Kwiecien 2003, OR 0.3; 95% CI 0.04 to 2.6), (Miller 2003, OR 0.5; 95% CI 0.2 to 1.5), (Stewart 2005, OR 1.45; 95% 0.73‐2.89), (Miller 2005, 49‐day cycle OR 1.29; 95% CI 0.7‐2.38), (Legro 2008, OR 1.24; 95% CI 0.34 to 4.53), (Machado 2010a, OR 1.15; 95% CI 0.41 to 3.21), (Jensen 2012, OR 1.07; 95% CI 0.72 to 1.59), (Stephenson 2013, OR 0.91; 95% CI 0.64 to 1.3)]. However, in an analysis by Stephenson 2013 that excluded women who were lost to follow up, discontinuation of the COC as allocated was higher in the continuous (tailored) group (p=0.02).

Discontinuation due to bleeding problems:Anderson 2003, Cachrimanidou 1993, Miller 2005, Machado 2010a, and Stephenson 2013 had more discontinuations due to bleeding problems in the extended or continuous cycle arm [Anderson 2003, OR 3.0 (95% CI 1.5 to 5.9), Cachrimanidou 1993, OR 3.6 (95% CI 1.6 to 8.2), Miller 2005, 49‐day cycle OR 7.75 (95% CI 1.32 to 45.48), 91‐day cycle OR 8.59 (95% CI 2.8 to 26.3), 364‐day cycle OR 8.87 (95% CI 3.54‐22.21), Machado 2010a, OR 8.01 (95% CI 1.09 to 59.17), Stephenson 2013, OR 3.79 (95% CI 1.75 to 8.22)]. The remaining seven studies did not show a difference.

Adherence: Participant adherence data were reported in five studies. Miller 2001, who defined compliance by the number of missed or late pills, found no difference between groups (28‐day = 1.3 and 49‐day = 1.2, p = 0.5). Anderson 2003 determined compliance as the percentage of total study days when participants took the designated pill for the given day. Compliance rates for both groups were not different (OR 0.7; 95% CI 0.3 to 1.3). Miller 2003 reported compliance data only for participants who missed three or more pills during the first and last trimesters (84‐day interval) of the study. No statistically significant difference between groups in either trimester was found. Miller 2005 defined compliance as hours of ring use (no more than 48 hours difference over 3 weeks) and nonuse (no more than 24 hours longer than the 7‐day hormone free interval). Compliance with ring use was not reported separately for each arm but was reported cumulatively as high (88.6‐98.9%). As compared to the 28‐day cycle, the 49‐day and 91‐day cycles were similarly compliant with the 7‐day hormone free interval. Stewart 2005 designated perfect compliance as a patch on every day during the treatment period (either 28 or 84 days) and a hormone‐free interval of no longer than 7 days. No difference between groups was found (OR 0.6; 95% CI 0.3‐1.19). In Jensen 2012 and Stephenson 2013, adherence was only reported for the continuous (flexible or tailored) arm.

Pregnancy: The number of pregnancies reported were extremely low and ranged from number 0 to 4. None of the studies were adequately powered for this outcome. The risk of pregnancy did not differ between regimens except in one trial that showed fewer pregnancies in the continuous‐administration group (Coutinho 1995, OR 0.1; 95% CI 0.02 to 0.97).

Endometrial assessment: Assessment of the endometrium by ultrasound and/or endometrial biopsy was done by Kwiecien 2003, Miller 2003, Miller 2005, and Legro 2008. In Kwiecien 2003, 14 women in the extended group (n=16) volunteered to undergo endometrial stripe assessments by ultrasound; the mean thickness was 3.3 mm (SD 0.73, range 2 to 4 mm). No participant met criteria for biopsy (stripe > 5 mm). Miller 2003 evaluated a volunteer subset of the study (28‐day cycles, n = 7; 336‐day cycles, n = 9), and two participants in the 336‐day cycle group underwent assessments because of prolonged bleeding. All evaluations were performed during cycle nine. All endometrial stripe measurements were less than 5 mm and no evidence of hyperplasia was found on biopsy. Miller 2005 obtained endometrial stripe measurements at study entry and exit in the majority of subjects. No differences between the 4 study arms or time points was found and measurements were well below the 10 mm safety cutoff [exit measurements: 28‐day 3.4 mm (SD 2.3), 49‐day 4.0 mm (SD 2.4), 91‐day 4.5 mm (SD 2.2), 364‐day 3.6 mm (SD 2.0)]. Miller 2005 also performed endometrial biopsies on 10 volunteer subjects in each of the 4 study arms but no results were reported. Legro 2008 assessed endometrial thickness by ultrasound and performed endometrial biopsies in most participants at baseline and study end. Both groups were noted to have a significant decrease in endometrial thickness at study end (mean difference ‐5.4 mm (95% CI ‐6.4 to ‐4.5) in the cyclic group and ‐6.0 mm (95% CI ‐6.7 to ‐5.2) in the extended group. There were no significant differences between groups in distribution of tissue histology. All subjects had atrophic, inactive, inadequate, OCP effect, or secretory histology at study end.

Bleeding patterns: Bleeding patterns were the main outcomes for eleven studies. The primary outcome in Stephenson 2013 was COC continuation, but bleeding episodes and bleeding quantity were also assessed. Most trials showed either no difference between groups or less bleeding and/or spotting with extended or continuous dosing of COCs. Women in the continuous arm in the Coutinho 1995 trial reported a mean of 10.7 less total bleeding days (95% CI ‐11.3 to ‐10.4) in the first trimester than the cyclic arm. The weighted mean differences for the second to fourth trimesters were similar. Miller 2001 found no statistically significant differences in the mean number of bleeding or spotting days for the four trimesters with one exception: the extended group had a mean of 4.5 and 3.9 less bleeding days (95% CI ‐7.1 to ‐1.9; 95% CI ‐6.8 to ‐1.04) in the first and third trimesters than the cyclic dosing group. Anderson 2003 evaluated bleeding patterns over the entire 364‐day study period, and reported no significant differences between groups for the mean bleeding plus spotting days. However, the extended arm had less bleeding‐only days (WMD ‐14.3; 95% CI ‐17.7 to ‐11.0). Kwiecien 2003 found no difference between groups in the mean total bleeding days (bleeding plus spotting) and mean spotting days in either of the study's two (84‐day) trimesters, but the extended group had less mean bleeding‐only days than the cyclic group for both the first (WMD ‐7.7; 95% CI ‐14.0 to ‐1.5) and the second trimester (WMD ‐8.9; 95% CI ‐12.8 to ‐4.9). No clear picture regarding bleeding patterns emerged for the Miller 2003 trial. The groups did not differ significantly for the mean total bleeding days during the four (84‐days) trimesters. While the cyclic group had fewer mean spotting days during the first two trimesters (but not the last two trimesters), women in the extended group had less mean bleeding days for the four trimesters. Miller 2005 provided bleeding days in mean percentages and medians over 91‐day intervals but no information regarding amenorrhea. Direct comparisons between study arms using the median data cannot be made as each of the 91‐day intervals had a different number of ring‐free days. No significant differences in the percentage of total bleeding days were found between the 28‐day group and the 3 extended dosing arms. Stewart 2005 presented bleeding data only in median days and the percentage of each study arm that achieved amenorrhea (not the number of amenorrhea days) for days 1 to 84 and days 1 to 56. For days 1 to 84, median total bleeding days were no different (extended = 14 days, cyclic = 16 days, p = 0.407) but in the extended group, median bleeding‐only days were significantly less (extended = 6 days, cyclic = 14 days, p < 0.001) and percentage achieving amenorrhea were greater (OR 11.04, 95% CI 1.45‐84.03). For days 1 to 56, the extended‐cycle group had significantly less median total bleeding days (extended = 6, cyclic = 10, p = 0.009), median bleeding‐only days (extended = 1, cyclic = 9, p < 0.001), and a greater percentage achieving amenorrhea (OR 30.33, 95% CI 4‐224.9). Cachrimanidou 1993 analyzed bleeding outcomes differently from the five previous studies in that they analyzed bleeding during the withdrawal week separately. Bleeding associated with the withdrawal week was decreased for the 70‐day cycle compared to 28‐day cycle, but for mean bleeding and mean spotting days (analyzed over 70 days for the 70‐day cycle and 84 days for the 28‐day cycle), there were less days for the standard 28‐day cycle group (p < 0.05). Legro 2008 found no significant difference in mean number of bleeding days between groups for the 168‐day study period (mean difference 3.7 days, 95% CI ‐6.2 to 13.6), but reported that the extended‐cycle group had fewer moderate/heavy bleeding days (mean difference 5.8 days, 95% CI 1.8 to 9.7 days). Bleeding decreased significantly over time in both groups. Machado 2010a did not present the number of bleeding days for participants, but found that 62.2% (95% CI 46.6 to 77.8) of women in the extended group were amenorrheic during the 6th 28‐day cycle, compared to 17.7% (95% CI 4.9 to 30.5) in the cyclic group (p=0.0001). The percentage of subjects reporting spotting was greater among the extended‐cycle group only during the 3rd through 6th cycles. In Jensen 2012, the mean number of total bleeding days (bleeding and spotting) was similar between groups in the first reference period, but was significantly less for the continuous (flexible) group in the second (WMD ‐2.3, 95% CI ‐4.38 to ‐0.22), third (WMD ‐3.4, 95% CI ‐5.58 to ‐1.22), and fourth (‐4.1, 95% CI ‐6.26 to ‐1.94) reference periods. Bleeding days (requiring sanitary protection) were consistently less in the flexible group in all reference periods. Spotting only days were similar in the first and second trimester, but the cyclic group had less in the third and fourth trimesters. Stephenson 2013 reported bleeding incidence rates rather than mean bleeding days, and found that women in the cyclic group had 1.06 bleeding episodes per month compared to 0.56 in the continuous (tailored) group (p<0.001). The rate of bleeding episodes lasting 3 days or more was significantly less in the tailored group (p<0.001), while the rate of shorter episodes (less than 3 days) was less in the cyclic group (p<0.001). Women in the cyclic arm reported bleeding on 15.9% of all days, compared to 9.6% of all days in the tailored arm (p<0.001).

Menstrual‐associated symptoms: Seven studies reported data on menstruation‐associated symptoms, which were collected with participant diaries (Cachrimanidou 1993; Miller 2001; Kwiecien 2003; Machado 2010a) or by questionnaire at set time points (Miller 2005; Legro 2008; Stephenson 2013). Stewart 2005 and Jensen 2012 only collected symptoms as adverse events (see adverse events). Cachrimanidou 1993 monitored the increased or decreased frequency of headaches, nervousness, nausea, dizziness, depression, acne, and dysmenorrhea. The extended‐cycle arm showed benefit for headache frequency (p < 0.05), but otherwise no important differences in other menstruation‐associated symptoms were found. Miller 2001 monitored the presence and severity of cramping, tiredness, headache, breast tenderness, and genital irritation. Genital irritation (p = 0.02), headache (p = 0.04), and tiredness (p = 0.05) were less severe in the 49‐day cycle group than in the cyclic group but otherwise no differences were found between groups. Kwiecien 2003 collected data on headache, nausea, bloating, breast tenderness, premenstrual syndrome, and menstrual pain. No significant difference was found between groups except that women in the extended group were less likely to report bloating (OR ‐11.0; 95% CI ‐19.8 to ‐2.2) and menstrual pain (OR ‐11.5; 95% CI ‐18.4 to ‐4.5) than those in the continuous regimen. Miller 2005 asked participants at 6 months and at study exit regarding the presence or absence of premenstrual syndrome, headaches, and pain with menses as compared to past contraceptive use. No differences between study groups was found. Legro 2008 had participants complete the Moos Menstrual Distress Questionnaire at baseline and at the end of the last study cycle. This questionnaire assesses scales for pain, water retention, autonomic reactions, negative affect, impaired concentration, behavior change, arousal, and control. Subjects in the extended‐cycle group reported greater improvements in menstrual pain (p=0.01) and behavior change (p=0.04) compared to the cyclic group at the end the 6‐month study period. Machado 2010a evaluated menstrual‐related symptoms using participant diaries. Presence and intensity of acne, dysmenorrhea, headache, mastalgia, nausea, irritability, edema, increased appetite, and breast swelling were measured. Only women in the extended‐cycle group were found to have significant decreases in dysmenorrhea, headache, nausea, and acne. They also were found to have increased appetite. Edema decreased in both groups, while mastalgia decreased only in the cyclic group. Stephenson 2013 asked subjects to report side effects at 12 months, including bloating, weight gain, headaches, skin changes, mood changes, abdominal cramps, tiredness, loss of interest in sex, breast tenderness, and aches and pains. No significant difference was found between groups in the intention‐to‐treat analysis. Loss of interest in sex was reported by a greater percentage of women in the continuous (tailored) arm, when analysis was restricted to women taking the COC as allocated (p=0.04).

Adverse events: No serious adverse events related to the study medication or regimen were reported in ten trials (Cachrimanidou 1993; Coutinho 1995; Miller 2001; Kwiecien 2003; Miller 2003; Stewart 2005; Legro 2008; Machado 2010a; Jensen 2012; Stephenson 2013). Anderson 2003 reported three adverse events likely related to study drug, including a pulmonary embolism (extended cycle), cholecystitis (28‐day cycle), and an exacerbation of pre‐existing depression (28‐day cycle). The adverse event data also showed less headaches in the extended‐cycle group than the traditional group (OR 0.7; 95% CI 0.5 to 1.0). Miller 2005 also reported three adverse events likely related to study drug, including cholecystitis with cholelithiasis (49‐day cycle), a known leiomyoma who underwent hysterectomy (364‐day cycle), and a deep venous thrombosis (91‐day cycle, testing positive for protein S deficiency and anticardiolipin antibodies). Although not statistically significant, Stewart 2005 found more headaches, breast tenderness, and nausea in the extended (84‐day cycle) than in the cyclic patch group.