Results of the search

The 2014 search produced 127 unduplicated citations from the main databases. In addition, we found 22 unduplicated trials through ClinicalTrials.gov and ICTRP. We included one new trial, which had been 'ongoing' in the previous update (Taylor 2014). Also, we added a new ongoing trial (WHO 2014).

We identified 22 trials that evaluated contraceptive effectiveness of repeated pre‐ and postcoital use of hormonal drugs. These trials included 12,407 women in Europe, Asia, and the Americas. Although nine studies evaluated more than one hormone regimen, we considered each arm of these studies separately, for the reasons explained in Risk of bias in included studies.

• Ten trials investigated pericoital use of a tablet containing LNG 0.75 mg:

  • nine examined postcoital use (Chernev 1995; He 1991; Klawe 1984; Kliment 1986; Nirapathpongporn; Sas; Seregely 1982; WHO 1987; WHO 2000); and

  • one studied pericoital use (Taylor 2014).

• One trial studied a) postcoital use of LNG 0.4 mg and b) LNG 0.75 mg as a divided dose pre‐ and postcoitally (Canzler 1984).

• Four trials evaluated postcoital use of LNG in one or more doses other than 0.75 mg (Echeverry 1974; Kesseru 1973; Larranga 1975; Schering 1978).

• Six trials evaluated pericoital use of one or more hormones other than LNG (Cox 1968; Mischler 1974; Rubio 1970; Szontagh 1969; Zanartu 1974; Zanartu 1976).

• One trial evaluated postcoital use of a different hormonal drug and a dose of LNG other than 0.75 mg (Moggia 1974).

Included studies

The included studies are briefly described below. In this review, we discuss each arm of Moggia 1974 and Canzler 1984 separately, with studies of comparable drugs and doses. Additional details can be found in Characteristics of included studies.

Excluded studies

We excluded 10 trials from our analysis. Four reports did not contain information on treatment regimen (Hetenyi 1988; Kulakov 1983; Szczurowicz; Unzeitig 1989). Two reports did not provide either pregnancy rates (Pearl indices) or sufficient data to allow us to calculate them (Krymskaya 1983; Serov 1983). One trial did not report pregnancy outcome (Orley). We excluded another trial, originally published in Bulgarian, after several attempts to have it translated (Vasilev 1983). We excluded two studies because we could not locate their full reports (published or unpublished). The information available was insufficient to evaluate the study quality and analyze the results from an abstract (Hurtado 1975) and an incomplete report (Czekanowski).

Incomplete outcome data

Seven of the 22 reports specified an intended duration of method use (Table 1). In the three studies in which this duration was less than six months, all women treated completed the intended use period (He 1991; Szontagh 1969; WHO 1987). In the other four studies, in which the intended duration of method use was six months (Chernev 1995; Taylor 2014; WHO 2000) or 10 months (Nirapathpongporn), 25% to 67% of participants used the method for the intended use period. Three of these study reports cited both the number of women who were known to have discontinued the method during observation and the number who were lost to follow‐up. Losses to follow‐up were as follows: WHO 2000, 4% by six months; Nirapathpongporn, 49% by 10 months; Taylor 2014, 18% by 6.5 months. A study with a low completion rate (33% at six months) did not provide data separately on the proportions of women who discontinued early or were lost (Chernev 1995).

Two reports did not specify an intended duration of follow‐up, but nevertheless presented detailed information about continuation, discontinuation, and loss to follow‐up in each month or three‐month period. In one of these trials, 68% of women completed at least six months of use, and 11% were lost to follow‐up in that time (Kesseru 1973). In the other trial, the corresponding figures were 60% and 27% (Larranga 1975).

The other 13 reports did not include sufficient information to allow an assessment of the ascertainment of outcome data. Loss to follow‐up may result in underestimation of the pregnancy rates during method use because women who were lost may have had undetected pregnancies while still using the method after the last study contact.

Selective reporting

All trials included in this review clearly defined their main objectives and interventions, although some trials lacked clear description of the drug regimen or how the drug was dispensed to participants. Only eight studies clearly specified the main study outcomes (Canzler 1984; He 1991; Kesseru 1973; Moggia 1974; Seregely 1982; Taylor 2014; WHO 2000; Zanartu 1976). Based on these eight reports as well as the reported results in other trials, pregnancy and side effects including bleeding problems were the main study outcomes in all included trials. In addition, several studies evaluated continuation rates and reasons for discontinuation (Echeverry 1974; Kesseru 1973; Larranga 1975; Moggia 1974; Nirapathpongporn; Schering 1978; Taylor 2014; WHO 1987; WHO 2000; Zanartu 1976). Two trials evaluated acceptability through a questionnaire in addition to measuring discontinuation (Taylor 2014; WHO 2000).

Most studies lacked detailed description of the trial procedures. Ten reports mentioned the frequency of follow‐up contacts. Only two studies specified how pregnancy was ascertained (He 1991; Taylor 2014). Although menstrual irregularities were one of the main outcome in all trials, only 10 trials reported on how the data on bleeding patterns were collected and evaluated (Canzler 1984; Echeverry 1974; He 1991; Larranga 1975; Moggia 1974; Seregely 1982; Taylor 2014; WHO 2000; Zanartu 1974; Zanartu 1976).

Other potential sources of bias

Only 14 reports specified inclusion and exclusion criteria (Canzler 1984; Echeverry 1974; He 1991; Kesseru 1973; Klawe 1984; Mischler 1974; Nirapathpongporn; Rubio 1970; Seregely 1982; Taylor 2014; WHO 1987; WHO 2000; Zanartu 1974; Zanartu 1976). Some of these reports failed to define these criteria clearly; for instance, two studies included only 'fertile' women but did not explain how fertility was assessed. The uniformity of participants' characteristics in some studies suggested that additional unstated criteria may also have been used. Nine studies did not specify any eligibility criteria (Chernev 1995; Cox 1968; Larranga 1975; Moggia 1974; Sas; Schering 1978; Szontagh 1969; Zanartu 1974; Zanartu 1976). The descriptions of the study populations in some studies are unclear and in others are limited or nonexistent.

Seven studies had sample sizes of fewer than 100 women (Canzler 1984; Cox 1968; Klawe 1984; Kliment 1986; Sas; Szontagh 1969; Taylor 2014), whereas three included more than 1000 women (Kesseru 1973; Mischler 1974; Seregely 1982). As previously noted, two trials had large sample sizes but a short duration of follow‐up (one treatment cycle), which limited the exposure to both pregnancy and the drug (He 1991; WHO 1987).

A few reports commented on possible relations between the treatment effect and dose, coital and dosing frequency, duration of treatment and time elapsing between coitus and pill intake. Canzler 1984 evaluated the potential role of the dose, coital frequency, time after coitus, duration of medication and the number of tablets taken; Kesseru 1973 explored length of treatment, coital frequency and the number of pills taken; Schering 1978 assessed the possible role of duration of treatment; Rubio 1970 evaluated the dose; and Mischler 1974 evaluated both the dose and number of pills taken per cycle. However, none of these associations were evaluated rigorously (statistically). Also, some such associations noted may be spurious. For example, an association between the number of tablets taken per cycle and the likelihood of side effects in a cycle may simply reflect the fact that women are more likely both to take more tablets and to have side effects in long cycles than in short cycles.

Most of the trials evaluated compliance with the treatment regimen and distinguished the results as drug or method failure. However, only eight trials described their methods of collection of adherence data (Canzler 1984; Echeverry 1974; He 1991; Larranga 1975; Moggia 1974; Seregely 1982; Taylor 2014; WHO 2000).

All studies evaluating doses of LNG other than 0.75 mg were sponsored by Schering AG (Echeverry 1974; Kesseru 1973; Larranga 1975; Moggia 1974). Canzler 1984 was co‐sponsored by another pharmaceutical company, VEB Jenapharm. The largest Hungarian trial to support introduction of LNG 0.75 mg for regular postcoital contraception was sponsored by its manufacturer, the pharmaceutical company Gedeon Richter, Ltd. (Seregely 1982).Three multicenter trials of LNG 0.75 mg for regular postcoital contraception were sponsored by the World Health Organization (WHO) (He 1991; WHO 1987; WHO 2000); Gedeon Richter, Ltd. provided pills for WHO 2000. In two trials, pills were provided by the pharmaceutical manufacturers but the trials were sponsored by the University of Chile (Zanartu 1974; Zanartu 1976).

Levonorgestrel (LNG) 0.75 mg

The 11 studies of the LNG 0.75 mg dose were conducted in numerous countries in Europe and Asia, as well as the USA, Brazil, and Cuba. The evaluated regimens of the drug were slightly different (see Characteristics of included studies).

The characteristics of the study populations varied considerably. Most of the trials had no age restriction or admitted a wide range of ages, but three enrolled only young (Canzler 1984; Chernev 1995; Seregely 1982) or only older women (Klawe 1984). In some studies, all participants met stringent criteria to demonstrate fertility, such as evidence of ovulation in prior cycles, or previous pregnancy with the current partner (Canzler 1984; He 1991; Kliment 1986; Nirapathpongporn; Sas; WHO 1987; WHO 2000), and some trials excluded women who had recently used hormonal contraceptives or who had a history of pelvic inflammatory disease (He 1991; Taylor 2014; WHO 1987; WHO 2000). Other trials admitted women with no evidence of fertility. Most of the trials excluded women who expected to have high coital frequency during the trial: in four trials the limit was four acts of sexual intercourse a month (Klawe 1984; Seregely 1982; Taylor 2014; WHO 2000), whereas others used a more general rule like "irregular" or "casual" or "infrequent" sex (Kliment 1986; Nirapathpongporn; Sas). Two trials did not limit coital frequency as their objective was to evaluate women at higher risk of conception (He 1991; WHO 1987). Canzler 1984 enrolled women who had up to 10 acts of sexual intercourse a month.

The studies evaluating pericoital use of LNG 0.75 mg included a total of 2700 women. The studies reported the duration of use of the method in cycles, months, or woman‐years (WY). If cycles and months were combined, the total number of such intervals in these studies was 13,514. Thus the average duration of use per participant was five cycles or months. However, two studies were designed to allow only one cycle of treatment (He 1991; WHO 1987). If those studies were excluded, the average duration of method use per participant was 6.3 cycles or months. Additional details on duration of treatment can be found in Table 1.

Coital frequency in these trials ranged from 1 to 15 acts per month or cycle. Average coital frequency in the eight studies that reported these data was about four acts per month or cycle, ranging from 2 to 7.5. More data on coital frequency and pill intake observed during the trial are presented in Table 2.

Doses other than LNG 0.75 mg

Six studies evaluated doses of LNG other than 0.75 mg. One study included five groups of women who received doses between 0.15 mg and 0.40 mg. The other five studies assigned women to only one dose of LNG (ranging from 0.35 mg to 1.0 mg) taken within a few hours after unprotected intercourse. In one trial, only one dose was recommended for any eight‐hour interval (Echeverry 1974). See Characteristics of included studies for more detail.

Inclusion criteria for most of these trials were much less stringent than for the trials of 0.75 mg tablets. None of the studies had a specific age restriction. In five of the trials all participants were parous, and in the other trial (Canzler 1984), all had biphasic menstrual cycles, suggesting the occurrence of ovulation. However, in three studies some or most participants were enrolled immediately postpartum or when lactating (Echeverry 1974; Larranga 1975; Moggia 1974). One study enrolled women who expected to have sex up to 10 times per month (Canzler 1984). Other trials enrolled women without regard to expected coital frequency.

The six trials included 5787 participants who used the method for a total of 53,347 cycles or months. The mean duration of use was thus 9.2 cycles or months per participant. The maximum duration of use in these trials ranged from 12 to 36 months. More details on duration of treatment can be found in Table 1. Coital frequency in these trials was higher than in the trials of levonorgestrel 0.75 mg. The available results on coital frequency and pill intake after admission to the trial are presented in Table 2. In most studies, the reported number of tablets taken was concordant with coital frequency, suggesting compliance with the assigned regimen.

Hormones other than LNG

Pregnancies per 100 WY in the trials that evaluated pericoital use of drugs other than LNG ranged from zero to 433.3. The highest rate was detected during the precoital use of megestrol acetate used up to 22 hours before intercourse, prompting the researchers to reduce the time interval between the pill intake and intercourse to a maximum of 14 hours. The pericoital use of several progestogens (e.g., ethynodiol diacetate, low doses of quingestanol acetate) was associated with high pregnancy rates, while use of other drugs resulted in reasonably low Pearl indices (Table 5). While most of the hormonal drugs other than LNG were not tested extensively in a large clinical trial, the postcoital use of quingestanol in doses ranging between 0.2 mg and 2 mg was evaluated in a total of 17,079 cycles in three large clinical trials (Mischler 1974; Moggia 1974; Rubio 1970). The Pearl indices ranged from zero to 168 pregnancies per 100 WY. We did not calculate a pooled pregnancy rate for all three studies because Mischler 1974 did not report the number of pregnancies. A pooled pregnancy rate for all quingestanol doses evaluated in Moggia 1974 and Rubio 1970 was 5.3 per 100 WY (95% CI 3.5 to 7.8). The use of the lowest doses of quingestanol was associated with the highest pregnancy rates.

These seven trials had a total of 3917 participants. Similar to the previous studies, the trials evaluating different drugs for pericoital contraception found that menstrual irregularities were the most common side effects. Several researchers noted that the postcoital regimens were well tolerated by the patients in spite of the menstrual problems (Rubio 1970; Zanartu 1974). However, Mischler 1974 concluded that the incidence of intermenstrual bleeding was "probably unacceptable" when quingestanol was used more than 12 times per cycles. The basis for this conclusion was unclear, because acceptability data were not described in the report.

Reported non‐menstrual side effects included occasional gastrointestinal symptoms (e.g., dyspepsia, nausea), breast discomfort, headaches and nervousness. All these side effects were mild and almost never caused discontinuation. None of these reports provided information on serious adverse events or pregnancy outcomes.