MEASURING CONTRACEPTIVE FAILURE James Trussell Office of Population Research Princeton University

Issues Efficacy versus effectiveness Typical versus perfect use Pearl index versus life table Non-completion of trial Common errors in literature Results from literature Communicating the risk of failure

Efficacy versus Effectiveness Efficacy: how well a method works under ideal circumstances Effectiveness: how well a method works in the real world Efficacy would be measured in a clinical trial whereas effectiveness would be measured in a survey or a chart review

Sources of Data Surveys: NSFG 1973, 1976, 1982, 1988, 1995, 2002 –Nationally representative –Retrospective –Underreporting of abortion –Overreporting of contraceptive failure leading to birth? Clinical trials –Hawthorne effect and inference beyond trial setting –Cycles of perfect use can be identified and pregnancy rates during perfect use can be estimated but adherence is self-reported

Results from the 1995 NSFG: % becoming pregnant in the first year of use uncorrected and corrected for underreporting of abortion MethodUncorrectedCorrected Pill Condom Spermicides Fu. Fam Plan Perspect 1999;31:56-63

Self-Reporting of Adherence Self reports on missed OCs compared with electronic recording on punched pills among 103 women for 3 cycles Agreement on only 45% of days Overreporting of no missed pills (53-59% versus 19-33%) Underreporting of missing 3+ pills (10-14% versus 30-51%) Potter. Fam Plan Perspect 1996;28:154-8

Typical Use versus Perfect Use Contraceptive failure during typical use can be measured in a clinical trial or in a survey Contraceptive failure during perfect use has been measured only in clinical trials, since retrospective reporting of adherence in surveys is likely to be terrible

What Is Typical Use? By definition, a woman is a user whenever she considers herself to be using a method Hence, typical use of a barrier method does not imply that it is actually used at every act of intercourse Typical use includes both inconsistent use and incorrect use

What Is Perfect Use? By definition, perfect use of a method requires actual use according to the directions for that method Perfect use of a barrier method requires that it be used correctly at every act of intercourse Perfect use does not imply no pregnancies

Logical Error Ingrained in Literature Suppose in a contraceptive trial there are 100 years of exposure to risk of pregnancy 15 pregnancies occur during a cycle of imperfect use 5 pregnancies occur during a cycle of perfect use What is the method-related pregnancy rate (pregnancy rate during perfect use)?

Method-Related Pregnancy Rate Traditional answer –5/100 = 5 per 100 woman-years of exposure Logical error –Denominator cannot be all exposure since by definition a method-related pregnancy can occur only during perfect use –If there are only 50 woman-years of perfect use, correct answer is 5/50 = 10 per 100 woman- years of exposure

Flaw in Design of Clinical Trials Information on perfect (correct and consistent) use is usually obtained only for cycles when pregnancy occurred Hence, pregnancy rates during perfect use cannot be estimated in most trials

Correct Analysis by Cycle Woman 1: █ █ █ █ █ █ Woman 2: █ █ █ █ █ Woman 3: █ █ █ █ Woman 4: █ █ █ Cycles: imperfect use perfect useall Rates: 1/3 1/152/18 P P P = pregnancy

Pearl Index versus Life Table Pearl index is a pregnancy rate: pregnancies per 100 woman-years of exposure –Ranges from 0 to 1300, not 0 to 100 –Rubber yardstick: women most likely to become pregnant do so early, leaving behind a pool increasing consisting of the more compliant and less fecund Life table methods produce estimates of the percent of women becoming pregnant within specific durations (e.g. 6 or 12 or 24 months) since initiating use

Pearl Index Is a Rubber Yardstick Two investigators using the same data obtained pregnancy rates of 7.5 and 4.4 during 100 woman-years of condom use One (who got 4.4) allowed each women to contribute up to 5 years of exposure whereas the other (who got 7.5) allowed each women to contribute only up to 1 year of exposure Which is correct?

Non-Completion of Trial In the ideal trial, all women would either become pregnant or complete the trial without becoming pregnant But in fact a high fraction stop for other reasons (LFU, medical reasons, personal reasons) What is the consequence?

Consequence of Non-Completion of Trial In life-table analysis, those who are censored are assumed to have the same failure rate as those who are observed, probably resulting in a downward bias In a Pearl index, more complex to assess: if those who are censored would have had a higher risk of pregnancy, observed Pearl index could be biased upward or downward

Example: Effect of Non-Completion 100 women start trial, 10 become pregnant at 6 months 50 women followed after 6 months, 1 of whom becomes pregnant at 12 months 40 women LFU at 6 months, 2 of whom would have become pregnant at 12 months Observed Pearl index = (10+1)/(50+25)=14.7 True Pearl index = (10+1+2)/( )=13.7 Observed life-table probability=11.8% True life-table probability=13%

Factors That Influence Failure Inherent efficacy of the method Imperfect use, the extent of which will depend on motivation to avoid pregnancy Frequency of intercourse, which declines with both age and marital duration Individual level of fecundity, which also declines with age Competence (honesty?) of the investigator

Common Errors Incorrect calculation of method failure Multiplying cycles by 1200 instead of 1300 to get pregnancies per 100 woman years of exposure Not including “learning” phase Discontinuing non-adherent women

Common Problems High percent not completing trial Underreporting of abortion Use of Pearl index when comparing risk of failure among methods

Problems in Comparing Methods Results come from different sources; results where available for typical use come from the NSFG, adjusted for underreporting of abortion Women choose which method to use and are not randomly assigned to methods. Women who choose to use spermicides are very different from those who choose to use IUDs

Results from Literature Table 27-1 from the 2007 edition of Contraceptive Technology Estimates of the annual risk of contraceptive failure during perfect and typical use of no method, spermicides, withdrawal, fertility awareness based methods, sponge, diaphragm, male and female condom, pill, patch, ring, injectable, IUD, implant, tubal ligation, and vasectomy

What Table 27-1 Shows Methods requiring adherence generally show a big difference between perfect-use and typical-use failure rates The most effective methods during typical use are those not requiring adherence The most effective methods are not those that protect against STIs

Communicating Risk of Failure Two studies have examined –How well do women understand contraceptive failure rates? –How to communicate contraceptive effectiveness? Result is a chart that appears in the new WHO Global Handbook for Family Planning Providers and the new edition of Contraceptive Technology Steiner. Obstet Gynecol 1996;88:24S-30S Steiner. Obstet Gynecol 2003;102:709-17

Comparing typical effectiveness of contraceptive methods More effective Less effective Less than 1 pregnancy per 100 women in one year About 30 pregnancies per 100 women in one year Injections: Get repeat injections on time LAM (for 6 months): Breastfeed often, day and night Pills: Take a pill each day Patch, ring: Keep in place, change on time Condoms, diaphragm, sponge: Use correctly every time you have sex Fertility-awareness based methods: Abstain or use condoms on fertile days. Newest methods (Standard Days Method and TwoDay Method) may be the easiest to use. How to make your method most effective After procedure, little or nothing to do or remember Vasectomy: Use another method for first 3 months Withdrawal, spermicide: Use correctly every time you have sex InjectablesPills LAM Male Condoms Female Condoms Diaphragm Spermicides IUD Female Sterilization Vasectomy PatchRing Fertility-Awareness Based Methods Withdrawal Implant Sponge Source WHO 2006, adapted with permission