Interpreting Evidence and Creating Clinical Guidance on Contraception

Kathryn M. Curtis, PhD

David Hubacher, PhD, MPH

The findings and conclusions in this chapter are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Clinical practice is the ultimate distillate of evidence, judgment, and experience. The safety, side effects, and benefits of treatments are established by clinical research. The clinician must determine whether results from studies are clinically relevant and useful. Incorporating research findings into clinical practice depends upon that determination. In this chapter, we provide a guide for interpreting published research and making judgments on clinical and epidemiologic studies, systematic reviews, and clinical practice guidelines.

Prior to initiating any clinical trial, regardless of design, investigators must register the study in a clinical trials database. Examples include ClinicalTrials.gov, a U.S. registry, and the EU Clinical Trials Register. These databases were created to provide information for patients, health care professionals, researchers, and the public on publicly and privately supported clinical trials. Investigators must also submit the key final results to the registry to ensure that all clinical studies, including those with negative findings, are reported. Most reputable peer-reviewed medical journals will not publish a manuscript unless trial registration was performed prior to study initiation.

The Hierarchy of Clinical Research (in Descending Order of Internal Validity)

Table 3.1 shows a general hierarchy of clinical study designs from the U.S. Preventive Services Task Force (USPSTF). Figure 3.1 provides an algorithm to classify various study designs.

Table 3.1 U.S. Preventive Services Task Force (USPSTF) Evidence Grading Scheme

Quality of Evidence
Level I	Properly powered and conducted RCT; wellconducted systematic review or meta-analysis of homogeneous RCTs
Level II-1	Well-designed controlled trial without randomization
Level II-2	Well-designed cohort or case-control analysis study
Level II-3	Multiple time series, with or without the intervention; results from uncontrolled studies that yield results of large magnitude
Level III	Opinions of respected authorities, based on clinical experience; descriptive studies or case reports; reports of expert committees
Strength of Recommendation
A	The USPSTF recommends the service. There is high certainty that the net benefit is substantial.
B	The USPSTF recommends the service. There is high certainty that the net benefit is moderate, or there is moderate certainty that the net benefit is moderate to substantial.
C	The USPSTF recommends selectively offering or providing this service to individual patients based on professional judgment and patient preferences. There is at least moderate certainty that the net benefit is small.
D	The USPSTF recommends against the service. There is moderate or high certainty that the service has no net benefit or that the harms outweigh the benefits.
I	The USPSTF concludes that the current evidence is insufficient to assess the balance of benefits and harms of the service. Evidence is lacking, of poor quality, or conflicting, and the balance of benefits and harms cannot be determined.
RCT, randomized controlled trial. Source: United States Preventive Services Task Force Procedure Manual, December 2015. https://www.uspreventiveservicestaskforce.org/Page/Name/procedure-manual; accessed June 19, 2019.

Figure 3.1 Algorithm for classification of types of clinical research. (Adapted from Grimes DA, Schulz KF, An overview of clinical research: the lay of the land, Lancet 359:57-61, 2002, Copyright 2002, with permission from Elsevier.)

Experimental Studies (Studies with the Exposure or Intervention Assigned by the Investigator)

Randomized Trials

A randomized trial is used to assign a treatment or intervention by chance and compare results. Randomized controlled trials (RCTs) are the most common type of randomized trial and involve at least two assignments: treatment(s) and control(s). Control assignments might be “standard treatment” or placebo, for example. A trial simply called a randomized trial is reserved for situations in which at least two experimental treatments are assigned (but no placebo or standard treatment). Participants theoretically have a random (unbiased) chance of being assigned to each group in the study, and the participant characteristics and unmeasurable influences should be nearly if not totally the same in each group. In crossover randomized trials, participants are randomly assigned to one treatment group and later assigned to the other group, and thus the participants serve as their own controls.

*Randomized Trials*
Advantages	Can provide evidence for causality. Prospective design permits person-time denominators and incidence calculations.
Disadvantages	Very expensive and time-consuming. Only a limited number of hypotheses can be evaluated in any one study.
Examples	A randomized controlled trial compared in a double-blind fashion the pregnancy rates in women using levonorgestrel 0.75 mg (n = 976) or ethinyl estradiol 100 mg plus levonorgestrel 0.5 mg (n = 979), with repeat doses in both groups 12 hours later, for emergency contraception.¹ In this study, the women receiving levonorgestrel constitute the treatment group, and the women receiving ethinyl estradiol/levonorgestrel (standard treatment), also known as the Yuzpe regimen, are the control group. Pregnancy occurred in 11 (1.1%) women in the levonorgestrel group compared with 31 (3.2%) in the Yuzpe regimen group. The crude relative risk of pregnancy for levonorgestrel compared with the Yuzpe regimen was 0.36 (95% CI, 0.18-0.70). Because the confidence interval (CI) does not include the null value of 1.0, the levonorgestrel group is superior. A crossover randomized trial assessed ovarian suppression during the use of combined oral contraceptives (COCs) with different durations of hormone-free interval.² Women were assigned to receive an ethinyl estradiol 20 µg plus norethindrone acetate 1 mg COC with either a 7-day hormone-free interval or a 4-day hormone-free interval. After 2 months, each group “crossed over” and used the other study drug for 2 months. Analysis of hormone levels in serum and cervical mucus suggested that the 4-day hormone-free interval regimen provided better ovarian suppression than the 7-day hormone-free interval regimen.

Nonrandomized Trials

In a nonrandomized trial, the intervention or exposure is still assigned by the investigator but in a nonrandom manner (e.g., by alternate assignment, odd/even subject numbers, day of the week). The essential feature of a trial involving comparisons (randomized or nonrandomized) is that participants are assigned to groups without clinical judgment or participant self-selection; many published reports labeled as trials are perhaps more accurately described as cohort studies, expanded case series, and/or prospective comparative studies. However, a trial can also be a single-arm study of a new drug or new indication for which clinical judgment is required to assess eligibility to participate. Without the benefit of randomization, nonrandomized trials may be subject to the same biases as observational studies.

*Nonrandomized Trials*
Advantages	May be used when randomization is not possible for ethical or logistical reasons.
Disadvantages	Subject to selection or allocation bias (a systematic difference in how patients are assigned to the study groups) and confounding (due to the potential for baseline differences between the study groups).
Example	A nonrandomized trial evaluated a computerized contraceptive decision aid for adolescent patients.³ Using alternate assignment (i.e., every other participant was assigned to the intervention or control groups), 456 adolescent females were assigned to the intervention group (standard education plus a contraceptive decision-making computer program), and 493 adolescent females were assigned to the control group (standard education only). The intervention group demonstrated increased short-term knowledge about oral contraceptives compared with the control group, but there were no differences in duration of oral contraceptive use over 1 year between groups.

Observational Studies (Nonexperimental Studies: Observation Without Intervention)

Observational studies that compare different groups (cohort, case-control, and cross-sectional) are higher-order studies than purely descriptive clinical reports (case series and case reports).

Cohort studies: In cohort studies, exposure information is collected from all subjects who are disease free, and subjects’ experiences are recorded forward from “time zero” to determine who develops disease. Cohort studies can be prospective (starting time zero in the present and moving forward through “real time”) or retrospective (starting time zero and exposures in the past and documenting subsequent disease that may have also occurred in the past).

*Cohort Studies*
Advantages	Can evaluate changes over time and, if prospective, minimize recall bias. Permits person-time denominators and incidence calculations.
Disadvantages	Can be expensive, lengthy in time, and subject to selection bias and surveillance bias making the two groups being compared unequal; not well suited for studying rare outcomes; loss to follow-up can be a major source of bias.
Examples	A prospective cohort study, the International Active Surveillance Study of Women Taking Oral Contraceptives, enrolled 85,109 women who were followed prospectively for 2 to 6 years and found no increased risk of venous thromboembolism from the use of newer drospirenone-containing COCs compared with levonorgestrel-containing COCs.⁴ A retrospective cohort study used data from the Danish national registries to identify1,626,158 women of reproductive age and looked back over 15 years for exposure to COCs and incidence of thrombotic events. Rates of thrombotic stroke and myocardial infarction were higher among COC users compared with nonusers.⁵

Case-control studies: Case-control studies select a group of individuals with a disease or condition (cases) and compare them with a carefully selected group of individuals who do not have the disease or condition (controls). The exposure history of those with disease and those with no disease is collected and compared to determine if exposure is positively or negatively associated with being a case. The control group must be a population that would naturally produce the cases.

*Case-Control Studies*
Advantages	Can be relatively quick and inexpensive because investigators don’t recruit participants and wait for the uncertainty of developing the disease/condition; good for rare outcomes.
Disadvantages	Subject to recall bias regarding past exposures and other errors. Usually relies on information recorded by others not with the intention of collecting data for research purposes and is, therefore, subject to missing or incomplete data. Must have sufficient (natural) levels of exposure for drawing possible associations with the disease or condition because the exposure is not controlled by the researchers; cannot estimate incidence rates, unless population-based.
Example	The World Health Organization (WHO) Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception identified 3,697 cases of stroke, myocardial infarction, or VTE among women of reproductive age, along with 9,997 controls (women without stroke, myocardial infarction, or VTE), and found no statistically significant increased risk of these thrombotic events based on an adjusted odds ratio (aOR) among oral progestin-only users (aOR 1.74; 95% CI, 0.76-3.99) or progestinonly injectable contraceptors (aOR 2.19; 95% CI, 0.66-7.26) compared with nonusers. Because the confidence interval includes the null value of 1.0, the treatment groups are not considered different than nonusers.⁶

Cross-sectional studies: Cross-sectional studies describe a group of individuals and assess exposure and disease simultaneously at one point in time.

*Cross-Sectional Studies*
Advantages	An appropriate method to estimate prevalence; generally quick and inexpensive. Often used for exploring associations between a condition and possible causes.
Disadvantages	Cannot establish cause-effect relationship, cannot assess changes over time.
Example	The National Survey of Family Growth is a periodic cross-sectional survey conducted in the United States since the 1960s; one of the latest iterations showed that oral contraceptives were the single most prevalent form of contraception (26% among current contraception users).⁷

Case Reports and Case Series

Case reports: A single description or handful of descriptions of anecdotal exposures/problems/conditions. Initial reports might lead to more systematic searches for similar situations; this is often described as a case series.

Case series: A study in which participants are sampled if they have the outcome of interest (a particular disease or disease-related outcome); may or may not have information on exposure

Both case reports and series serve to draw attention to problems of unknown cause and often lay the groundwork for hypotheses that rule out the role of chance/coincidental occurrence. Traditionally, case reports and case series were generated from personal clinical practice, but today, the source of reports can be external, like surveying a group on a listserv for similar cases and reporting as a case series.

Case reports and case series do not have a comparison group and do not attempt to estimate the risks of disease from an exposure or estimate incidence or absolute risk; in this way, they are different from cohort studies and case-control studies.

*Case Reports and Case Series*
Advantages	Easy to assemble information and requires little study planning.
Disadvantages	Descriptive only: does not take a scientific approach to concluding anything about cause Does not allow for calculations of absolute risk.
Examples	A case report described migration of a transcervical female sterilization microinsert, including details of the placement procedure, detection of migration of the microinsert, and resolution of the problem, along with an accompanying literature review on the topic.⁸ A case series of 218 pregnancies among etonogestrel implant users is described; 45 women had insufficient data to assess why pregnancy occurred, 46 were determined to have been pregnant at the time of implant insertion, 84 had unrecognized noninsertion, 30 failed for other reasons, and 13 were deemed to be product failures.⁹

Bias: Distortions Due to Study Design and Other Factors

The concerns over possible bias generally increase the more an investigation strays from standard experimental methodology. Identifying potential sources of bias and then attempting to mitigate their impact in the ascertainment of data (both exposure and disease) are essential aims of epidemiologic research. In addition, a large body of sophisticated statistical analysis and techniques have been developed to further improve the validity of research results.¹⁰

Detection/Surveillance/Diagnostic Bias: Systematic errors in methods of ascertainment, diagnosis, or verification of disease status. Can occur when not everyone in the study population has equal access to or utilization of medical interventions and diagnostic tests.

Publication Bias: Negative (null) studies and studies that confirm old results tend not to be published. An important source of bias in meta-analysis.

Reporting or Recall Bias: Inaccurate memory and selective recall introduce errors.

Selection Bias: Occurs when procedures for selecting study participants or influences enabling participation affect the internal and external validity of the measurements. Differences in characteristics between those selected for study and those not selected may be a result of preferential prescribing, family history, preferential referral of patients, or a healthy user effect.
For case-control studies, the source of the controls is important. Hospitalbased controls are less likely to be representative of the general population than population-based controls. In cohort studies, if the selection methods for the exposed and unexposed groups lead to groups that are not representative of the target population, selection bias can result.

Information or Observer Bias: A flaw in measuring exposure or outcome that produces different results between comparison groups. Nonresponse by subjects or patients lost to follow-up can produce differences in cohort studies.

Confounding: Three characteristics are necessary, but not sufficient, for a factor to be a confounder: a risk factor for the disease, associated with the exposure of interest, and not an intermediary step in the pathway between exposure and disease. Factors such as age, body weight, and smoking are often confounders in the exposure-disease relationship. In the data analysis phase, multivariable approaches and statistical tests for examining relationships between participants’ background factors and the role of such explanatory factors on disease are essential. Often, it is impossible to fully address the relationships due to errors in the measurement of confounding factors and statistical limitations (too few outcome events in relationship to the permutations of different sets of background factors). In essence, it is important to remember that statistical modeling, including multivariable analyses, cannot always make up for confounding or missing data.

The following fictitious retrospective cohort study illustrates these concepts. Returning women sophomores at unnamed state university participated in a survey to recount their freshman experiences. All women who were sexually active in their first year (n = 3,991) were included in the analysis to explore the relationship between the exposure, condom use (reported as their primary form of contraception), and the outcome, unintended pregnancy.

Crude Relative Risk Analysis
		Unintended Pregnancy?		Total	Proportion Who Experienced Unintended Pregnancy
		Yes	No
Were condoms primary contraceptive method?	Yes No	99 33	1,909 1,950	2,008 1,983	4.9% 1.7%
	Total	132	3,859	3,991

The crude relative risk is calculated as follows: 4.9%/1.7% = 3.0 (95% CI, 2.0 to 4.4). The 95% confidence interval is calculated with standard statistical analysis software. From the crude analysis, women who used condoms as their primary method of contraception were three times as likely to experience unintended pregnancy as those who did not use condoms as their primary method.

Also included in the questionnaire was the question: Did you often go to fraternity/sorority functions/parties as a freshman? As it turned out, 2,408 students did not go to such functions often, while 1,583 did. Condom use and unintended pregnancy were then re-examined separately in each stratum labeled Greek life (yes or no).

No Greek Life: Relative Risk Analysis
		Unintended Pregnancy?		Total	Proportion Who Experienced Unintended Pregnancy
		Yes	No
Were condoms primary contraceptive method?	Yes No	9 20	608 1,771	617 1,791	1.5% 1.1%
Were condoms primary contraceptive method?	Total	29	2,379	2,408
Relative risk: 1.5%/1.1% = 1.4 (95% CI, 0.6-2.8)

Yes Greek Life: Relative Risk Analysis
		Unintended Pregnancy?		Total	Proportion Who Experienced Unintended Pregnancy
		Yes	No
Were condoms primary contraceptive method?	Yes No	90 13	1,301 179	1,391 192	6.5% 6.8%
Were condoms primary contraceptive method?	Total	103	1,480	1,583
Relative risk: 6.5%/6.8% = 1.0 (95% CI, 0.5-1.7)

In each stratum (Greek life yes or no), there is no association between condom use and unintended pregnancy (the confidence intervals of both RRs include the null value of 1.0). Without consideration of Greek life as a risk factor for unintended pregnancy, one might erroneously conclude that only poor condom use was to blame. As it turned out, both condom use and unintended pregnancy were disproportionately more common among freshmen who participated in Greek life than those who did not. If the above-estimated crude relative risk of 3.0 is adjusted using the Mantel-Haenszel approach to control for the confounding role of Greek life, the risk of unintended pregnancy from condom use disappears (adjusted relative risk of 1.1; 95% CI, 0.7 to 1.7). If alcohol consumption were examined as a fourth variable, then logistic regression techniques could be used to control for Greek life and alcohol consumption simultaneously.

A Guide to Epidemiologic Terms Commonly Used

Relative Risk

The ratio of the risk of the disease or condition among those exposed and the risk among the unexposed, or the ratio of the cumulative incidence rate in the exposed and the unexposed, is also called risk ratio. In its simplest definition, relative risk compares the rate of disease in two groups, one of which has been exposed to something that is believed to either increase or decrease the risk of that disease, usually in a cohort study or randomized trial. A null value of 1.0 suggests neither an adverse effect nor a protective effect of the exposure; a value over 1.0 suggests elevated risk of the outcome, while a value under 1.0 suggests reduced risk.

Odds Ratio

The odds ratio is the measure of association calculated in case-control and cross-sectional studies. The odds ratio is a good estimate of the relative risk when the disease or condition is rare. The odds ratio does not characterize incidence or causation; it only provides an internal measure of relative risk. An important scientific mantra is “Association ≠ Causation.”

Confidence Interval

By convention, we often use 95% confidence intervals. The lower and upper bounds of the interval constitute the range of error around the point estimate for that particular measurement. In practical terms, it means that if the exact study were to be done 100 times, the point estimate would fall between those values 95 times. For an odds ratio or relative risk to be statistically significant, the confidence interval must not include the null value of 1.0. When comparing confidence intervals of a rate or a measure (such as blood pressure), if the ranges overlap, the differences are not considered statistically significant.

The tighter (narrower) the range of the confidence interval, the more precise the estimate. The wider the confidence interval (CI), the more imprecise the estimate, usually because of small numbers of study subjects.

p Value

The p value characterizes the role (probability) that chance may play in the results. By convention, the threshold used for chance to be ruled out is a p value below 0.05 (significant). A p value of 0.05 means that there is a 5% probability that the result occurred by chance. The lower the p value, the more likely the result is real. (Note: a leading zero is a zero to the left of the decimal point used for clarity, but some journals do not allow them.)

Attributable Risk

Attributable risk is the difference in actual incidence of disease or condition between exposed and unexposed groups; it provides a realistic estimate of the change in incidence that is attributable to the exposure. A modest increase in relative risk will produce only a small number of cases when clinical events are rare, such as venous thromboembolism (VTE) and arterial thrombosis in young women. If the absolute risk is very low, a statistically significant increase in relative risk may mean little or nothing in practical, real numbers. For example, COC use is associated with a threefold increase in relative risk of VTE compared with nonuse.¹¹

However, the baseline risk of VTE is relatively rare in women of reproductive age—about 5 cases per 10,000 women per year.¹¹ Therefore, COC users have an absolute risk of about 15 cases of VTE per 10,000 women per year, 10 of which are attributable to COC use—still a rare event.

Number Needed to Treat

The number needed to treat is the number of individuals who must be treated, usually over a 1-year time period, to produce one instance of either a positive or a negative effect.

Statistical Concepts in Hypothesis Testing

Statistical tests provide an objective methodology for evaluating whether the quantitative results obtained from an experiment or observation represent a true deviation from chance. While a full discussion of statistical tests and interpretation of results is beyond the scope of this chapter, some basic principles warrant review. First, the type of measures for endpoint analyses (e.g., means, medians, proportions, etc.) defines the statistical test used to evaluate the study hypothesis. The second major principle is that all statistical tests reflect probabilities. The degree to which we accept a result as statistically significant is somewhat arbitrary. By convention, we typically regard an outcome as statistically significant if the result has less than a 5% chance of being observed, if the null hypothesis is true.

When designing a clinical study, the investigator must consider the anticipated magnitude of the difference in the outcome between the study groups (e.g., effect size), the natural variability in the outcome in the study population (e.g., variance), and choose a level of statistical significance for evaluation of the result.

A power analysis estimates a target sample size based upon a predefined effect size and variance for the appropriate statistical test. While studies frequently report primary and secondary outcomes, the power analysis generally reflects the sample size requirement for the primary outcome only. To ensure that the study remains adequately powered to evaluate the primary outcome, investigators typically will increase the number of participants enrolled above the minimum required by the sample size calculation to account for the anticipated dropout rate.

A power analysis must consider concepts of Type 1 and Type 2 errors.

Type 1 error (also known as α) refers to the probability of falsely concluding that a difference exists (a false-positive result). In most studies, investigators will select an α of 0.05 for evaluation of the primary outcome. In practical terms, this represents a 1 in 20 chance of a Type 1 error.
Type 2 error (also known as β) refers to the probability of falsely accepting the null hypothesis of equality, when in fact a true difference exists (a false-negative result). Since scientists tend to be cautious, we consider the implications of a Type 2 error as less worrisome than Type 1 error. For this reason, we typically select a β of 0.2 when considering sample size. The probability of avoiding a Type 2 error is called power (power = 1 − β). A study with a β of 0.2 has 80% power (e.g., 4 in 5 chance) to exclude a Type 2 error. The most common cause of a Type 2 error is a sample size being too small.

Clinicians interpreting the literature must also consider the difference between clinical significance and statistical significance. Clinical significance refers to the magnitude of difference that would reflect a clinically important result and is not influenced by sample size. In contrast, increasing sample size can result in highly statistically significant results with very small effect sizes of dubious clinical relevance. Very large database studies published in major journals frequently report highly statistically significant risk estimates of clinically unimportant effect size. Ideally, to guide clinical practice, a result should be both clinically and statistically significant.

Putting Risk into Perspective—Interpreting Findings from Epidemiologic Studies

Epidemiology is the science of detecting and understanding disease patterns in large populations. Epidemiologic studies alone do not prove causation; they identify associations between diseases and certain factors that are the first step in assessing causality.

A relative risk in the range of 1.0 to 2.0 represents an increased risk but a weak association. As explained above, confidence intervals around the relative risk help determine statistical significance of the risk. A relative risk (from a randomized trial of cohort study) of 1.5 and an odds ratio (from a case-control study) of 1.5 are mathematically equivalent and signal the same levels of risk; the hierarchy of study designs and validity attached to the estimates then differentiates the interpretation.

The clinical significance of an increase in risk is influenced by the rate of the disease in the general (unexposed) population. If the rate of the disease in the unexposed population is 10% and the relative risk is 1.4, then the risk of disease in the exposed population is 14%. If the rate of disease in the unexposed population is only 1%, then the same relative risk of 1.4 increases the actual disease risk to only 1.4%.

Criteria that strengthen the conclusion that an epidemiologic finding or association represents a true cause-and-effect relationship include the following¹²:

1. The strength of the association (the larger the relative risk, the more likely it is real).
2. Consistency, uniformity, and agreement among many studies.
3. A dose-response relationship (either with dose of a drug or an increasing effect with increasing duration of exposure).
4. Biologic plausibility of the finding (known mechanisms by which exposure could cause or influence disease).
5. An appropriate temporal relationship (the amount of time between exposure and development of disease is appropriate according to the pathogenesis of the disease).
6. Coherence of the finding with what is known about the natural history, biology, and other characteristics of the exposure and outcome relationship.

The results section of a published study objectively provides epidemiologic findings that can be used to advance science and our understanding of possible cause-and-effect relationships. Researchers typically provide outcome measures and statistical tests; together this information is used to support the conclusions of the study report. Statisticians often transform simple outcome measures such as number of incident cases, percentages, and means, to composite measures with built-in comparisons; additionally, the measures usually have a mathematical test to determine whether any differences can be considered outside the realm of chance.

Example of Interpreting Findings

A published cohort study from 2017 reported that women who currently or recently used any hormonal contraceptive had a higher risk of breast cancer than women who had never used hormonal contraceptives (relative risk, 1.20; 95% CI, 1.14 to 1.26).¹³

The relative risk of 1.20 means a 20% increased chance of developing breast cancer among women who currently or recently used any hormonal contraceptive compared with women who had never used any hormonal contraceptive. The 95% confidence interval demonstrates that if one were to conduct this exact study 100 different times, the estimated relative risk would fall between 1.14 and 1.26 in 95 of those completed studies.

The components and calculation of the relative risk are the estimated incidence rate of breast cancer among users of any hormonal contraceptive (68 cases per 100,000 per year) divided by the rate among “never users of hormonal contraception” (55 cases per 100,000 per year).

Further, the authors adjusted the relative risks by seven other factors that, if not considered, might have biased the summary measure through confounding. An important limitation for this adjustment is that a significant amount of information about important confounding factors was unavailable.

Finally, the published risk difference was estimated at 13 additional breast cancer cases per 100,000 women per year among women who currently or recently used hormonal contraception (95% range of 10-16). This converts to “number needed to treat” in the following way: to produce one instance of breast cancer, 7,690 women will need to use a hormonal contraceptive for 1 year (100,000 divided by 13 = 7,690). The estimated range of number needed to treat is 6,250 to 10,000. Therefore, while this study observed a small increase in relative risk of breast cancer associated with hormonal contraception use that is statistically significant, the absolute risk of breast cancer associated with hormonal contraceptive use is quite low (13 cases per 100,000 users per year).

Synthesizing the Evidence

While it is important for clinicians to understand how to interpret individual studies, results from a single study are generally insufficient to determine a course of clinical action or to develop an evidence-based recommendation. A systematic review synthesizes the body of evidence on a specific question and uses explicit methods to identify, critically appraise, and summarize the findings across studies, providing an overall assessment of the evidence along with a discussion of potential biases.¹⁴^,¹⁵ Systematic reviews may include meta-analysis, in which the results of individual studies are statistically
combined to provide a more precise measure of effect. Looking across the totality of the evidence on a particular topic provides the clinician with a distilled review of the consistency of findings and the strength of the evidence on which to base a clinical decision. In addition to systematic reviews and meta-analyses, there are a number of other types of reviews that readers and researchers may find useful; Grant et al.¹⁶ describe 14 review types including scoping reviews (preliminary assessment of the scope of the literature on a particular topic), rapid reviews (completeness may be determined by time constraints), and state-of-the-art reviews that address current issues.

The Cochrane Database of Systematic Reviews is a primary source for systematic reviews of the evidence in health care and contains several reviews on contraception and other reproductive health care topics (https://fertility-regulation.cochrane.org). Reviews are developed through a standardized methodology, peer-reviewed, and updated with new evidence on a regular basis. Other sources include systematic reviews published in the literature and can be found through searching databases such as MEDLINE®, from the U.S. National Library of Medicine.

Reporting the Evidence

The final step for investigators of individual studies and systematic reviews/meta-analyses is to clearly communicate the findings. This is most often done through publication in the peer-reviewed literature but also in the form of reports, oral presentations, and communication with scientific and lay media. It is important to “tell the story” of the research, including a precise statement of the research question, a detailed accounting of the research methods, a clear and organized summary of the results, an objective reflection on the quality of the evidence including strengths and limitations, and implications for practice and future research. Many journals encourage or require that authors follow reporting guidelines based on study type, including guidelines for randomized trials (CONSORT guidelines), observational studies (STROBE), and systematic reviews and meta-analyses (PRISMA).¹⁷^,¹⁸^,¹⁹ These guidelines often have extensions specific to a particular type of intervention or medical condition. The EQUATOR Network (http://www.equatornetwork.org) provides links to many different types of reporting guidelines and extensions, as well as tips and training for publishing high-quality articles.

Translating Evidence into Clinical Guidance

Systematic reviews still need to be translated into specific recommendations for clinical action. Clinical practice guidelines are tools that provide recommendations for clinical care based on the sum and strength of the evidence, as well as other factors such as benefits and harms, patient preferences, feasibility, and cost. The Institute of Medicine (IOM) defines clinical practice guidelines as “recommendations, intended to optimize patient care, that are informed by a systematic review of evidence and an assessment of
the benefits and harms of alternative care options,” and defines several characteristics of trustworthy guidelines (Table 3.2).¹⁴ Table 3.1 provides the USPSTF grading system for strength of recommendations. Clinical practice guidelines need to be implemented in the context of improving quality of care, along with development of tools and job aids for providers that may include electronic clinical decision support tools. Sources of guidelines include federal agencies, professional organizations, and other sponsoring organizations. Examples of clinical practice guidelines related to family planning and reproductive health are included in Table 3.3.

Table 3.2 Trustworthy Clinical Practice Guidelines

Clinical practice guidelines are statements that include recommendations intended to optimize patient care that are informed by a systematic review of evidence and an assessment of the benefits and harms of alternative care options. Trustworthy clinical practice guidelines should:

Be based on a systematic review of the existing evidence
Be developed by a knowledgeable, multidisciplinary panel of experts and representatives from key affected groups
Consider important patient subgroups and patient preferences, as appropriate
Be based on an explicit and transparent process that minimizes distortions, biases, and conflicts of interest
Provide a clear explanation of the logical relationships between alternative care options and health outcomes and provide ratings of both the quality of evidence and the strength of the recommendations
Be reconsidered and revised as appropriate when important new evidence warrants modifications of recommendations

Source: Institute of Medicine, Clinical Practice Guidelines We Can Trust, The National Academies Press, Washington, DC, 2011.

Medical Eligibility Criteria for Contraceptive Use and Selected Practice Recommendations for Contraceptive Use

Two key examples of clinical practice guidelines for contraception are the Medical Eligibility Criteria for Contraceptive Use (MEC) and the Selected Practice Recommendations for Contraceptive Use (SPR), produced by the World Health Organization (WHO) for a global audience.²⁰^,²¹ The MEC provides recommendations for safe use of specific contraceptive methods by women with medical conditions and other characteristics, such as age and postpartum status (Table 3.4). Recommendations are provided using a

numeric classification from 1 (safe to use) to 4 (unacceptable health risks) (Table 3.5). The SPR provides recommendations for contraceptive management issues, such as when a woman can start a specific method, what exams and tests (if any) are needed and how to manage bleeding problems and other issues with contraceptive use. Use of these evidence-based guidelines can help assure that women are not exposed to inappropriate risks and are not unnecessarily denied access to contraceptive methods and services when choosing and using contraception.

Table 3.3 Examples of Clinical Practice Guidelines for Family Planning and Reproductive Health

Guideline	Organization	Link
U.S. Medical Eligibility Criteria for Contraceptive Use, 2016	Centers for Disease Control and Prevention	https://www.cdc.gov/reproductivehealth/contraception/mmwr/mec/summary.html
U.S. Selected Practice Recommendations for Contraceptive Use, 2016	Centers for Disease Control and Prevention	https://www.cdc.gov/reproductivehealth/contraception/mmwr/spr/summary.html
Providing Quality Family Planning Services	Centers for Disease Control and Prevention and Office of Population Affairs	https://www.cdc.gov/reproductivehealth/contraception/qfp.htm
Clinical Guidelines on Contraception and Abortion	Society of Family Planning	https://www.societyfp.org/Resources/Clinical-guidelines.aspx
Sexually Transmitted Disease Treatment Guidelines	Centers for Disease Control and Prevention	https://www.cdc.gov/std/tg2015/default.htm
Long-Acting Reversible Contraception: Implants and Intrauterine Devices	American College of Obstetricians and Gynecologists	https://www.acog.org/Clinical-Guidance-and-Publications/Practice-Bulletins/Committee-on-Practice-Bulletins-Gynecology/Long-Acting-Reversible-Contraception-Implants-and-Intrauterine-Devices
Contraception for Adolescents	American Academy of Pediatrics	http://pediatrics.aappublications.org/content/134/4/e1257

Table 3.4 Summary Chart for U.S. Medical Eligibility Criteria for Contraceptive Use

Condition	Sub-Condition		Cu-IUD		LNG-IUD		Implant		DMPA		POP		CHC
			I	C	I	C	I	C	I	C	I	C	I	C
Age			Menarche to <20 yrs:2		Menarche to <20 yrs:2		Menarche to <18 yrs:1		Menarche to <18 yrs:2		Menarche to <18 yrs:1		Menarche to <40 yrs:1
			≥20 yrs:1		≥20 yrs:1		18-45 yrs:1		18-45 yrs:1		18-45 yrs:1		≥40 yrs:2
							>45 yrs:1		>45 yrs:2		>45 yrs:1
Anatomical abnormalities	a) Distorted uterine cavity		4		4
	b) Other abnormalities		2		2
Anemias	a) Thalassemia		2		1		1		1		1		1
	b) Sickle cell disease^‡		2		1		1		1		1		2
	c) Iron-deficiency anemia		2		1		1		1		1		1
Benign ovarian tumors	(including cysts)		1		1		1		1		1		1
Breast disease	a) Undiagnosed mass		1		2		2^*		2^*		2^*		2^*
	b) Benign breast disease		1		1		1		1		1		1
	c) Family history of cancer		1		1		1		1		1		1
	d) Breast cancer^‡
		i) Current	1		4		4		4		4		4
		ii) Past and no evidence of current disease for 5 years	1		3		3		3		3		3
Breastfeeding	a) <21 days postpartum						2^*		2^*		2^*		4^*
	b) 21 to <30 days postpartum
		i) With other risk factors for VTE					2^*		2^*		2^*		3^*
		ii) Without other risk factors for VTE					2^*		2^*		2^*		3^*
	c) 30-42 days postpartum
		i) With other risk factors for VTE					1^*		1^*		1^*		3^*
		ii) Without other risk factors for VTE					1^*		1^*		1^*		2^*
	d) >42 days postpartum						1^*		1^*		1^*		2^*
Cervical cancer	Awaiting treatment		4	2	4	2	2		2		1		2
Cervical ectropion			1		1		1		1		1		1
Cervical intraepithelial neoplasia			1		2		2		2		1		2
Cirrhosis	a) Mild (compensated)		1		1		1		1		1		1
	b) Severe^‡ (decompensated)		1		3		3		3		3		4
Cystic fibrosis^‡			1^*		1^*		1^*		2^*		1^*		1^*
Deep venous thrombosis (DVT)/Pulmonary embolism (PE)	a) History of DVT/PE, not receiving anticoagulant therapy
		i) Higher risk for recurrent DVT/PE	1		2		2		2		2		4
		ii) Lower risk for recurrent DVT/PE	1		2		2		2		2		3
	b) Acute DVT/PE		2		2		2		2		2		4
	c) DVT/PE and established anticoagulant therapy for at least 3 months
		i) Higher risk for recurrent DVT/PE	2		2		2		2		2		4^*
		ii) Lower risk for recurrent DVT/PE	2		2		2		2		2		3^*
	d) Family history (first-degree relatives)		1		1		1		1		1		2
	e) Major surgery
		i) With prolonged immobilization	1		2		2		2		2		4
		ii) Without prolonged immobilization	1		1		1		1		1		2
	f) Minor surgery without immobilization		1		1		1		1		1		1
Depressive disorders			1^*		1^*		1^*		1^*		1^*		1^*
Key:
1 No restriction (method can be used)			3 Theoretical or proven risks usually outweigh the advantages
2 Advantages generally outweigh theoretical or proven risks			4 Unacceptable health risk (method not to be used)

Condition	Sub-Condition		Cu-IUD		LNG-IUD		Implant		DMPA		POP		CHC
			I	C	I	C	I	C	I	C	I	C	I	C
Diabetes	a) History of gestational disease		1		1		1		1		1		1
	b) Nonvascular disease
		i) Non-insulin dependent	1		2		2		2		2		2
		ii) Insulin dependent	1		2		2		2		2		2
	c) Nephropathy/retinopathy/neuropathy^‡		1		2		2		3		2		3/4^*
	d) Other vascular disease or diabetes of >20 years’ duration^‡		1		2		2		3		2		3/4^*
Dysmenorrhea	Severe		2		1		1		1		1		1
Endometrial cancer^‡			4	2	4	2	1		1		1		1
Endometrial hyperplasia			1		1		1		1		1		1
Endometriosis			2		1		1		1		1		1
Epilepsy^‡	(see also Drug Interactions)		1		1		1^*		1^*		1^*		1^*
Gallbladder disease	a) Symptomatic
		i) Treated by cholecystectomy	1		2		2		2		2		2
		ii) Medically treated	1		2		2		2		2		3
		iii) Current	1		2		2		2		2		3
	b) Asymptomatic		1		2		2		2		2		2
Gestational trophoblastic disease^‡	a) Suspected GTD (immediate postevacuation)
		i) Uterine size first trimester	1^*		1^*		1^*		1^*		1^*		1^*
		ii) Uterine size second trimester	2^*		2^*		1^*		1^*		1^*		1^*
	b) Confirmed GTD
		i) Undetectable/non-pregnant β-hCG levels	1^*	1^*	1^*	1^*	1^*		1^*		1^*		1^*
		ii) Decreasing β-hCG levels	2^*	1^*	2^*	1^*	1^*		1^*		1^*		1^*
		iii) Persistently elevated β-hCG levels or malignant disease, with no evidence or suspicion of intrauterine disease	2^*	1^*	2^*	1^*	1^*		1^*		1^*		1^*
		iv) Persistently elevated β-hCG levels or malignant disease, with evidence or suspicion of intrauterine disease	4^*	2^*	4^*	2^*	1^*		1^*		1^*		1^*
Headaches	a) Nonmigraine (mild or severe)		1		1		1		1		1		1^*
	b) Migraine
		i) Without aura (includes menstrual migraine)	1		1		1		1		1		2^*
		ii) With aura	1		1		1		1		1		4^*
History of bariatric surgery^‡	a) Restrictive procedures		1		1		1		1		1		1
	b) Malabsorptive procedures		1		1		1		1		3		COCs: 3
													P/R: 1
History of cholestasis	a) Pregnancy related		1		1		1		1		1		2
	b) Past COC related		1		2		2		2		2		3
History of high blood pressure during pregnancy			1		1		1		1		1		2
History of Pelvic surgery			1		1		1		1		1		1
HIV	a) High risk for HIV		2	2	2	2	1		2^*		1		1
	b) HIV infection						1^*		1^*		1^*		1^*
		i) Clinically well receiving ARV therapy	1	1	1	1	If on treatment, see Drug Interactions
		ii) Not clinically well or not receiving ARV therapy^‡	2	1	2	1	If on treatment, see Drug Interactions
Abbreviations: C=continuation of contraceptive method; CHC=combined hormonal contraception (pill, patch, and, ring); COC=combined oral contraceptive; Cu-IUD=copper-containing intrauterine device; DMPA = depot medroxyprogesterone acetate; I=initiation of contraceptive method; LNG-IUD=levonorgestrel-releasing intrauterine device; NA=not applicable; POP=progestin-only pill; P/R=patch/ring ‡ Condition that exposes a woman to increased risk as a result of pregnancy. *Please see the complete guidance for a clarification to this classification: www.cdc.gov/reproductivehealth/unintendedpregnancy/USMEC.htm.
Condition	Sub-Condition		Cu-IUD		LNG-IUD		Implant		DMPA		POP		CHC
			I	C	I	C	I	C	I	C	I	C	I	C
Hypertension	a) Adequately controlled hypertension		1^*		1^*		1^*		2^*		1^*		3^*
	b) Elevated blood pressure levels (properly taken measurements)
		i) Systolic 140-159 or diastolic 90-99	1^*		1^*		1^*		2^*		1^*		3^*
		ii) Systolic ≥160 or diastolic ≥100^‡	1^*		2^*		2^*		3^*		2^*		4^*
	c) Vascular disease		1^*		2^*		2^*		3^*		2^*		4^*
Inflammatory bowel disease	(Ulcerative colitis, Crohn’s disease)		1		1		1		2		2		2/3^*
Ischemic heart disease^‡	Current and history of		1		2	3	2	3	3		2	3	4
Known thrombogenic mutations^‡			1^*		2^*		2^*		2^*		2^*		4^*
Liver tumors	a) Benign
		i) Focal nodular hyperplasia	1		2		2		2		2		2
		ii) Hepatocellular adenoma^‡	1		3		3		3		3		4
	b) Malignant^‡ (hepatoma)		1		3		3		3		3		4
Malaria			1		1		1		1		1		1
Multiple risk factors for atherosclerotic cardiovascular disease	(e.g., older age, smoking, diabetes, hypertension, low HDL, high LDL, or high triglyceride levels)		1		2		2^*		3^*		2^*		3/4^*
Multiple sclerosis	a) With prolonged immobility		1		1		1		2		1		3
	b) Without prolonged immobility		1		1		1		2		1		1
Obesity	a) Body mass index (BMI) ≥30 kg/m²		1		1		1		1		1		2
	b) Menarche to <18 years and BMI ≥ 30 kg/m²		1		1		1		2		1		2
Ovarian cancer^‡			1		1		1		1		1		1
Parity	a) Nulliparous		2		2		1		1		1		1
	b) Parous		1		1		1		1		1		1
Past ectopic pregnancy			1		1		1		1		2		1
Pelvic inflammatory disease	a) Past
		i) With subsequent pregnancy	1	1	1	1	1		1		1		1
		ii) Without subsequent pregnancy	2	2	2	2	1		1		1		1
	b) Current		4	2^*	4	2^*	1		1		1		1
Peripartum cardiomyopathy^‡	a) Normal or mildly impaired cardiac function
		i) <6 months	2		2		1		1		1		4
		ii)>6 months	2		2		1		1		1		3
	b) Moderately or severely impaired cardiac function		2		2		2		2		2		4
Postabortion	a) First trimester		1^*		1^*		1^*		1^*		1^*		1^*
	b) Second trimester		2^*		2^*		1^*		1^*		1^*		1^*
	c) Immediate postseptic abortion		4		4		1^*		1^*		1^*		1^*
Postpartum (nonbreastfeeding women)	a) <21 days						1		1		1		4
	b) 21 days to 42 days
		i) With other risk factors for VTE					1		1		1		3^*
		ii) Without other risk factors for VTE					1		1		1		2
	c) >42 days						1		1		1		1
Postpartum (in breastfeeding or nonbreastfeeding women, including cesarean delivery)	a) <10 minutes after delivery of the placenta
		i) Breastfeeding	1^*		2^*
		ii) Nonbreastfeeding	1^*		1^*
	b) 10 minutes after delivery of the placenta to <4 weeks		2^*		2^*
	c) ≥4 weeks		1^*		1^*
	d) Postpartum sepsis		4		4

Condition	Sub-Condition		Cu-IUD		LNG-IUD		Implant		DMPA		POP		CHC
			I	C	I	C	I	C	I	C	I	C	I	C
Pregnancy			4^*		4^*		NA^*		NA^*		NA^*		NA^*
Rheumatoid arthritis	a) On immunosuppressive therapy		2	1	2	1	1		2/3^*		1		2
	b) Not on immunosuppressive therapy		1		1		1		2		1		2
Schistosomiasis	a) Uncomplicated		1		1		1		1		1		1
	b) Fibrosis of the liver^‡		1		1		1		1		1		1
Sexually transmitted diseases (STDs)	a) Current purulent cervicitis or chlamydial infection or gonococcal infection		4	2^*	4	2^*	1		1		1		1
	b) Vaginitis (including trichomonas vaginalis and bacterial vaginosis)		2	2	2	2	1		1		1		1
	c) Other factors relating to STDs		2^*	2	2^*	2	1		1		1		1
Smoking	a) Age <35		1		1		1		1		1		2
	b) Age ≥35, <15 cigarettes/day		1		1		1		1		1		3
	c) Age ≥35, ≥15 cigarettes/day		1		1		1		1		1		4
Solid organ transplantation^‡	a) Complicated		3	2	3	2	2		2		2		4
	b) Uncomplicated		2		2		2		2		2		2^*
Stroke^‡	History of cerebrovascular accident		1		2		2	3	3		2	3	4
Superficial venous disorders	a) Varicose veins		1		1		1		1		1		1
	b) Superficial venous thrombosis (acute or history)		1		1		1		1		1		3^*
Systemic lupus erythematosus^‡	a) Positive (or unknown) antiphospholipid antibodies		1^*	1^*	3^*		3^*		3^*	3^*	3^*		4^*
	b) Severe thrombocytopenia		3^*	2^*	2^*		2^*		3^*	2^*	2^*		2^*
	c) Immunosuppressive therapy		2^*	1^*	2^*		2^*		2^*	2^*	2^*		2^*
	d) None of the above		1^*	1^*	2^*		2^*		2^*	2^*	2^*		2^*
Thyroid disorders	Simple goiter/hyperthyroid/hypothyroid		1		1		1		1		1		1
Tuberculosis^‡ (see also Drug Interactions)	a) Nonpelvic		1	1	1	1	1^*		1^*		1^*		1^*
	b) Pelvic		4	3	4	3	1^*		1^*		1^*		1^*
Unexplained vaginal bleeding	(suspicious for serious condition) before evaluation		4^*	2^*	4^*	2^*	3^*		3^*		2^*		2^*
Uterine fibroids			2		2		1		1		1		1
Valvular heart disease	a) Uncomplicated		1		1		1		1		1		2
	b) Complicated^‡		1		1		1		1		1		4
Vaginal bleeding patterns	a) Irregular pattern without heavy bleeding		1		1	1	2		2		2		1
	b) Heavy or prolonged bleeding		2^*		1^*	2^*	2^*		2^*		2^*		1^*
Viral hepatitis	a) Acute or flare		1		1		1		1		1		3/4^*	2
	b) Carrier/Chronic		1		1		1		1		1		1	1
Drug Interactions
Antiretroviral therapy All other ARV’s are 1 or 2 for all methods.	Fosamprenavir (FPV)		1/2^*	1^*	1/2^*	1^*	2^*		2^*		2^*		3^*
Anticonvulsant therapy	a) Certain anticonvulsants (phenytoin, carbamazepine, barbiturates, primidone, topiramate, oxcarbazepine)		1		1		2^*		1^*		3^*		3^*
	b) Lamotrigine		1		1		1		1		1		3^*
Antimicrobial therapy	a) Broad spectrum antibiotics		1		1		1		1		1		1
	b) Antifungals		1		1		1		1		1		1
	c) Antiparasitics		1		1		1		1		1		1
	d) Rifampin or rifabutin therapy		1		1		2^*		1^*		3^*		3^*
SSRIs			1		1		1		1		1		1
St. John’s wort			1		1		2		1		2		2
Updated in 2017. This summary sheet only contains a subset of the recommendations from the U.S. MEC. For complete guidance, see: http://www.cdc.gov/reproductivehealth/unintendedpregnancy/USMEC.htm. Most contraceptive methods do not protect against sexually transmitted diseases (STDs). Consistent and correct use of the male latex condom reduces the risk of STDs and HIV.														CS266008-A
Reprinted from https://www.cdc.gov/reproductivehealth/contraception/pdf/summary-chart-us-medical-eligibility-criteria_508tagged.pdf Use of this material does not imply an endorsement by the Centers for Disease Control and Prevention (CDC) or Health and Human Services (HHS) of any particular organization, service, or product.