Evidence-based medicine (EBM) is considered the gold standard of optimal care by healthcare professionals worldwide, with significant and conscious focus at academic centers. Although its philosophy dated back to the 1800s, the term was coined in 1990 by Dr. Gordon Guyatt to describe his new residency program curriculum based on the teaching and practices of his mentor, Dr. David Sackett. This term later was introduced widely to academic medicine cultures in a 1991 American College of Physicians (ACP) Journal Club editorial that launches Dr. David Sackett’s title as the father of EBM.^1,2

Sackett defines EBM as “the integration of best research evidence with clinical expertise and patient values” including “patient’s preferences, concerns and expectations.”³ It is the intentional and rigorous effort to make clinical decisions based on the highest level of evidence acquired from the most exhaustive search, rather than relying on personal experience alone. Its process encourages practice changes as more evidence emerges, so that clinical decisions would continually evolve. EBM has changed the standard of practice of numerous medical therapies over the decades. One well-known example is the discontinuation of diethylstilbestrol (DES) during pregnancy as evidence of harm accumulated over time.

In formulating a research question, key parameters are Patient/population and Problem, Intervention, Comparison, and Outcome (PICO)⁴ (see Table 10-1). This guideline helps the conscientious clinician target the literature search specific to the patient currently on the exam table.

An example of such a question would be: In an African-American teenage multiparous woman at 34 weeks and 3 days estimated gestational age who presents with confirmed preterm premature rupture of membranes (PPROM), is expectant management or tocolytic medications associated with a lower risk of respiratory issues and a shorter hospital stay for the neonate?

Once a clear research question is formulated, the clinician can start with primary research data on the topic via the PubMed, CINAHL, EMBASE, Web of Science, and PsychINFO search engines. Most initial PubMed searches would return hundreds to thousands of articles. It may help to narrow such searches by applying filters. For example, a simple keyword search on “misoprostol induction labor” yields 808 articles. Apply more filters in the advanced search section such as “clinical trials” reduces to 333 entries, and to 271 trials in “adult: 19–44 years,” and to 92 that are published in “core clinical journals” and 21 that have been published in the last “10 years.”

For topics that have been well researched, preappraised reviews of all available evidence on this topic may be available on databases such as the Cochrane Library, National Guidelines Clearinghouse, UpToDate, ACP Journal Club, Clinical Evidence, NHS Clinical Knowledge Summaries, Dynamed, and Physicians’ Information and Education Resource (PIER).⁵ Professional organizations such as the American Congress of Obstetricians and Gynecologists (ACOG), American Pediatrics Association (APA), American Heart Associations (AHA), and US Preventive Services Task Force (USPSTF) also provide expert reviews of evidence-based medicine with added expert opinion on questions for which research data is limited.

Once a narrow number of published, peer-reviewed primary and secondary reviews have been identified, the next task is to assess the strength of the evidence and interrogate its integrity in terms of the ability to provide the fairest and most truthful evidence. This information can be identified through the research designs, data collection methods, and analysis techniques that minimize bias and confounders. Some red flags include secondary gain by the authors (business affiliations) to show positive outcomes, poor follow-up (i.e. >20%, because participants with poor outcomes might have been unable to return to clinic for follow-up), selection bias (illiteracy and tremor vs. written research forms), randomization flaw (e.g. all odd numbers, so participants can figure it out), and differences in the baseline characteristics between experimental and control groups.

Research designs can differ in various aspects, including observation vs. intervention, retrospective review vs. cross-sectional assessment vs. prospective evaluation, single vs. double vs. multiple-arms, and study size. See Figure 10-1 to see the hierarchy of evidence.

The lowest level of evidence starts with personal clinical experience and anecdotes, expert opinions through expert committee reports, and deductions from mechanistic reasoning. Next are case series or observational studies. Uncontrolled studies at least include an experimental intervention, but they still are less strong than case-control studies, which also compare with a nonintervention. Stronger studies are nonrandomized trials with some historical or contemporary controls. However, the strongest study design is the large randomized controlled trial (RCT), which is double blind, and in which results with statistical significance present the best single-study evidence. On that same principle, a systematic review of many RCTs that yield similar results would provide the most clear-cut evidence, which is the most trustworthy.

OBSERVATIONAL RESEARCH DESIGNS

Observational studies serve the main purpose of studying epidemiology (e.g. prevalence of disease), associations and causations of disease, and prognoses.

Case Reports

A case report simply describes a typically rare condition that is not already well known, or a novice intervention or lesser-known adverse event. This is a formalized and published form of clinical anecdote shared among colleagues.

Case Series

Case series describe the characteristics of a group of patients with a similar disease outcome, with or without specific exposure.

Cross-sectional Studies

Cross-sectional studies capture a single point in time and usually report statistics such as prevalence of disease and prevalence of a particular exposure. It can report averages, such as “normal cervix size in nulliparous women,” or other descriptive statistics. It can involve a survey that asks about a diagnosis, any history of a prior factor such as smoking, or a particular medication. Cross-sectional studies can establish correlations of risk factors and diseases, but not causation of said diseases.

Case-Control Studies

Case-control studies match patients with a condition (case) with those without (control) and retrospectively trace the cases to assess the proportion of each group that is exposed to a particular possible cause. It can be conducted as a survey or through a review of a medical record chart.

MIXED OBSERVATIONAL AND EXPERIMENTAL RESEARCH DESIGNS

Some study designs are applicable to both observational and experimental intention depending on the research question. Experimental designs help most with therapy efficacy, safety, and dosing.

Case Report

As described previously, a case report can detail a newly discovered surprise treatment or innovative surgical technique.

Cohort Study

A cohort study is a prospective observational study of a large group of patients with similar exposure characteristics who are followed over time (years to decades) to collect possible outcomes. The exposure is being studied as a possible causative agent to some outcome characteristics being studied. This exposure can be freely occurring whether the study exists or not (such as divers who are exposed to deep water pressure) or a result of an intentional experimental intervention (such as an experimental or newly approved drug). The patients are divided into exposed and non-exposed cohorts and their outcome characteristics are compared against each other to look for differences.