Biostatistics
Melissa Orr, BS
Nicolas S. Piuzzi, MD
Dr. Piuzzi or an immediate family member serves as a paid consultant to or is an employee of RegenLab and Stryker; has received research or institutional support from Regeneron and Zimmer; and serves as a board member, owner, officer, or committee member of American Association of Hip and Knee Surgeons and ISCT. Neither Melissa Orr nor any immediate family member has received anything of value from or has stock or stock options held in a commercial company or institution related directly or indirectly to the subject of this chapter.
ABSTRACT
It is important for orthopaedic surgeons to have a basic overview of research methodology, basic statistical principles, and a review of the current literature regarding biostatistics in orthopaedic research. The practice of evidence-based medicine requires orthopaedic surgeons to keep abreast of the latest clinical studies, with the ability to critically appraise research relevant to their individual practice.
Keywords: clinical relevance; evidence-based medicine; patient-reported outcome measures; statistical significance; study design
Introduction
It is imperative for orthopaedic surgeons to have a basic understanding of biostatistics and comprehension of research methodology, specifically (1) how to evaluate the validity of evidence; (2) the basics of clinical study design and common statistical tests; (3) interpretation of clinical relevance versus statistical significance; and (4) patient-reported outcome measures (PROMs) commonly used in orthopaedics.
Evidence-Based Medicine
Although higher levels of statistical methods are being included in medical literature, investigations have demonstrated that disparities often exist in the correct understanding and interpretation of results among the medical community.1,2 In surgery, specifically, descriptive studies predominate and the statistical methods chosen for a given study affect interpretation and application to practice.3 Evidence-based medicine (EBM) is defined as “the conscientious, explicit, and judicious use of the current best evidence in making decisions about the care of individual patients.”4 This translates to a process of integrating individual clinical expertise with external clinical evidence. Individual expertise refers to judgment acquired by clinical practice, and external clinical evidence refers to relevant patient-centered clinical research.5 Clinical evidence should inform but not replace individual clinical expertise. In practice, EBM involves applying knowledge from clinical trials, meta-analyses, and reviews to patient care for which an understanding of biostatistics is of key importance.6 Thus, for EBM to promote consistent treatment strategies and to establish standards of orthopaedic surgery practice, surgeons should have an understanding of study design and methodology along with biostatistics. The cycle of EBM7,8 is depicted in Figure 1.
Evaluating the Validity of Evidence
All health care professionals should deliver care based on the best available evidence to ensure outcomes. However, it is also the responsibility of practitioners to understand how external evidence can be applied to their clinical practices. Evaluating evidence occurs on both an internal (within the study) and external (outside of the study) level. Questions that can be used to evaluate clinical evidence are provided in Table 1.
Evidence-Based Orthopaedic Surgery
The application of EBM to orthopaedic surgery is not without challenges. In surgery, on-the-spot decisions are made and often affected by socialized knowledge within this specialized community.8 Although randomized
controlled trials (RCTs) are the highest level of evidence (Table 2), surgery is a complex intervention with variability often unsuited to RCT regulations, including placebo (sham surgery) and double blinding of treatment, resulting in lower percentage of RCT in orthopaedics compared with other fields.9 To adopt EBM to surgical practice, the health care practitioner needs to have appropriate knowledge to best interpret and understand the application of the literature to the question at hand. Furthermore, the development of large databases of prospective cohort studies can present high-quality evidence with more diverse patient populations.
controlled trials (RCTs) are the highest level of evidence (Table 2), surgery is a complex intervention with variability often unsuited to RCT regulations, including placebo (sham surgery) and double blinding of treatment, resulting in lower percentage of RCT in orthopaedics compared with other fields.9 To adopt EBM to surgical practice, the health care practitioner needs to have appropriate knowledge to best interpret and understand the application of the literature to the question at hand. Furthermore, the development of large databases of prospective cohort studies can present high-quality evidence with more diverse patient populations.
 Figure 1 Diagram shows the cycle of evidence-based medicine. Ask: formulating an answerable question. Acquire: a thorough search of relevant literature. Appraise: critical evaluation of evidence and application to current question. Apply: translating conclusions in the context of the current clinical problem. Act: evaluating the process by integrating the physician’s clinical judgment with the patient’s perspective. |
Table 1 Questions to Evaluate Relevant Literature | ||||
|---|---|---|---|---|
|
Table 2 Levels of Evidence | ||||
|---|---|---|---|---|
|
Proper Study Design
Hypothesis
When relying on EBM to guide decision making in clinical practice, a research hypothesis is tested by investigators. The null hypothesis (H0) states that there is no statistical difference between groups. The null hypothesis is deemed true until a study presents significant data to support its rejection. The alternative hypothesis (H1) is the presence of an effect.
Example H0: Body mass index has no effect on complication rate after total hip arthroplasty
Example H1: Body mass index has an effect on complication rate after total hip arthroplasty
Consider the example of probing the association of body mass index (BMI) with complication rate after total hip arthroplasty. A 2020 study compared the rate of surgical complications between patients undergoing total hip arthroplasty and found a significantly higher rate of complications for patients with BMI outside of the normal to overweight range.10 Thus, researchers are able to reject the null hypothesis in favor of the alternative.
Dependent and Independent Variables
Independent variables are what are expected to influence dependent variables; change in a dependent variable is the effect of change in the independent variable. For the aforementioned example, the independent variable is BMI and the dependent variable is complication rate. It is important to note that the relationship between independent and dependent variables is not always linear. In the aforementioned example, BMI both higher and lower than a normal-to-overweight range was associated with increased complications.
When designing a clinical study, deciding if the goal is to describe events or to study a treatment is categorized into two distinct study designs: analytic and descriptive studies11 (Figure 2).
 Figure 2 Diagram shows the study designs and categories. |
Descriptive Studies
Descriptive, or observational, studies describe a situation or events. No explanations of the relationship between any variables are offered. However, evidence from descriptive studies can prompt a hypothesis for additional studies. Examples of descriptive studies include cross-sectional, correlational (or ecologic), case series, and case reports (Table 3).
Cross-sectional studies involve collecting data from many individuals at a single point in time. For example, a study gave a questionnaire to patients about to undergo a knee replacement asking if they expected the surgery to benefit their pain level and improve function and examined associations between preoperative characteristics (eg, age, sex, comorbidities) with these expectations.12 Such studies can examine the relationship between patient variables and health outcomes as well as provide information on the prevalence of a given event. Advantages of cross-sectional studies include being low in cost and time commitment. Disadvantages include not being able to explain the event.
Table 3 Examples of Descriptive Studies | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Case reports or case series are made by clinicians on patients. Generally they describe rare events (eg, diseases, complications) or generate new hypotheses (ie, diagnostic methods or treatment strategies).11 A case report should present novelty,13 including unexpected presentation of a disease, unexpected associations between a disease and symptoms, adverse events, new or emerging diseases, or unusual adverse effects of medications. As described in a 2021 study, an example of a case report would be a postoperative skin lesion (scattered pruritic bullae) around the incision site 7 weeks after total knee arthroplasty.14 A case series exists in a research study that tracks subjects with a known exposure, such as patients who have received a similar treatment, or examines their medical records for exposure and outcome. An example of a case series, as described in a 2019 study, is an evaluation of 10 male patients with an age range of 10 to 25 years who all had a diagnosed traumatic displaced pure chondral fracture of the knee and underwent internal fixation.15 Within the orthopaedic literature, case series are the most commonly reported research.16 Disadvantages of case series include selection bias and lack of a control group.
Correlational studies determine a potential relationship between two variables and represent average exposure levels within a given population. The disadvantage of correlational studies is that correlation does not imply causation, and the correlation does not imply valid statistical association. Furthermore, relationships observed at the group level may not always apply to
individuals as well as potential confounding variables that are not taken into account. An example of a correlational study, as described in a 2021 study, is examining the association of preoperative prescription drug use with length of stay after total hip arthroplasty.17 In such a study, confounding factors of age, race, BMI, smoking status, and insurance are controlled. Although the results would not claim that preoperative drug use causes an increased length of stay, such a study would highlight an association and make clinicians aware of an increased risk within a patient population.
individuals as well as potential confounding variables that are not taken into account. An example of a correlational study, as described in a 2021 study, is examining the association of preoperative prescription drug use with length of stay after total hip arthroplasty.17 In such a study, confounding factors of age, race, BMI, smoking status, and insurance are controlled. Although the results would not claim that preoperative drug use causes an increased length of stay, such a study would highlight an association and make clinicians aware of an increased risk within a patient population.
Analytic Studies
Analytic studies answer a scientific hypothesis and use a sample to make inferences about the target population as a whole. The main categories of analytic studies are RCTs, cohort studies, and retrospective case-control trials.
RCTs constitute the gold standard of EBM.18 Participants are in a defined population and randomized into a treatment or control group. For example, a trial in which patients undergoing a revision total knee arthroplasty who are identified to be at risk for wound complications are randomized to receive either standard of care or closed incision negative-pressure therapy.19 Treatment groups can be a new or existing treatment, and the control group could be an existing treatment or no treatment at all (placebo).20 Participants are followed prospectively and treatment groups are compared. The disadvantages of RCTs are that they can be costly, time intensive, and not always feasible or ethical in a surgery discipline. Finally, because of selection bias, the results may not be generalizable to the entire population. For the aforementioned example, it would be unknown if the results could be applied to all revision cases or only those identified as high risk for complication. Application of RCTs to the population can be improved when they occur multi-institutionally. Because of the challenges involved with RCT and surgery, RCTs made up only 8% of the original research published in The Bone & Joint Journal between 2012 and 2017.9
Cohort studies are used to compare groups with similar baseline characteristics such as demographics, but who have undergone different exposures. These groups are followed either retrospectively or prospectively. Such studies can be used to approximate incidence or the proportion of new cases of a disease within a certain period of time. Cohorts are typically stratified by specific risk factors, which allow them to be followed prospectively to observe outcomes. As a result, inferences can be made about the prognosis of a risk factor.
Cases, or patients with a disease of interest, and control patients (patients without the disease) can be compared with retrospective case-control studies. The comparison is made across the level of exposure to a risk factor. Unlike cohort studies that select groups based on exposure status, case-control studies select groups based on disease status. The differences in exposure between cases and control patients help to find protective factors and risks associated with outcomes of interest. A challenging part of this study design is defining the base population and in the selection of control patients. These studies tend to be longitudinal in nature and provide an odds ratio as the primary outcome measurement. The odds ratio is defined as the odds of disease in exposed individuals compared with odds of disease in unexposed individuals. When examining rare diseases or events, this provides a good approximation of relative risk. Overall, if the odds ratio is less than 1, odds are decreased for a given outcome, and if odds ratio is greater than 1, the odds are increased for a given outcome.
Related posts:
Regulation of Orthopaedic Products
Structure and Biology of Normal and Diseased Bone
Polytrauma Care
Rotator Cuff Disease, Calcific Tendinitis, Adhesive Capsulitis, Throwing Shoulder, and Instability
Ligament Injuries of the Wrist
Articular Cartilage of the Knee: Defects, Degeneration, and Preservation
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
