Observational Studies

William A. Cantrell, BS

Kurt P. Spindler, MD

Dr. Spindler or an immediate family member has received royalties from nPhase; serves as a paid consultant to or is an employee of Cytori-Scientific Advisory Board, Mitek, and NFL; has received research or institutional support from National Institutes of Health (NIAMS and NICHD); and serves as a board member, owner, officer, or committee member of the American Academy of Orthopaedic Surgeons, the American Orthopaedic Society for Sports Medicine, and the Orthopaedic Research Society. Neither Dr. Cantrell 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.

INTRODUCTION

Observational studies are conducted to evaluate a patient population and then to determine prognosis of outcomes of interest. The studies first report the distribution of outcomes and potential predictors (risk factors) for the outcomes of interest in the population. Then, by analyzing these distributions, the associations between predictor and outcome can be understood, and thus prediction of outcomes can be inferred.¹ To gain these insights, different types of observational studies can be employed. The most common types of observational studies are a prospective cohort and retrospective cohort. Other types include cross-sectional and case-control. This chapter will define these study types, describe the differences between them, and suggest the situations for which each study design is optimal.

CROSS-SECTIONAL STUDIES

The cross-sectional study design provides insight into the population being studied at a single point or moment in time.¹ Thus, they are best suited for the descriptive studies, as opposed to predictive studies. When using a cross-sectional design to gain descriptive information, the investigator must recognize that only prevalence, and not incidence, can be determined. Prevalence describes the number of patients in a population who have a condition at a single point in time, whereas incidence is defined as the proportion of the population who have newly developed the condition over a set period of time.¹ Prevalence holds significant weight in the clinical environment because it informs clinicians of a patient’s likelihood of having a certain condition before obtaining a thorough workup, which, in statistical language, is labeled that patient’s prior probability of having the disease.¹ A good example of the cross-sectional study design and the importance of understanding prevalence is the study by Sher et al. published in The Journal of Bone and Joint Surgery (JBJS) in 1995, where MRI findings on 96 asymptomatic shoulders were described.² The study reported that 54% of patients (25/46) above 60 years of age with an asymptomatic shoulder had a rotator cuff tear visible on MRI of that shoulder, with 52% of those (13/25) being full-thickness tears.² The high prevalence of rotator cuff tears in patients with asymptomatic shoulders is a reminder to clinicians to be cautious when correlating rotator cuff as only cause of shoulder pain and symptoms.²

Although the cross-sectional design is better for a descriptive study, it can have limited use to study associations. When doing so, it is of the utmost importance to carefully draft an appropriate hypothesis, because in this case that hypothesis defines the predictors and outcomes.¹ Sometimes determining which variables serve which roles is rather straightforward, such as when using demographic factors such as age or sex,
as they are never dependent and so always serve as predictors. However, teasing out which is the cause and which is the effect sometimes is unknown. For example, if an obese patient has a low Marx activity score, it is challenging at that moment in time to determine which is the predictor and which is the outcome, and so those roles are dependent on the hypothesis applied. Therefore, in general, cross-sectional studies are primarily for descriptive purposes.

STRENGTHS AND WEAKNESSES OF CROSS-SECTIONAL STUDIES

Convenience is one of the major strengths of a cross-sectional design, as these studies are quick to complete and relatively inexpensive.¹ As a result, these types of studies are often included as an initial component of randomized controlled trials or cohort studies. One common reason for this approach is to obtain information, including demographic data, about the patients at baseline. In other study types, especially cohort and randomized controlled trials, losing patients to follow-up is a major concern because of the bias it can introduce into the results; however, because follow-up is not required for cross-sectional studies, this potential problem does not affect them.¹

Although these strengths are often useful, it is important to recognize and acknowledge the limitations of the cross-sectional study design. First and foremost, determining causal relationships is quite difficult in this approach, especially because of the lack of time-related data.¹ It can be difficult to distinguish whether a factor has a causative influence on a disease or whether it is merely associated with the disease being present. Answers to these types of questions can be explored better with the study designs discussed later in this chapter. Another area in which the cross-sectional study design does not perform particularly well is when the investigator intends on studying rare diseases or rare events; in this situation, the cross-sectional design is impractical because of the exceptionally large sample size that would be required.¹ Despite this limitation, one way that the cross-sectional study design has been applied with utility in this context is through a case series approach, where observations are reported on a sample that exclusively contains patients who have the desired condition or who have experienced the event in question.¹ The case series works well for providing information about disease characteristics, but it does not work well for answering comparative questions.¹

PROSPECTIVE COHORT STUDIES

In a cohort study, a group of patients is selected and observed over a period of time.¹ This can be done either prospectively, where patients are identified by the investigator and then followed into the future, or retrospectively, where patients are identified and their past records are examined.¹ The objective of a prospective cohort study is to measure or record potential predictors (risk factors especially modifiable ones) of outcome and then to follow up with the patients periodically to determine patient outcomes.¹ One of the most commonly known studies of this kind is the Framingham cohort study, which has given significant insight into many areas including the heart health, hypertension, and even noncardiac diseases such as dementia.³^,⁴^,⁵

In orthopaedic surgery, one of the major prospective cohort studies is the Multicenter Orthopaedics Outcome Network (MOON).⁶^,⁷ Over 3,500 patients who underwent anterior cruciate ligament reconstruction (ACLR) were enrolled.⁷ For these patients, numerous hypothesized predictors were recorded at the time of surgery, including body mass index (BMI), smoking status, age, preoperative physical examination, and a Marx activity score to indicate the patient’s preinjury level of activity. Baseline information for a variety of patient-reported outcome measures (PROMs), namely IKDC, KOOS, WOMAC, and Marx activity rating score, were also obtained. Each surgeon documented intraoperative data, including the presence of meniscus pathology (and treatment conducted) and the location and severity of cartilage defects. For follow-up time points, PROMs were completed by patients who were contacted either passively (via a direct mail questionnaire) or actively (via phone calls to the patients) at 2, 6, and 10 years postoperatively.⁷ From this cohort, numerous important insights have been gained, including the ACLR prognosis for PROMs at 10 years after ACLR and the modifiable risk factors of patient-reported outcomes (PROs) at that time.⁶ PROs were initially measured at 2 years postoperatively and then stayed constant at the 6- and 10-year follow-up points.⁶^,⁷ Because of the prospective design and high rate of follow-up (greater than 80%), the risk factors for worse outcomes at 10-year follow-up were able to be identified in a level I prognostic cohort. They were lower baseline scores on the PROMs, higher BMI, smoking, having a surgery for meniscus pathology before index ACLR, undergoing lateral meniscectomy during the index ACLR, undergoing a revision ACLR, grade 3 or 4 defects in the articular cartilage of any knee compartment, and undergoing a subsequent surgical procedure on the ipsilateral knee after the index ACLR.⁶ Knowledge of these predictors helps us better identify which patients are at a higher risk of worse outcome, better educate patients on modifiable risk factors, and encourage clinicians to act in ways to reduce risk of these factors developing (eg, choosing meniscectomy over meniscus repair in situations where repair success is in avascular zone because of the significant negative influence of a second surgery on outcomes).

Another example of an excellent prospective cohort study is the Multicenter ACL Revision Study (MARS).⁸ Patients
who underwent revision ACLR were enrolled in the study prospectively. In a similar fashion to MOON, preoperative and intraoperative data collection occurred on a variety of predictors including demographics, pathology, and surgical technique used. PROMs were also administered at baseline and then at 2 years follow-up.⁸ One of the most influential studies from this cohort showed that allograft had worse PROMs as compared with autograft for revision ACLRs at 2 years and the graft re-rupture rate for allografts was significantly higher.⁸

Only gold members can continue reading. Log In or Register to continue