Contributed Commentary by Kerina Bonar and Rob Sambrook

December 4, 2018 | Study designs and real-world data sources are evolving to meet drug development needs. The availability of real-world data, for example, continues to open new avenues of research beyond reliance on randomized controlled trials (RCTs). Cross-sectional surveys (CSS) and medical chart reviews (MCRs) are both common study designs, and each has its strengths and limitations. However, many of those limitations can be overcome through a hybrid design approach that uses both types of study designs to enhance each another. Just as in nature, hybrid studies have “hybrid vigor” in that they offer many benefits to researchers over each single research element. Whereas at one time, hybrid studies were difficult to do, they can now be performed efficiently, thanks to the availability of comprehensive, electronic medical records (EMR) databases.

Cross-Sectional Surveys

CSSs are an effective and inexpensive study design for gathering pertinent epidemiological data quickly to support drug development across the lifecycle. They select participants based on the inclusion/exclusion criteria defined in the study protocol to profile a cohort of individuals at a single point in time. That is to say, there is no baseline established, no outcome measured, and no follow-up data collected. Rather, certain population characteristics are identified that merely describe the cohort under study at a snapshot in time.

Although associations between certain characteristics within the survey cohort can be examined, CSS designs inherently lack an ability to examine the temporality of events e.g., cause and effect. For example, a CSS cannot be used to determine whether a risk factor preceded the onset of disease. Yet, CSS designs can provide the first step in identifying a possible association between variables and can suggest further avenues of research. The causality of any relationship appearing in a CSS would need further investigation using other suitable study designs such as longitudinal studies.

Medical Chart Reviews

Medical Chart Reviews (MCRs) can be performed manually when patient charts are in physical files, or electronically when the records are in a database. Importantly, they can capture one point in time or they can be longitudinal to explore causality and provide insights beyond incidence and prevalence.

These studies can be easier to conduct than CSSs in that patient informed consent is not always needed; in certain countries, it can be waived. For example, in a number of countries, conducting an MCR does not require approval from a Regional Ethics Committee or patient informed consent, provided that the data remain fully anonymized. This is a particularly strong benefit in disease areas such as oncology or psychiatry in which recruiting patients or maintaining a high response rate may be challenging.

Yet, medical charts are rarely complete for sponsors’ research purposes. A sponsor studying diabetes, for instance, may want to track patients’ Body Mass Index (BMI), but may find that BMI is not recorded at every patient visit as a routine part of patient care. This factor may contribute to the fact that MCRs appear to be underutilized. Other issues could be incomplete documentation, difficulty in understanding physician notes, poor quality photocopies, and the variability between medical records.

Hybrid CSS-MCR Studies

A hybrid CSS-MCR study enhances the MCR design by supplementing the capture of chart data with a survey component so that additional data can be collected. Each component of the hybrid study is conducted in the same manner as standalone CSS and MCR studies. After data are collected from the chart review, patients are surveyed to provide missing information or evaluate outcomes. By thus combining these study types, sponsors reap the benefits of:

• The ability to examine causality via longitudinal follow-up
• Reduced cost, as compared to a prospective study designed
• Lower participant burden, since surveys can begin where the chart review left off
• Potentially higher response rates given that surveys can be shorter

Hybrid studies, however, must be designed in a way that ensures that the findings are robust and meaningful, and generalizable to the population of interest. There is a risk of introducing bias such as selection bias and misclassification bias in both the survey and MCR study designs. These biases and other statistical issues, including uncontrolled confounding, could render the findings of a hybrid study less reliable, so it is advisable to rely on the expert advice of an epidemiologist when designing these studies.

Hybrid CSS-MCR studies can be advantageous in generating RWE to answer epidemiology questions efficiently and effectively, often negating the need for prospective research. However, such studies must be designed in a well-considered manner with design input from epidemiological experts so that they produce robust, meaningful, and generalizable results.

Kerina Bonar is an epidemiologist in ICON Commercialisation and Outcomes. She has several years’ experience working across a variety of observational study designs in pharma, public health and more recently consultancy.

Rob Sambrook is Divisional Principal, heading the epidemiology practice in ICON Commercialisation and Outcomes. He manages a team of 20 epidemiologists and analysts, with a strong focus on design, implementation, analysis and reporting for programs gathering and synthesizing real world data (RWD). He can be reached at Robert.sambrook@iconplc.com