Over the course of any one clinical trial, literally thousands of pieces of data are collected for every enrolled patient. Where is all that data stored, and how does it reach the pharmaceutical companies? Thus far, this department has considered how the research data is verified (March 2001) and how it is used to determine a medication’s efficacy (June 2001). The topic of study design has even been dissected (July 2001). However, the complexities of what data is collected and how it is recorded have largely been ignored. This month, we pull back the curtain to reveal the core of every research project.

The record
Whether called the “case record book”, the “case book”, the “case report form”, or most commonly, the CRF, all data collected during a study is contained in its binder and is divided into sections specific to each study-related patient visit. The pages usually consist of two-part forms that provide copies for both the investigative site and the pharmaceutical company.

The CRF is actually a copy of information maintained elsewhere in documents such as a patient’s medical chart (referred to as the “source”). Each investigative center must copy the relevant source data into the CRF. One role of the clinical trials monitor is to verify that the appropriate data was completely and accurately transcribed.

The pages of the CRF are designed by the sponsoring pharmaceutical company and are typically of the “fill-in-the-blank” format. In fact, with the common exception of a summary page located at the end of each study visit’s section, there is no place for an investigator or other staff to include extraneous information. Most pages are designed to follow the flow of the patient visits as outlined in the study protocol and consist of spaces for reporting the results of every diagnostic assessment and clinical measurement.

Certain parts of the case book are consistent from visit to visit. For example, each page includes the patient’s initials and randomization number (see “Clinical Trials Track”, July 2001), as well as the investigative site’s identification number. Every visit section ends with a summary page to be signed by the overseeing physician and includes a medication dispensation log, adverse events form, and a concomitant medications page—although these last two items may be charted either on a single form or as individual pages repeated in each visit’s section.

In addition, there are some consistencies between the CRFs for every trial, regardless of protocol specifics. These include pages meant to capture the complete medical and surgical history of a patient, a verification checklist that the patient is eligible to participate in the trial, a log of previous medications used by the subject, and a conclusion page that assesses the patient’s outcome (this is also known as a “termination page”). Most CRFs also have sections aimed at capturing data related to “screen failures”—patients who sign an informed consent and begin the study, but for certain reasons, such as outlying laboratory test results, cannot be included in the trial. Specially designed pages that record the outcome of patients who discontinue early from a study, as well as the reason for their discontinuation, are also common. (Patients may “early term” because of their own desires, the occurrence of an adverse event, a doctor’s decision, or a choice made by the sponsoring pharmaceutical company resulting from a patient’s violation of the terms of the research trial.)

It should also be noted that there is a difference between the CRFs meant for Phase I and II trials and those used in Phases III and IV. In particular, concomitant medication logs are not used in early trials because patients are not allowed to ingest non-study-related therapeutics. In addition, the necessarily lengthy blood records common in early research phases are not required in later phases.

Dirty data
Four sets of data that must be transcribed into the CRF present unique problems in that they are often riddled with errors and require vigilant review. Specifically, problems in maintaining the CRF come from the adverse events page, the concomitant medication log, the patient history page, and the drug dispensation record.

Adverse events are the most complicated for a site to track, but because of their use in maintaining patient safety and determining the side-effect profile for the experimental medication, they are of extreme importance. Universally, sites must record several things related to every reported event: the event, its onset date, resolution date, and severity. The problems that make this task complex result from patients not knowing, for example, what day they had a migraine or how many days they ran a fever. A major complication, however, occurs when the monitor, while reviewing source documents, discovers events that were not transcribed into the CRF or finds references to medications that were taken to relieve unrecorded symptoms. A further problem emerges when events that are reported as ongoing at one visit are not cited in subsequent visit records or are listed only at much later visits. (This final issue has to do with data continuity and questions of whether a specific event resolved and then recurred or was truly ongoing.)

Concomitant medication logs record the names of all nonstudy drugs used by patients along with their start and stop dates and dosing regimens. The problems with these records are similar to those cited for adverse events. Another concern about concomitant medications is that certain ones may be excluded from use during the trial. (For example, during a migraine-relief study, patients may be prohibited from using aspirin.) As a result, careful tracking of concomitant medication use is crucial.

Completing patient histories is often error prone, and drug dispensation logs sometimes “don’t add up”. When collecting patient histories, less data is more; it is a tedious task to cull a patient’s relevant medical and surgical history from lengthy hospital records. The opposite is true for tracking drug accountability, particularly in Phase III/IV trials in which patients self-administer their medications. Here, more information is crucial. For example, a common problem is a patient who leaves the clinic with 50 once-per-day pills, returns 30 days later with 25, and claims not to have missed a single daily dose. Missed doses may negatively impact the determination of a drug’s efficacy, especially if the skipped days were consecutive.

Finally, it should be mentioned that adverse events and concomitant medications are often “coded” using a specialized dictionary of terms. The purpose is to clarify the data. For example, in the absence of all other adverse events, if a site recorded that a patient suffered an ongoing “migraine” at one visit, listed a continuing complaint of “headache” at the next visit, and then resolved a citation of “headaches” at the third, proper coding would resolve the question of whether these are different events. Further, it would eliminate concern about whether the event was a migraine or a headache—an important medical distinction.

The inner sanctum
To be useful, CRF data must eventually be entered into a computer database; this is typically accomplished through double data entry, in which the entire data set is entered twice and then compared for inconsistencies. Databases are necessarily rigid and designed with numerous self-checks. (For instance, in the coding example just discussed, the database would query what it would consider to be three separate events, each containing missing start and/or stop dates.) As a result, once all queries are resolved, the data set on which all statistical analyses are based is consistent.

CRF design
For a typical Phase III/IV trial, a CRF contains 15 pages per visit per patient. Each page may contain as many as 50–60 individual data points, and a total of 15 visits is not uncommon. Given that a typical Phase III study involves around 500 patients, over 6 million pieces of data are obtained for an average trial. As a result, it is crucial that the CRF be designed to facilitate data collection.

A great deal of time is spent developing the clinical database to ensure that it can correctly handle the data that will be stored in its memory banks and to be certain that it accurately reflects the queries posed within the CRF. This part of clinical research is hidden from the patients—and often from the investigators—but is one of the most critical components of the research process.

CRFs should be carefully planned to limit the possibilities of confusing, missing, or inconsistent data. Some simple things can have a dramatic impact, like maintaining the adverse event log on a summary sheet as opposed to including a new form for each visit. Further, the use of on-site, electronic data entry (a “digital CRF”) is slowly emerging as a means of reducing errors and speeding the data collection and review processes. However, nothing that is done in the design phase alone will replace the study-specific training given to coordinators and the vigilant review of the data performed by monitors. After all, it’s the data that represents the sum of all the research knowledge gained.