Subscribe Bookmark RSS Feed
Richard_Zink

Staff

Joined:

May 27, 2014

Discovering unreported adverse events using your findings data

When designing case report forms (CRFs) for a clinical trial, it is important to minimize or eliminate redundancies in the collected information. Such redundancies can lead to inconsistencies that require a query to the clinical site for resolution. In a poorly designed CRF, data conflicts can be so numerous that they can delay locking the study database. This, in turn, affects all downstream analysis and reporting activities.

For example, since the effects of the study drug on the unborn are typically not known, it is important to regularly administer one or more pregnancy tests to women of child-bearing potential (WOCBP) during the course of a clinical trial. However, some women may not be required to take regular tests, such as those who are post-menopausal or surgically sterile. Since these reasons are highly unlikely to change, a CRF that collects a reason for not taking a pregnancy test at each and every instance is not only collecting a lot of unnecessary data, but it is also generating the opportunity for the data to contradict itself over the course of the study.

However, though not duplicated by poor CRF design, some reported adverse events can be confirmed using the "redundant" data that may be collected as part of the study database. For example, hypertension is typically defined as a systolic blood pressure (SBP) exceeding 140 mmHg with a diastolic blood pressure (DBP) in excess of 90 mmHg. If a subject without hypertension at baseline experiences hypertension over the course of the trial, this should be reported as an AE. Similarly, if an individual with hypertension at baseline experiences a worsening of his or her hypertension over the course of the trial, this should be reported as an AE.

Generally, the sponsor may have few data that coincide with the occurrence of a particular adverse event. However, vital signs are typically collected as part of routine medical care, so hypertension is quite straightforward to confirm using SBP and DBP measurements from the study database. If a study subject has the appropriate systolic and diastolic blood pressures (i.e., > 140 and > 90, respectively) without an AE of hypertension, check the subject's medical history. If the individual is not hypertensive at baseline, either the medical history is incomplete or an adverse event has gone unreported. If the subject is hypertensive at baseline, consider whether the individual has gotten worse over the course of the study to warrant the reporting of an event.

JMP Clinical 3.1 has a new time-to-event feature for SDTM findings data (Figure 1, above). This particular analysis allows you to define an event using multiple findings test codes, such as SBP and DBP for the hypertension example above. For the nicardipine data included with JMP Clinical, there are 517 subjects with hypertension in their medical history and 254 subjects who experience hypertension as an adverse event. However, the time-to-event feature identifies 52 subjects who meet the definition of hypertension based on study blood pressures (> 140 and > 90) who do not have a hypertension history or event term. Conversely, there are 31 subjects with a reported hypertension AE  who do not meet the blood pressure definition. There may be legitimate reasons for these inconsistencies, but these subjects are definitely worth a second glance.

Ideally, identifying these inconsistencies should be part of the regular data-cleaning activities so that they can be corrected during the course of the study. Other adverse events that can be confirmed through findings data could include thrombocytopenia (platelets < 20,000 cells/µL), neutropenia (neutrophils < 1500 cells/µL), ocular hypertension (intraocular pressure > 21 mmHg), pyrexia (temperature > 100° F), tachycardia (heart rate > 100 bpm) or bradycardia (heart rate < 60 bpm).