Discerning between disease, drug reactions

Novel method used with cancer population

One of the chief problems with adverse event reports is these often cast a net so wide that many disease symptoms are swept up with the reports. This particularly is an issue with oncology clinical trials. Investigators have developed a strategy that should make it clearer which symptoms are related to the medication and which are related to the disease.

"I've been interested in adverse drug events, injuries related to the medical use of drugs," says Steven M. Belknap, MD, a research assistant professor at Northwestern University Feinberg School of Medicine in Chicago, IL.

"I was interested in cancer because it's often difficult to distinguish the toxicities of a cancer drug from the effects of the cancer itself and from comorbid conditions," he adds. "So we looked at the quality of the data reported in clinical trials and found it striking that the data quality is quite poor."

There was no good evidence that the adverse events were caused by the drug.

"It was difficult to distinguish the adverse events from all the other things going on with the patient," he says.

Belknap previously has looked at adverse event reports and was struck by how inaccurate these can be.

For instance, he came across a case where a patient had allegedly died from a reaction to an antibiotic. But when he looked through the record he found that the antibiotic had been ordered, but it was never given to the patient.

"In another case, the adverse event occurred before the first dose of the drug was given," Belknap says. "There are many problems like that."

These findings convinced him and other investigators that there might be a more accurate method of evaluating causality in adverse events in a clinical setting.

"We did that and had an amazing result," he says. "The drugs we thought were the main problems were not the main problems."

Once investigators found the root cause of the problems they assisted the hospital in reducing adverse events, he adds.

In the cancer research, investigators developed structured, symptom-specific case report forms that improved ADE reporting.1

They found that of 115 serious adverse drug events reported to the IRB the overall completeness of adverse event descriptions was 2.4 times greater for the structured case report forms versus the corresponding forms from IRB reports.1

The key to separating cancer disease symptoms from cancer drug side effects is to know the expected adverse events for any particular drug. Start with the medication label, Belknap advises.

"Cancer chemotherapy drugs are toxic and can be beneficial," he explains. "Applying this approach to cancer, we thought we'd look at how people were distinguishing cancer drug policy from other things happening to patients."

Investigators found that the ADE reporting lacked key elements needed to make causality, including the reporting done for clinical trials.

Questions guide process

A validated causality assessment tool, such as the Naranjo instrument, which was used in the ADE study, could have helped with this process. For instance, researchers could ask these simple questions:

  • Are there previous conclusive reports of this reaction?
  • Did the occurrence of adverse events occur after the drug was administered?

"We were interested in looking at how cancer drugs were reported, and the simplest thing was to look at cancer drugs and clinical trials done at 49 National Cancer Institute-designated centers," Belknap says.

When researchers examined the adverse event reports that were submitted to the IRB, they found that the forms and instruments used to collect the AE data did not explicitly ask for information that would be useful to describe adverse events, he says.

"We had a panel of experts convene, and they came up with a set of 34 items of criteria that we consider essential for describing an adverse event," he explains. "The median number of items present on the forms at the centers was four and the most was 11, so very few of the items that were necessary were being requested."

To capture the important information, all 34 items were needed and a form would be the best way to do so, he adds.

"The next step is to take the 34 items and set up tools that would extract all those and get better methods for doing it," Belknap says.

"We did not publish an instrument that has been tested," Belknap says. "Our next step would be to take our results and the criteria we used to judge these forms and then create an instrument."

One reason for the inaccurate data regarding adverse events is that researchers will make it a priority to protect subjects by withdrawing them from trials after they've had an AE, but they pay less attention to documenting the AE, he notes.

"The patients often are managed correctly," he says. "But there's not as much attention paid to making sure we learn from these AEs."

This is unfortunate because with better AE information, investigators and sponsors could learn better strategies for protecting patients during treatment with toxic cancer medications, Belknap says.

"But the first thing to do is get high quality data to obtain these insights," he adds. "If you don't get the data then you won't have information that can be turned into practical information that people can use."

Another strategy that will elicit better quality data on AEs would be to have two experts review each adverse event and agree on its cause.

"I did another study that looked at all the patients who came into an emergency room at an academic medical center, and we described adverse events and got a very good agreement when we had two experts review each record," Belknap says. "We implemented a system for inpatients in the hospital, doing much the same thing, and this approach has general applicability and not just for cancer drugs."

Reference

  1. Dorr DA, Burdon R, West DP, et al. Quality of reporting of serious adverse drug events to an institutional review board: a case study with the novel cancer agent, Imatinib Mesylate. Clin Cancer Res. 2009;15(11):3850-3855.