Question 14 of 20 Intermediate Interpretation
How do you interpret a disproportionality statistic in spontaneous reports?
Short answer
A disproportionality ratio compares how often an event is reported with one drug against how often it is reported with all other drugs in the database. A value above 1, especially with a lower confidence bound above 1 and enough cases, flags a signal of disproportionate reporting to investigate. It is not a measure of risk or incidence.
Why
GVP Module IX Addendum I explains that the expected reporting proportion is estimated from reports that do not mention the drug, on the assumption that they reflect background reporting. If a drug's reports mention an event more often than expected, the ratio exceeds 1. Take a simple proportional reporting ratio.
Suppose 40 of a drug's 2,000 reports mention pancreatitis, which is 2 percent, while 1,000 of 200,000 reports for all other drugs do, which is 0.5 percent. The ratio is 4. Using the lower confidence bound rather than the point estimate makes a flag less likely when cases are few.
The addendum notes that limiting false positives works better by raising the minimum number of cases than by raising the ratio threshold. Interpretation needs care because spontaneous reporting is selective. Publicity, stimulated reporting, the indication itself and masking by heavily reported drugs all distort ratios.
A disproportionality flag is a reason to review cases clinically, not a conclusion that the drug causes the event.
A strong answer covers
- Ratio of observed to expected reporting proportion
- Lower confidence bound and minimum case count
- Worked arithmetic of a simple ratio
- Biases: stimulated reporting, indication, masking
- A flag triggers clinical review, not a causal conclusion
Common mistakes
- Reading a disproportionality ratio as a relative risk or incidence.
- Ignoring confounding by the condition the drug treats.
- Treating a ratio from very few cases as reliable.
The interviewer might ask next
How would confounding by indication show up in a disproportionality analysis, and how would you check for it?