Page 85 - Safety Risk Management for Medical Devices

64 Safety Risk Management for Medical Devices

11.6.1.1 Method 1—Using published data
Method 1 is the preferred method because it relies on actual data. The steps to follow
for Method 1 are as follows:

1. Define search criteria, and exclusion criteria for the published papers.
2. Search for published papers, or other formal sources using the search criteria
from Step 1. There are many databases such as PubMed, or Cochrane that can
be used for this purpose.
3. Filter the results of the search based on the exclusion criteria.
4. Review the remaining papers to ensure appropriateness of the selected papers
for your System.
5. From each paper extract the number of observations of the Harms in the HAL in
each severity class of: (Catastrophic, Critical, Serious, Minor, Negligible)
6. Sum the total number of observations per severity class, per harm from all
papers. This makes the numerator for P 2 computations.
7. For each Harm in the HAL sum all the observations in all severity classes from all
the selected papers. This makes the denominator for P 2 computations.
Example: Let’s say we have selected 14 published papers that are relevant to the
System under analysis. And, let’s say the Harm of interest is Hemorrhage, intracranial.
The total number of reported intracranial hemorrhages in all papers 5 79
Total number of deaths from intracranial hemorrhage (catastrophic) 5 2
Total number of symptomatic, persistent intracranial hemorrhages (critical) 5 15
Total number of symptomatic, treated transient intracranial hemorrhages (serious) 5 25
Total number of asymptomatic untreated intracranial hemorrhages (minor) 5 37

In no case the physicians considered an intracranial hemorrhage an inconve-
nience/discomfort (negligible).

Therefore, P 2 (catastrophic) 5 2/79 5 2.5%
P 2 (critical) 5 15/79 5 19.0%
P 2 (serious) 5 25/79 5 31.7%
P 2 (minor) 5 37/79 5 46.8%
P 2 (negligible) 5 0/79 5 0.0%

Total 5 100%

Tip If you have access to registry data, it’s a preferred source over published clinical studies.
The reason is that in clinical studies in many cases patients are selected who are most likely to
benefit from the therapy. Whereas in registries all users of a device are taken into consider-
ation. And, the patient population is much larger than those in clinical studies. Therefore the
data is more representative of the “real world.” Also, in comparison with complaint data, regis-
try data is superior because for registries data is actively sought, while complaint data-
gathering is passive and likely subject to underreporting.