• Provide information not available from the medical device reporting regulation

• Provide actual use of safety and effectiveness information.

ECRI (formerly the Emergency Care Research Institute), a not-for-profit health services research agency, maintains a version of the MDR and MAUDE databases with additional search capabilities. Both versions of the databases were used for this report. Searches were performed in December 2004. The strategy summarized in Table E.3 was used to search the MDR database.

This search produced 1,762 reports relevant to CSF shunts submitted to the MDR database from 1978 through 1996. Because of misclassifications leading to the inclusion of non-shunt devices and overlapping product codes that include CSF devices that are not implanted shunts (e.g., external ventricular drains, CSF shunt programmers, and devices for passing shunt catheters subcutaneously), the records were reviewed by the author (SJH) to identify misclassifications and reports not related to implanted CSF shunts. The final result was 1,555 reports regarding implanted CSF shunts.

The MAUDE database was searched in June 2004 with the strategy ((((“CEREBROSPINAL FLUID” OR CSF OR VP OR V-P OR VA OR VENTR* OR HYDROCEPH* OR CNS OR CENTRAL) AND SHUNT) OR (“NEURO VALVE” OR PUDENZ OR “SHUNT KIT” OR “POSTERIOR FOSSA” OR OSV OR DELTA)))). This produced 2,985 records of reports submitted from 1992 through 2004. These were reviewed for accuracy and a number of misclassifications by FDA product code were identified. Thirty-four were clearly shunts misclassified as another device (Table E.4). Others were non-shunt devices misclassified with a shunt code. Cleaning of the list resulted in 2,298 records of reports to the MAUDE database regarding problems with CSF shunts. An amended search in December 2004 increased the number of reports to 2,472.

In Part 2, we presented a classification of complications based on review of the literature. Both the MDR and MAUDE reports list reporter-classified outcomes as: death, serious injury, malfunction, or other. Because the MDR database could only be obtained in a text file, the “EFFECT TYPE” was identified by text searching the 1,762 reports that were identified in line S9 of the search strategy in Table E.3. All effect types other than “DEATH,” “SERIOUS INJURY,” or “MALFUNCTION” are grouped as “OTHER.” The MAUDE database was available in spreadsheet format allowing a more detailed search. The responses “OTHER,” “NO