Cerebrospinal fluid shunt failure: late is different from early.

Six hundred and seventy-one operations to insert or revise simple, linear cerebrospinal fluid shunts were performed in a single institution from 1976 through 1989. The results were analyzed using actuarial techniques. Failures within the first 8 months were defined as 'early', and subsequent failures were 'late'. Patient age did not quite attain significance as a determinant of the risk of late shunt failure (p = 0.054). The following factors had no effect on the risk of late failure: interval since last operation, etiology of hydrocephalus, identity of surgeon, burr hole site, type of valve, and whether the shunt was new or revised. Modes of shunt failure in the two phases were compared: there were proportionately more infections among early failures and proportionately more fractures/separations among late failures (chi-squared = 42.9, d.f. = 6, p < 0.000005). Shunts failing during the early phase were more often revised by removal and replacement of the entire system, whereas shunts failing later were more often revised by replacement of valve and/or abdominal catheter (chi-squared = 33.7, d.f. = 3, p < 0.000005). The late phase of shunt survival conformed very closely to an exponential decay model with a constant monthly risk of failure of 0.5% (r2 = 0.98). Contrary to intuition, cerebrospinal fluid shunts did not encounter increasing monthly risks of failure as they grew older. The contempt of familiarity has discouraged study of the mechanisms of shunt failure. Experience suggests that most early shunt failures are attributable to infection and other technical misadventures, but the mechanisms of late failure are distinct and incompletely understood.