This is a value judgment. Only 3% of detectors are likely to fail in the first year, and annual replacement would be very expensive, so that doesn't make sense. At 15 years, the chances are better than 50/50 that your detector has failed, and that seems too big a risk to take. Manufacturers' warranties for the early detectors typically ran out in 3-5 years. So, in ten years there is roughly a 30% probability of failure before replacement. This seemed to balance safety and cost in a way that made sense to the responsible technical committees.
If a 30% failure probability still seems too high, remember that replacement on a schedule is only a backup for replacement based on testing. A national study found that when home smoke detectors fail, tend to fail completely. Regular monthly testing will help discover detector failure as well as a dead or missing battery.
The same study showed all the inoperable detectors tested in 1992 were at least 5 years old and predated a 1987 change in product standards that reduced sensitivity to reduce nuisance alarms. Changes in detector chip design, among other improvements, make it likely that electronic failure now occurs at a rate much less than 4 times per million hours of operation.
Replacing detectors after 10 years protects against the accumulated chance of failure, but monthly testing is still your best means of making sure detectors work. Today's detectors are even less vulnerable than the older models to failure.