Yes, if one of two cases is true:
1. You test a representative sample size and conduct an engineering analysis of the relevant characteristics and factors, and show that your analysis is at least as capable as the component maker's original analysis and your conclusion of a longer life has an objective basis, such as that you can assure that the component always will be subject to less stress than the component maker assumed.
2. You test a statistically valid sample size to achieve a high confidence level. What confidence level would be acceptable would have to be determined. I'd guess that this requirement will depend on the device risk, with a moderate or high risk device requiring a quite high confidence level in this kind of study. This approach, being entirely dependent on testing, also would require a rationale to justify that the testing conditions are representative of all of the worst cases that are allowed by your instructions for storage and use, your worst case manufacturing stresses, and so forth.