Table 3.
Discrepancies in sample size calculations reported in trial publications compared with protocols
| Component of sample size calculation | No of trials with discrepancy | ||
|---|---|---|---|
| Total | Not pre-specified* | Different from protocol description | |
| Outcome measure (n=31)† | 7 | 3 | 4 |
| Estimated delta (n=33)† | 12 | 6 | 6: 3 larger in protocol; 3 larger in article |
| Estimated event rates (n=16)‡ | 3 | 3 | 0 |
| Estimated standard deviation (n=14)§ | 5 | 2 | 3: 2 larger in protocol; 1 larger in article |
| Alpha (n=33)† | 2 | 2 | 0 |
| Power (n=34)† | 9 | 2 | 7: 5 larger in protocol; 2 larger in article |
| Calculated sample size (n=30)† | 8 | 0 | 8¶: 7 larger in protocol; 1 larger in article |
| Any component (n=34)** | 18 | 8 | 16 |
*Reported in publication but not mentioned in protocol.
†Among trials reporting component in publication.
‡Among trials reporting event rates for binary outcome measures in publication.
§Among trials reporting standard deviations for continuous outcome measures in publication.
¶Greater than 10% difference in calculated sample size.
**Among trials reporting any component in publication.