Results

Calculations of study sample size scenarios are presented in Table 3. For instance, assuming a 7-year study (4 years enrollment, 3 years minimum follow-up, 7 years maximum follow-up, 5 years average follow-up) with a newly identified structural disease cumulative event rate of 35% after 7 years and an annual dropout rate of 3%, a total of 1568 patients (355 primary outcome events) would be required to detect a 25% reduction in the hazard rate for the newly identified structural disease with central lymph node dissection with 80% power (Scenario 1, Table 3) at a type I error of 5% (two-sided). However, given that patients would all be cN0 (by preoperative US and intraoperative inspection), and thus represent low-risk PTC, the estimated newly identified structural disease rate would likely be significantly lower. PTC without pre- or intraoperative central neck nodal disease (i.e., cN0) has a low newly identified structural disease rate, even in the setting of histopathological presence of lymph node metastasis, with rates ranging from 2% to 9 % (12,14–20). Thus, a more realistic sample size scenario is presented in Table 3 as Scenario 2. Given the same assumptions as for Scenario 1, but with a cumulative 10% newly identified structural disease rate after 7 years, a total of 5840 patients (377 primary outcome events) would be required to achieve satisfactory statistical power. The key determinates of the study sample size scenarios are the reduction in hazard rate and the newly identified structural disease cumulative event rate as depicted in Fig. 2

In addition, detecting an increased complication rate in patients undergoing central neck dissection would require large sample sizes as well. For example, assuming rates of permanent hypoparathyroidism of 5% and 2.5% in those with and without central neck dissection (6), respectively, would require approximately 2000 patients to detect this difference with 80% power. Similarly, to detect differences in rates of unintentional permanent recurrent laryngeal nerve injury of 2% versus 1% with 80% power would require more than 5000 total patients (6). In contrast, postoperative levels of serum calcium and intact parathyroid hormone at postoperative day 1, prior to starting postoperative prophylaxis with calcium carbonate and/or calcitriol, could be measured with more modest sample sizes (25). However, the significance of detecting differences in serum calcium and intact parathyroid hormone on postoperative day 1 is of limited clinical value.

TABLE 3. EVALUATION OF SAMPLE SIZE SCENARIOS
Annual dropout rate
 
1%
2%
3%
4%
5%
Reduction in hazarda
N
Events
N
Events
N
Events
N
Events
N
Events
Scenario 1b
 20% 2441 586 2481 586 2522 586 2653 586 2605 586
 25% 1517 355 1542 355 1568 355 1593 355 1620 355
 33% 829 186 843 186 857 186 871 186 885 186
 40% 539 116 548 116 558 116 567 116 576 116
Scenario 2c
 20% 9043 623 9215 623 9389 623 9566 623 9746 623
 25% 5624 377 5731 377 5840 377 5950 377 6062 377
 33% 3078 197 3136 197 3196 197 3256 197 3318 197
 40% 2005 123 2044 123 2082 123 2122 123 2162 123

aProphylactic central lymph node dissection compared with no prophylactic central lymph node dissection. For both scenarios, assuming a 7-year study (4 years enrollment, 3 years minimum follow-up, 7 years maximum follow-up, 5 years average follow-up); 5% Type I error (two-sided); 80% power.

bAssuming the following newly identified structural disease (i.e., persistent, recurrent or distant metastatic disease) cumulative rates over the study period: Year 1=10%, Year 2=20%, Year 3=25%, Year 4=27.5%, Year 5=30%, Year 6=32.5%, Year 7=35%

cAssuming the following newly identified structural disease (i.e., persistent, recurrent or distant metastatic disease) cumulative rates over the study period: Year 1=3%, Year 2=5%, Year 3=6%, Year 4=7%, Year 5=8%, Year 6=9%, Year 7=10%.

 

FIG. 2. Total required sample size required in a prospective randomized controlled trial of prophylactic central lymph node dissection in cN0 PTC as a function of the cumulative event rate (newly identified structural disease) and reduction in hazard rate of 20%, 25%, 33%, and 40%. All scenarios assume a fixed annual drop-out rate of 3%.

 

Click here to enlarge image

Click here for a larger image