Trial sequential meta-analysis of laparoscopic versus open pancreaticoduodenectomy: is it the time to stop the randomization?

Abstract

Background: The advantages of LPD compared with OPD remain debatable. The study aimed to compare the laparoscopic (LPD) versus open (OPD) for pancreaticoduodenectomy.

Methods: A meta-analysis of randomized studies (RCTs) comparing LPD and OPD was made. The results were reported as relative risk (RRs) or mean differences (MDs). The trial sequential analysis was used to test the type I and type II errors defining the required information size (RIS). The primary outcome was mortality, major morbidity, and postoperative pancreatic fistula (POPF). R1 resection, post-pancreatectomy hemorrhage, delayed gastric emptying, biliary fistula, reoperation, readmission, operative time (OT), lymph nodes harvested, and length of stay (LOS) were also studied.

Results: Four RCTs, counting 818 patients, were found. The RRs for mortality, major morbidity, and POPF were 1.16, 1.04, and 0.86, without significant differences. The RISs were 35,672, 16,548, and 8206. To confirm this equivalence, at least 34,854, 15,730, and 7338 should be randomized. OT was significantly longer in LPD than OPD, with an MD of 63.22. The LOS was significantly shorter in LPD than in OPD, with - 1.76 days. The RISs were 1297 and 1273, excluding a false-positive result. No significant differences were observed for the remaining endpoints, and RISs suggested that more than 3000 patients should be randomized to confirm the equivalence.

Conclusion: The equivalence of LPD and OPD for mortality, major morbidity, and POPF is affected by type II error. The RISs to demonstrate a superiority of one of the two techniques seem unrealistic to obtain.

Fig. 1

Flow diagram according to the PRISMA guidelines

Fig. 2

90-day mortality. Legend: A Forest plot; B–F the x-axis is the number of patients yet randomized; the y-axis is the cumulative z-score value representing the effect of each arm; and the blue line is the cumulative z-score obtained cumulating the studies. The dotted red horizontal lines are the conventional boundaries (p-value < 0.05). When z-curve crosses the conventional boundaries and the required information size (RIS) is not reached, the result is a false positive (“type I error”). When z-curve does not cross the conventional boundaries and RIS is not reached, the result is a false negative (type II error). The dotted black near-logarithmic lines are the monitoring boundaries. When the z-curve crosses the monitoring boundaries, the result is a true positive. The inverse dotted black lines are the futility boundaries (area in which any further randomization is useful). B At current RR equal to 1.16; C assuming that LPD could decrease by 50% the RR of mortality rates; D assuming that OPD could reduce by 50% the RR of mortality rates; E bearing that LPD could decrease by 25% the RR of mortality rates; and F assuming that OPD could reduce by 25% the RR of mortality rates

Fig. 3

Major morbidity. A Forest plot of meta-analysis; B–F the x-axis is the number of patients yet randomized; the y-axis is the cumulative z-score value representing the effect of each arm; and the blue line is the cumulative z-score obtained cumulating the studies. The dotted red horizontal lines are the conventional boundaries (p-value < 0.05). When the z-curve crosses the conventional boundaries and the required information size (RIS) is not reached, the result is a false positive (“type I error”). When z-curve does not cross the conventional boundaries and RIS is not reached, the result is a false negative (type II error). The dotted black near-logarithmic lines are the monitoring boundaries. When the z-curve crosses the monitoring boundaries, the result is a true positive. The inverse dotted black lines are the futility boundaries (area in which any further randomization is useful). B at current RR equal to 1.04; C assuming that LPD could decrease by 50% the RR of major morbidity rates; D assuming that OPD could reduce by 50% the RR of major morbidity rates; E bearing that LPD could decrease by 25% the RR of major morbidity rates; and F assuming that OPD could reduce by 25% the RR of major morbidity

Fig. 4

Clinically relevant postoperative pancreatic fistula. A Forest plot of meta-analysis; B–F the x-axis is the number of patients yet randomized; the y-axis is the cumulative z-score value representing the effect of each arm; and the blue line is the cumulative z-score obtained cumulating the studies. The dotted red horizontal lines are the conventional boundaries (p-value < 0.05). When z-curve crosses the conventional boundaries and the required information size (RIS) is not reached, the result is a false positive (“type I error”). When z-curve does not cross the conventional boundaries and RIS is not reached, the result is a false negative (type II error). The dotted black near-logarithmic lines are the monitoring boundaries. When the z-curve crosses the monitoring boundaries, the result is a true positive. The inverse dotted black lines are the futility boundaries (area in which any further randomization is useful). B at current RR equal to 1.04; C assuming that LPD could decrease by 50% the RR of clinically relevant postoperative pancreatic fistula rate; D assuming that OPD could drop by 50% the RR of clinically relevant postoperative pancreatic fistula rate; E bearing that LPD could reduce by 25% the RR of clinically relevant postoperative pancreatic fistula rate; and F assuming that OPD could decrease by 25% the RR of clinically relevant postoperative pancreatic fistula rate