The harms of early cessation of trials on systematic reviews

A novel laparoscopic pancreaticoduodenal training model: optimization of the learning curve and simplification of postoperative complications

PURPOSE

Laparoscopic pancreaticoduodenectomy requires a long learning curve. A preoperative training system was established to optimize the surgeons' learning curve and reduce the incidence rate of complications at the beginning of the curve.

METHODS

The laparoscopic pancreaticojejunostomy model, and choledochojejunostomy and gastrojejunostomy training systems were developed, and corresponding evaluation systems were also defined. Surgeons B and C performed laparoscopic pancreaticoduodenectomy after completing training session. Surgical outcomes, postoperative complications and their learning curves were analyzed.

RESULTS

Patients operated by surgeons B and C experienced shorter operative durations following training session than those in nontrained group (called A) ( P <0.001). B and C began entering the inflection point at the 26th and 20th case in learning curve, respectively. The incidence of postoperative pancreatic fistula in group B was 3.3%, significantly lower than 13.1% in group A ( P =0.047). Patients in group B showed significantly lower incidence of biliary-enteric anastomosis leakage (0% vs. 8.2%, P =0.029) and Clavien-Dindo classification greater than or equal to 3 (3.3% vs. 14.8%, P =0.027) compared with those in group A. The incidence of surgical site infection in groups B (3.3%, P =0.004) and C (4.9%, P =0.012) was significantly lower than that in group A (19.7%). Moreover, the length of postoperative hospital stay was significantly shorter in groups B (12.5±5.9 days, P =0.002) and C (13.7±6.5 days, P =0.002) compared with group A (16.7±8.5 days).

CONCLUSIONS

The laparoscopic pancreaticojejunostomy training model and evaluation system can shorten the operative duration, lower the risk of postoperative complications, and shorten the length of hospital stay.

Laparoscopic versus open pancreatoduodenectomy for pancreatic or periampullary tumours: a multicentre, open-label, randomised controlled trial

BACKGROUND

The benefit and safety of laparoscopic pancreatoduodenectomy (LPD) for the treatment of pancreatic or periampullary tumours remain controversial. Studies have shown that the learning curve plays an important role in LPD, yet there are no randomised studies on LPD after the surgeons have surmounted the learning curve. The aim of this trial was to compare the outcomes of open pancreatoduodenectomy (OPD) with those of LPD, when performed by experienced surgeons.

METHODS

In this multicentre, open-label, randomised controlled trial done in 14 Chinese medical centres, we recruited patients aged 18-75 years with a benign, premalignant, or malignant indication for pancreatoduodenectomy. Eligible patients were randomly assigned (1:1) to undergo either LPD or OPD. Randomisation was centralised via a computer-generated system that used a block size of four. The patients and surgeons were unmasked to study group, whereas the data collectors, outcome assessors, and data analysts were masked. LPD and OPD were performed by experienced surgeons who had already done at least 104 LPD operations. The primary outcome was the postoperative length of stay. The criteria for discharge were based on functional recovery, and analyses were done on a modified intention-to-treat basis (ie, including patients who had a pancreatoduodenectomy regardless of whether the operation was the one they were assigned to). This trial is registered with Clinicaltrials.gov, number NCT03138213.

FINDINGS

Between May 18, 2018, and Dec 19, 2019, we assessed 762 patients for eligibility, of whom 656 were randomly assigned to either the LPD group (n=328) or the OPD group (n=328). 31 patients in each group were excluded and 80 patients crossed over (33 from LPD to OPD, 47 from OPD to LPD). In the modified intention-to-treat analysis (297 patients in the LPD group and 297 patients in the OPD group), the postoperative length of stay was significantly shorter for patients in the LPD group than for patients in the OPD group (median 15·0 days [95% CI 14·0-16·0] vs 16·0 days [15·0-17·0]; p=0·02). 90-day mortality was similar in both groups (five [2%] of 297 patients in the LPD group vs six [2%] of 297 in the OPD group, risk ratio [RR] 0·83 [95% CI 0·26-2·70]; p=0·76). The incidence rate of serious postoperative morbidities (Clavien-Dindo grade of at least 3) was not significantly different in the two groups (85 [29%] of 297 patients in the LPD group vs 69 [23%] of 297 patients in OPD group, RR 1·23 [95% CI 0·94-1·62]; p=0·13). The comprehensive complication index score was not significantly different between the two groups (median score 8·7 [IQR 0·0-26·2] vs 0·0 [0·0-20·9]; p=0·06).

INTERPRETATION

In highly experienced hands, LPD is a safe and feasible procedure. It was associated with a shorter length of stay and similar short-term morbidity and mortality rates to OPD. Nonetheless, the clinical benefit of LPD compared with OPD was marginal despite extensive procedural expertise. Future research should focus on identifying the populations that will benefit from LPD.

FUNDING

National Natural Science Foundation of China and Tongji Hospital, Huazhong University of Science and Technology, China.

Effects of simulated sample sizes on the mortality effect estimates in three randomized intensive care unit trials

BACKGROUND

Randomized clinical trials (RCTs) are occasionally stopped prematurely before reaching their planned sample sizes. It has been suggested that early stopped RCTs are associated with under- and overestimation of the effect estimates. We simulated the effect of hypothetical premature stopping of three large RCTs done in the intensive care unit (ICU) setting.

METHODS

In this post hoc study, we simulated the impact of stopping trials early by calculating mortality effect estimates continuously after the inclusion of each individual patient in three large RCTs, that is the 6S trial on hydroxyethyl starch vs Ringer's acetate in sepsis in ICU, the TRISS trial on lower vs higher haemoglobin threshold for transfusion in septic shock in ICU and the SUP-ICU trial on pantoprazole in patients at risk for gastrointestinal bleeding in the ICU.

RESULTS

The three trials included a total of 5087 patients; 798 from the 6S trial, 998 from the TRISS trial and 3291 patients from the SUP-ICU trial. The premature mortality effect estimates showed considerable fluctuations until at least 20%-30% of the sample size was included. The premature estimates became stable after inclusion of 205 patients (26% of the final sample size) in the 6S trial, 133 patients(13%) in the TRISS trial and 1926 patients(59%) in the SUP-ICU trial.

CONCLUSIONS

In this post hoc study of three international RCTs within intensive care, we found that the simulated interim mortality effect estimates showed considerable fluctuations until at least 20%-30% of the sample size was included, but remained instable until the final sample sizes had been included. Thus, this study illustrates the necessity for cautious interpretations of prematurely stopped trials.