Impact of Unplanned Intra-Operative Conversions on Outcomes in Minimally Invasive Pancreatoduodenectomy

Background Minimally-invasive pancreatoduodenectomy (MIPD) is fraught with the risk of complication-related deaths (LEOPARD-2), a significant volume-outcome relationship and a long learning curve. With rates of conversion for MIPD approaching 40%, the impact of these on overall patient outcomes, especially, when unplanned, are yet to be fully elucidated. This study aimed to compare peri-operative outcomes of (unplanned) converted MIPD against both successfully completed MIPD and upfront open PD. Methods A systematic review of major reference databases was undertaken. The primary outcome of interest was 30-day mortality. Newcastle–Ottawa scale was used to judge the quality of the studies. Meta-analysis was performed using pooled estimates, derived using random effects model. Results Six studies involving 20,267 patients were included in the review. Pooled analysis demonstrated (unplanned) converted MIPD were associated with an increased 30-day (RR 2.83, CI 1.62- 4.93, p = 0.0002, I2 = 0%) and 90-day (RR 1.81, CI 1.16- 2.82, p = 0.009, I2 = 28%) mortality and overall morbidity (RR 1.41, CI 1.09; 1.82, p = 0.0087, I2 = 82%) compared to successfully completed MIPD. Patients undergoing (unplanned) converted MIPD experienced significantly higher 30-day mortality (RR 3.97, CI 2.07; 7.65, p < 0.0001, I2 = 0%), pancreatic fistula (RR 1.65, CI 1.22- 2.23, p = 0.001, I2 = 0%) and re-exploration rates (RR 1.96, CI 1.17- 3.28, p = 0.01, I2 = 37%) compared upfront open PD. Conclusions Patient outcomes are significantly compromised following unplanned intraoperative conversions of MIPD when compared to successfully completed MIPD and upfront open PD. These findings stress the need for objective evidence-based guidelines for patient selection for MIPD. Supplementary Information The online version contains supplementary material available at 10.1007/s00268-023-07114-1.


Introduction
The adoption of MIPD into the pancreatic surgical armamentarium has been cautious owing to an appreciation of the technical challenges, high morbidity, pronounced volume-outcome relationship and long learning curves [1][2][3]. The initial randomized controlled data supporting feasibility of MIPD were from high-volume centres [4,5] with considerable experience in minimally-invasive surgery. The largest study (MITG-P-CPAM) [6] has shown a marginal, though statistically significant, benefit in terms of post-operative length of hospitalization (15 vs. 16 days). In general, studies comparing MIPD to open PD [7,8] demonstrate longer operative duration for MIPD with a reduced intraoperative blood loss and comparable postoperative morbidity and mortality. The Dutch multi-centre LEOPARD-2 trial [9] was the first 'real world' data to emerge in MIPD. The trial was prematurely terminated owing to increased complication-related 90-day mortality in the MIPD arm (10% vs. 2%) on interim analysis. It paved the way for introspection around the role of MIPD and the need for a systematic approach to the training of surgeons, as well as the uptake of MIPD in a controlled environment to obviate the risk of inadvertent harm to patients.
Not all PDs are equal [10]! MIPDs are fraught with an additional layer of complexity when compared to open PD, namely, the risk of conversion. The rates of this event have been reported to approach 40% [11]. Despite a previous nationwide training programme in LPD [12], the conversion rate during the LEOPARD-2 trial was 20% [9]. Experience from minimally-invasive colorectal [13] and liver [14] resections suggest that an unplanned intraoperative conversion not only nullifies the benefits afforded by minimally-invasive surgery, but may also worsen surgical outcomes. The impact of unplanned intraoperative conversions during MIPD on overall patient outcomes is yet to be fully elucidated [15]. This study aimed to compare perioperative outcomes of (unplanned) converted MIPD versus successfully completed MIPD and upfront open PD.

Study methodology Search strategy
Major databases (Medline, Google Scholar and Cochrane Library) were comprehensively searched to identify all relevant studies published between January 2000 and December 2022 using the MeSH keywords provided in Supplementary table 1.
The review was registered with PROSPERO (CRD42022355044) and performed in strict adherence to the PRISMA 2020 guidelines [16] (Fig. 1).

Inclusion Criteria
Studies fulfilling the following PICOS criteria were deemed eligible for inclusion in the systematic review: • P (Population): Patients undergoing MIPD (LPD or RPD) • I (Intervention of interest): Intra-operative Conversion ((unplanned) converted MIPD). Conversion was defined as any resection starting with a laparoscopic or robotic approach, but requiring either laparotomy or hand assistance for reasons other than trocar placement or specimen extraction • C (Comparator): • Data extraction and quality assessment Two authors (MK and MMW) independently extracted relevant data from the screened full-text articles according to a standardized Cochrane data extraction sheet [21], which included the following: Name of the first author, year of publication, sample size, baseline demographic characteristics (including age, sex, BMI), ASA grade, intraoperative characteristics including type of MIPD-laparoscopic or robotic or hybrid, operative duration, blood loss, transfusion requirements and markers of oncologic adequacy including R0 resection rates and lymph nodal harvest, post-operative outcomes such as CR-POPF, DGE, PPH, overall and major (C CD 3) complications, medical complications including cardiac (myocardial infarction, cardiac arrest), pulmonary (re-intubation rates, need for ventilator support [ 48 h, pneumonia) and AKI, UTI, VTE including DVT and PE, sepsis, SSI, need for post-operative PCD, LoS, readmissions, 30-day mortality and costs. The authors (MK and MMW) independently judged the quality of the studies using the NOS [22]. Any disagreement was resolved through mutual discussion, and the accuracy of the extracted data was adjudicated further by the senior author (SGB).

Data synthesis and analysis
The meta-analysis was run through 'meta' package in R software, version 4.2.3 (R Core Team, 2023) [23]. Outcomes expressed as median and inter-quartile range were converted to mean and standard deviation for pooled analysis [24]. The pooled effect size of converted versus completed and converted versus open on outcome was estimated using a random-effects model to account for both within-study and between-study variation and provides a more conservative estimate of the overall effect size. RR and associated 95% confidence intervals were calculated for dichotomous data by Mantel-Haenszel models, while SMD and associated 95% confidence intervals were calculated for continuous data using inverse-variance methods by restricted maximum-likelihood estimation process. Statistical heterogeneity among the included studies was assessed using I 2 , with an I 2 of 0-30, 30-60, 50-70, and [ 75% representing low, moderate, substantial, and considerable heterogeneity, respectively. The two-sided test was performed for all analyses, 95% confidence interval reported, and level of significance was set at 0.05. Publication bias was assessed using funnel plots and Egger's test.

Baseline demographics
Six retrospective cohort studies (four from the United States [25][26][27][28], one from Korea [29] and one multi-national European study [30]) including a total of 20,267 patients, published between 2017 and 2022 were suitable for inclusion (  [30], and Stiles et al. [27] with the propensity matched cohorts taken from Hester et al. [26]. Conversion rates ranged between 9.2 and 25.2%, with higher rates being noted in registry-based studies. There were no differences in age or BMI between groups. While Villano et al. [28] included patients with malignancy alone, the percentage of patients undergoing PD for malignancy ranged from 55-92% in other studies. There was no significant difference between the groups in terms of the use of neoadjuvant chemotherapy. (Unplanned) converted MIPD were associated with significantly longer operative duration compared to successfully completed MIPD [27] and upfront open PD [27,29]. Conversion was associated with a increased median intraoperative blood loss (500 vs. 275 ml; p = 0.005), and rate of blood transfusions (17.2 Vs. 42.2%; p \ 0.01) [25,30]. In studies that reported the outcome, conversion did not impact lymph node harvest [29], though patients who underwent an (unplanned) converted MIPD had lower rates of R0 resection compared to successfully completed MIPD and upfront open PD (71.9% vs. 77.8% vs. 77.7% p = 0.004) [28] ( Table 2).

Risk of bias and quality assessment
Of the included studies, 2 studies [27,28] received a NOS score of 9 with the remaining 4 studies [25,26,29,30] scoring 8, suggesting a high quality and low risk of bias for all studies (Table 1).
Significantly higher rates of respiratory complications (pneumonia, need for re-intubation and [ 48 h of ventilator dependence) were noted in the (unplanned) conversion group, while there were no significant differences in renal complications or thromboembolic events [25]. Patients in the (unplanned) conversion group also experienced higher incidence of sepsis including SSIs [25,26] Fig. 2a) (Table 4). In the 2 studies [28,29] comparing 90-day mortality between (unplanned) converted MIPD and upfront open PD, Connie et al. [29] had zero event rates, precluding a pooled analysis.

Discussion
These data demonstrate that (unplanned) converted MIPD were associated with an increased 30-and 90-day mortality and overall morbidity compared to successfully completed MIPD. Patients undergoing (unplanned) converted MIPD experienced significantly higher 30-day mortality, CR-POPF and re-exploration rates compared upfront open PD. Observational, case-matched studies on MIPD [31] [32] [33] demonstrate longer operative times, but less operative blood loss and shorter hospitalization in LPD with complication rates and oncological outcomes comparable to OPD. However, registry-based studies [34,35] have advised caution owing to increased mortality rates after LPD, especially in low-volume centres. The four RCTs [4][5][6]9] published to date comparing LPD vs. OPD have been unable to demonstrate a clear indication of post-operative morbidity and mortality, as they were likely underpowered to detect these differences. The conversion rates in these trials ranged from 3 to 25%. Unfortunately, since converted patients were predominantly analysed in the laparoscopic group on an ''intention-to-treat'' basis, the true implications of an intra-operative conversion were not readily evident. The meta-analysis by Zhang et al. [36] comparing LPD vs. open PD highlighted the advantages of LPD in terms of lesser intra-operative blood loss, higher R0 resection rates and lymph node yield, lower perioperative overall morbidity, and shorter length of hospitalization. No difference in survival was noted. In keeping with the IDEAL framework for surgical innovation, all novel interventions should preferably be evaluated against the current standard in a randomized controlled trial (RCT) [37]. And so, an updated meta-analysis of RCTs comparing LPD vs. open PD confirmed a significantly lower blood loss and surgical site infection rate in the LPD cohort, while the approaches were similar with respect to other outcomes [38]. The benefits of MIPD in terms of improvements in optics, surgical instrumentation, and increased access to training [12,39] have led to an increased interest amongst surgeons to attempt MIPD.
This study presents the most updated appraisal of the literature on the impact of unplanned intra-operative conversions in MIPD acknowledging that they constitute an inherent problem in the learning curve of minimally-invasive surgery. Interestingly, there were no significant differences in the rates of pancreas-specific complications (CR-POPF, DGE), re-explorations rates, or readmissions rates. Though less frequent in comparison to LPD, unplanned conversions in RPD had more significant consequences, which may be attributed to the longer duration in RPD to actually convert to an open procedure with consequently greater blood loss. It could also reflect a problem of selection bias wherein only the most difficult procedures were converted in RPD, whereas LPD procedures were converted more easily [40]. The significantly higher 30-day mortality, CR-POPF and re-exploration rates in the unplanned conversion MIPD cohort compared to the upfront open PD cohort is also intriguing bearing in mind that the data in the present study is from retrospective series, wherein patients for MIPD would have been following a strict selection policy.
Though broadly defined into three phases [41], that is, competency, proficiency, and mastery, there exists little standardization in literature on what constitutes an  [42] based on intra-operative parameters such as operative duration and/ or intraoperative blood loss rather than postoperative outcomes such as complication rates or LoS. Additionally, there exists a significant correlation between study sample sizes and number of procedures needed to surpass the learning curve, questioning the meaningfulness and applicability of these results [43]. It is paramount that future studies investigating the subject take into account not only surgeon (previous surgical [44] and simulation [45] experience, procedure-specific training and clinical fellowships [46]) and patient (BMI [47], comorbidities and tumour factors [10]) characteristics, but institutional expertise as well, which includes annual procedural volumes and team familiarity [48]. Patient selection appears to be of paramount importance in MIPD, given the risks associated with unplanned conversions. The current review highlights patient (elderly, male, smokers, ASA III/ IV, recent history of weight loss), pancreatic gland, and tumour ([ 4 cm, pancreato-biliary tumours) characteristics associated with higher risk of intra-operative conversion, that need to be factored in while selecting patients for MIPD. The Miami guidelines state that trainees should have passed the learning curve for open PD ([ 60) before undertaking training in MIPD. Centres should be performing at least 50 PD annually in addition to minimum annual volume of 20 MIPDs [1]. This may be operationalized by devising strict national and international surgical society guidelines. Further, utilization of risk prediction scores like Difficulty scoring system (DSS) [49] or PD-ROBOSCORE [50] will likely aid better patient selection for MIPD.
The study is not without limitations. The definition of conversion varied between centres/ studies. Categorization of conversion into elective and emergency in future studies may facilitate rational comparison of outcomes. Secondly, studies might have had surgeons at different stages of learning curve, and a uniform definition of learning curve might enable assessment of the impact of surgeon experience and centre volume on the risk of conversion. Finally, we combined LPD and RPD together due to non-availability of stratified data, though it is obvious that both the approaches differ not only in terms of impact of conversion, but risk factors as well.
As surgical innovations become more complex and the burden of age and comorbidities in the surgical patient population continues to increase, understanding the benefits and risks associated with surgical interventions becomes ever more important. We need to move beyond the traditional endpoints of mortality and resource use towards more pertinent measures of morbidity, patient-reported outcomes, and functional status. At the present time, the implementation of MIPD must be guided by an appreciation of surgeon training and (also institutional) capability and optimum patient selection, as unplanned conversions are fraught with the attendant risk of morbidity and mortality.