Learning Curves and Timing of Surgical Trials: Robotic Kidney Transplantation with Regional Hypothermia

Abstract

Purpose:

Outcomes of surgical procedures can be affected by multiple factors including surgical skill and learning curve (LC). These factors need to be considered for optimal timing of surgical trials. We used the LC cumulative summation (CUSUM) method to describe the number of cases associated with competency of a single surgeon learning the technique of robotic kidney transplantation (RKT).

Methods:

Thirty-three patients underwent Vattikuti Urology Institute technique of RKT at a center that recently adopted this procedure (study group). Anastomoses times and short-term functional outcomes were compared with an established RKT program (reference group). LCs were evaluated using CUSUM analysis using target values from the reference group.

Results:

Mean ± standard deviation for console time, rewarming time (RWT), arterial anastomosis, venous anastomosis, and ureterovesical anastomosis times for the study group was 187 ± 34.6 minutes, 58.03 ± 17.81 minutes, 19.36 ± 5.91 minutes, 21.97 ± 6.78 minutes, and 22.55 ± 4.24 minutes, respectively, significantly longer than reference group (p < 0.001 for all). Mean ± standard deviation for serum creatinine at discharge and 1 month in the study group was 1.43 ± 0.57 mg/dL and 1.23 ± 0.35 mg/dL, respectively, similar to the reference group (p = 0.074 at discharge and p = 0.163 at 1 month). The LC was short, with competence achieved for RWT within 9, proficiency within 16, and mastery within 21 cases. Longer anastomosis times during the LC did not affect graft function.

Conclusions:

The LC of RKT is short, with improving skill up to 20–25 cases. The procedure is reproducible by surgeons experienced with open transplant and robotic surgery for other procedures, with comparable outcomes and low complication rates at a new center during adoption.

Introduction

Assessment of complex surgical procedures differs from drug trials in the inherent multifactorial nature of the procedures. The outcomes of a procedure can be affected by many factors, including initial surgical skill, inherent skill of the surgeon, standardization of the procedure, difficulty in blinding, and the effect of training.

Although randomized trial is the best way to evaluate a procedure, the learning curve (LC) of surgery is an important consideration while planning these trials. A procedure may be needed to be performed a certain number of times by a given surgeon to become competent and safe. Randomization before the LC is complete may result in bias in favor of the established procedure. In contrast, if the randomization is too late, there is a possibility of denying the benefit of an intervention that is better than the standard because of a delay in evidence generation. LC cumulative summation (CUSUM) has been suggested as a method to study LCs during the development of surgical procedures.¹

The Vattikuti Urology Institute technique of robotic kidney transplantation (RKT) with regional hypothermia was developed in 2013 by utilizing the IDEAL guidelines.^2,3 LCs of this procedure have previously been evaluated using target values from open kidney transplant (OKT).⁴ The procedure has since been adopted at various centers around the world.^5,6 This has provided an opportunity to evaluate the LCs with values derived from established RKT series.

There were three main objectives of the study: the first objective was to evaluate the LCs of RKT at a center that recently adopted this procedure, the second objective was to evaluate the impact of this LC on short-term outcomes, and the third was to determine the appropriate time to start a randomized trial comparing RKT with the established standard (OKT).

Patients and Methods

Study groups and surgeon characteristics

We included 33 consecutive patients undergoing RKT at a tertiary care center in Istanbul, Turkey, between January 2015 and October 2016, henceforth, referred to as the study group. The surgeon adopting the procedure (V.T.) had performed >500 robotic surgeries and 450 OKTs before the start of this study.

The reference group comprised 64 RKTs performed at a tertiary care center in New Delhi, India, between January 2015 and October 2016 by a single surgeon (R.A.) who was beyond the LC for RKT. The experience of R.A. before the start of this study was 125 RKTs, >2000 OKTs, and >500 robotic procedures.

The study center's team visited the reference center for 1 week to observe cases. In addition, the study center team was mentored for the first three cases in situ by the surgeon from the reference center (R.A.). Data were captured by independent data managers at each center who were not directly involved in patient care. Institutional review board approval was obtained from both the institutions.

Definition of variables

The baseline variables used in this study were age, gender, body mass index (kg/m²), ABO blood group incompatibility defined as transplantation performed across the ABO blood group barrier, preemptive transplant defined as transplantation performed before the initiation of chronic maintenance hemodialysis, graft side (side of the donor kidney procured), and multiple renal arteries defined as more than one renal artery.

Operative variables used in the study were warm ischemia time (time calculated from the clamping of donor renal artery and vein until the start of on-table perfusion of the graft), console time (time calculated from docking of the robot to completion of the console part of surgery) in minutes, total operative time (time calculated from incision until the last skin suture is placed), arterial, venous, and ureterovesical anastomosis times (calculated from the start of anastomosis until the tying of the last suture of the anastomosis), estimated blood loss calculated by suction outputs in milliliters, and rewarming time (RWT) (time calculated from introduction of the graft kidney in the recipient peritoneal cavity until its revascularization).

Postoperative and short-term follow-up variables used were serum creatinine at discharge, serum creatinine at 1 month of follow-up in milligrams/deciliter, complications classified per Clavien–Dindo classification, and delayed graft function defined as the need for dialysis postoperatively.

Statistical analysis

Data were reported as mean ± standard deviation for continuous variables. Proportions and percentages were used to express categorical data. Student's t-test and Fisher exact test were used to compare continuous and categorical data, respectively. Statistical analysis proceeded in two steps. First, the descriptive, operative, and postoperative characteristics of the study group were compared with those of the reference group. Second, LCs were evaluated using the LC–CUSUM method as described hereunder. A p-value of <0.05 was considered significant. SPSS statistical software version 22.0.0 (IBM Corp, Armonk, NY) was used for statistical analysis.

LC assessment

LCs were analyzed using the LC–CUSUM methodology as described previously.⁴ In brief, CUSUM analysis involves defining a target value, and sequential summation of differences between measured values and target values. The sequential summed value is plotted on the y-axis with the number of cases on the x-axis. The CUSUM charts plateau when the measured value matches the target value or the difference between the two values is minimal. As the attained value becomes smaller than the target, the curve shows a downward slope, indicating attainment of a higher skill. CUSUM for RWT, arterial, venous, and ureterovesical anastomosis times of the study group was plotted using target values derived from the reference group.

Accounting for “skill”

An inherent limitation of measuring the learning process is the “skill” component of learning. A surgeon may not achieve the “target” value even after learning, and the outcomes may or may not differ. Conversely, a “learning” surgeon may even surpass the “target.” In addition, the “reference” may itself be a moving target. Indeed, the “skill” and “time to complete a step of an operation” of surgeons may have a normal distribution, with 95% of surgeons lying on this bell-shaped curve. In this context, we defined “learning” as three phases: competence, proficiency, and mastery, arbitrarily defining “competence” as +2SD of mastery, and proficiency as +1SD of “mastery” values. Since the reference center surgeon had the highest experience of the procedure, the values from this surgeon were defined as “mastery” values.

Results

Baseline characteristics

Descriptive characteristics of the study and reference groups are summarized in Table 1. The study group had younger patients (37.97 ± 11.54 years vs 43.64 ± 12.34 years in the reference group; p = 0.03) and a higher proportion of females [12/33 (36.4%) vs 9/64 (14.1%) in the reference group; p = 0.018]. All grafts were procured laparoscopically from live donors. The warm ischemia time was significantly shorter in the study group (1.89 ± 0.51 minutes vs 2.80 ± 1.17 minutes in the reference group; p < 0.001). Induction immunosuppression was used in 29 cases [29/33 (87.9%), all antithymocyte globulin] in the study group and 61/64 cases (95.3%, basiliximab 31, ATG 30) in the reference group. Triple immunosuppression (tacrolimus/cyclosporine, mycophenolate mofetil, and glucocorticoid) was used in both groups as the maintenance therapy.

Table 1.

Descriptive Statistics of Consecutive Patients Undergoing Robotic Kidney Transplantation in the Study Group (Istanbul) and the Reference Group (Delhi) from January 2015 to October 2016

Variable	Study group	Control group	p-value
Number of patients	33	64
Age (years), mean ± SD	37.97 ± 11.54	43.64 ± 12.34	0.030
Body mass index (kg/m²), mean ± SD	23.06 ± 3.38	24.35 ± 6.33	0.278
ABO blood group incompatible, N (%)	0	5 (8%)	0.163
Preemptive, N (%)	14 (42.4%)	17 (26.6%)	0.167
Graft side, N (%)			0.484
Left	29 (87.9%)	59(92.2%)
Right	4 (12.1%)	5 (7.8%)
Multiple renal arteries, N (%)	5 (15.2%)	7 (11.3%)	0.747
Recipient gender, N (%)			0.018
Male	21 (63.6%)	55 (85.9%)
Female	12 (36.4%)	9 (14.1%)
Warm ischemia time, minutes, mean ± SD	1.89 ± 0.51	2.80 ± 1.17	<0.001

SD = standard deviation.

Operative and perioperative outcomes

The operative and perioperative outcomes are summarized in Table 2. There was one elective conversion to OKT in the reference group because of extensive intraperitoneal adhesions. No conversions were needed in the study group. The mean console time, RWT, arterial anastomosis time, venous anastomosis time, and ureterovesical anastomosis times in the study group were 187 ± 34.6 minutes, 58.03 ± 17.81 minutes, 19.36 ± 5.91 minutes, 21.97 ± 6.78 minutes, and 22.55 ± 4.24 minutes, respectively. These times were significantly longer than those of the reference group (p < 0.001 for all). The estimated blood loss in the study group was not significantly different from that in the reference group (p = 0.643). The mean serum creatinine at discharge and at 1 month was similar in both the groups (p = 0.074 and p = 0.163, respectively).

Table 2.

Operative and Perioperative Parameters of Consecutive Patients Undergoing Robotic Kidney Transplantation in the Study Group (Istanbul) and the Reference Group (Delhi) from January 2015 to October 2016

Variable	Study group Mean ± SD	Control group Mean ± SD	p-value
Console time (min)	187.27 ± 34.67	131.34 ± 23.87	<0.001
Rewarming time (min)	58.03 ± 17.81	44.87 ± 6.02	<0.001
Arterial anastomosis time (min)	19.36 ± 5.91	12.35 ± 2.79	<0.001
Venous anastomosis time (min)	21.97 ± 6.78	13.00 ± 2.13	<0.001
Ureterovesical anastomosis time (min)	22.55 ± 4.24	16.30 ± 4.44	<0.001
Total operative time (min)	275.00 ± 45.12	221.93 ± 25.70	<0.001
Estimated blood loss (mL)	176.67 ± 43.20	185.74 ± 132.03	0.643
Creatinine at discharge (mg/dL)	1.43 ± 0.57	1.23 ± 0.35	0.074
Creatinine at 1 month (mg/dL)	1.20 ± 0.43	1.33 ± 0.39	0.163

Perioperative complications

Clavien–Dindo grade greater than or equal to three complications occurred in two patients in the study group, and one patient in the reference group (two re-explorations for abdominal distension and unexplained ileus, respectively, in the study group; one exploration in the reference group for postoperative decrease in urine output with ambiguous Doppler parameters). All three re-explorations were negative. There was no vascular complication, delayed graft function, or graft loss in either group. One minor wound infection in the study group needed conservative management. One graft needed biopsy in the study group, whereas 9 of 64 (14%) needed biopsy in the reference group (mild acute tubular necrosis in 4 cases and acute cellular or humoral rejection in 5 cases). All biopsies were effectively performed percutaneously under ultrasound reference. There were no lymphoceles or ureteral complications.

LC analysis

The target values for the different phases of LC are shown in Table 3. These values were derived from means and standard deviations of the reference group. LC–CUSUM analysis revealed that the study surgeon could achieve competence for RWT after 9 cases, proficiency after 16 cases, and mastery after 21 cases. Phase 1 of the LC (competence) was absent for ureterovesical anastomosis time, as criteria for competence were met from the first case onward. Proficiency was achieved in 23 cases (Fig. 1B). For the arterial anastomosis, competence was achieved in 7 cases, and proficiency in 26 cases (Fig. 1C). For the venous anastomosis, competence could be achieved only after 29 cases.

FIG. 1.

(A) Learning curves (LCs) for rewarming time, (B) LCs for arterial anastomosis time, (C) LCs for ureterovesical anastomosis time, (D) LCs for venous anastomosis time. CUSUM, cumulative summation; SD, standard deviation.

Table 3.

Target Values for Various Steps of the Procedure, Used to Define the Phases of Learning Curve at the Study Center

	Target values (minutes)
Step of procedure	Competence	Proficiency	Mastery
Arterial anastomosis	17.9	15.1	12.4
Venous anastomosis	17.3	15.1	13.0
Ureterovesical anastomosis	25.2	20.7	16.3
Rewarming time	56.9	50.9	44.9

Discussion

The Vattikuti Urology Institute technique of RKT with regional hypothermia² has been shown to be associated with reduced pain and decreased blood loss compared with OKT. Functional outcomes have been shown to be similar to those of OKT, with the benefit of a lower complication rate.⁷ However, the data thus far have been retrospectively analyzed, and the true evaluation of this new procedure can only be done through a multicenter randomized trial. The results of such a trial, however, may be affected by the initial LC.

Considering the multifactorial nature of surgical learning, the evaluation of LC in this study accounted for both the initial skill of the surgeon and the differences in the final plateau of an individual surgeon. Although it may be difficult to match two surgeons in terms of skill, the number of cases is often considered a surrogate of experience. The selection of this study center was an attempt to match the robotic surgery experience with the reference surgeon. A perfect match of surgeons may not be possible. Both surgeons were beyond the LC in robotic surgery and OKT. The only difference was the experience in performing RKT, and this provided a unique opportunity to evaluate the LC for this procedure. For a surgeon with no robotic experience, the LC would include the cases needed to learn robotic surgery, and a similar LC would exist for transplant surgery. It has previously been demonstrated by Sood, et al. that the LC for a surgeon who has experience with both transplant surgery and robotic surgery is shorter than that of a surgeon with experience in one field.⁷ The second unique aspect about our methodology was the definition of target values. LC–CUSUM analysis is a useful way to identify small changes to the target of a process. Although competence refers to the bare minimum skill required, which is “adequate but not exceptional,”⁸ proficiency refers to a level of skill that is higher than the minimum and has been defined as “very skilled, and experienced at something.”⁹ Mastery is “extreme skill or command” over the procedure.¹⁰ The LC progresses through the steps of competence and proficiency before reaching a steady state of mastery. Although the time taken by a surgeon to complete a task is not the only indicator of his or her skill, the baseline assumption is that the time taken to complete various steps of the procedure would continue to improve during the learning phase before reaching a steady state to one's ability.

This study has several interesting findings. First, the learning phase of RKT is short for a surgeon who is proficient in both routine robotic surgery and open renal transplantation, with attainment of competence within six to nine cases for most of the steps of the procedure. Further improvement is seen in the next 20–25 cases. It is not surprising that the study center was already competent for the steps like ureterovesical anastomosis, as the study surgeon (V.T.) was experienced in other robotic urology procedures. Vascular anastomosis was the most difficult step of the procedure, as the study surgeon (V.T.) had not routinely performed robotic vascular anastomoses. Venous anastomosis was found to be the most difficult step of the procedure with criteria for proficiency and mastery not achieved within the study period (Fig. 1D). The difference between the competence time of arterial and venous anastomoses in this study may be explained by the much longer length of venous anastomosis and the higher number of throws needed to complete this step.

Second, short-term outcomes of the study group are no different from the reference group even during the learning process, despite the higher anastomosis times and RWTs. Various studies have shown that the anastomosis times,¹¹ and thus RWT,¹² have an impact on transplant outcome. A warm ischemia time of <10 minutes¹³ and total anastomosis time (indirectly RWT) of <45 minutes have been suggested to prevent graft dysfunction during OKT.¹¹ Every 10 minutes of RWT may be equal to 1 hour of cold ischemia time toward renal damage.¹¹ The previously less-than-ideal creatinine at discharge of RKT series using techniques without regional hypothermia also supports the hypothesis that regional hypothermia may be protective.^14
–16

Our study is not without limitations. The multifactorial nature of the LC means that it is impossible to control for all factors. Both the surgeons were experienced with robotic surgery and OKT. Therefore, the results are generalizable only to surgeons experienced in both robotic surgery and OKT. In the absence of more established RKT series, we had to rely on arbitrary definitions of “competence,” “proficiency,” and “mastery” based on mean values with standard deviations from a single center. Ideally, a singular cutoff derived from the mean value of a group of RKT surgeons might have been used. The follow-up of this study was short term; therefore, it is not a testament to outcomes of RKT and indeed that answer can only be provided by a well-designed randomized trial.

Conclusions

The Vattikuti Urology Institute technique of RKT with regional hypothermia has a short LC, with competence being achieved for RWT in about nine cases. A steady state may be expected within 10 cases, with improving skill, and averages, after 20–25 cases. The procedure is reproducible by surgeons experienced with open transplant and robotic surgeries for other urologic procedures, with comparable outcomes and low complication rates at a new center during adoption. The contemporary series of RKT may be well beyond the LC, and therefore, it is an appropriate time to start a randomized trial comparing it with the established standard.

Footnotes

Author Disclosure Statement

The authors declare that they have no conflict of interest.

Informed consent was obtained from all individual participants included in the study.

Abbreviations Used

References

Pennell

, Hirst

, Campbell

, et al. Practical guide to the idea, development and exploration stages of the IDEAL framework and recommendations. Br J Surg, 2016; 103:607–615.

Menon

, Sood

, Bhandari

, et al. Robotic kidney transplantation with regional hypothermia: A step-by-step description of the Vattikuti Urology Institute-Medanta technique (IDEAL phase 2a). Eur Urol, 2014; 65:991–1000.

McCulloch

, Altman

, Campbell

, et al. No surgical innovation without evaluation: The IDEAL recommendations. Lancet, 2009; 374:1105–1112.

Sood

, Ghani

, Ahlawat

, et al. Application of the statistical process control method for prospective patient safety monitoring during the learning phase: Robotic kidney transplantation with regional hypothermia (IDEAL phase 2a-b). Eur Urol, 2014; 66:371–378.

Breda

, Gausa

, Territo

, et al. Robotic-assisted kidney transplantation: Our first case. World J Urol, 2016; 34:443–447.

Tuğcu

, Şener

, Şahin

, et al. Robotic kidney transplantation: The Bakırköy experience. Turk J Urol, 2016; 42:295–298.

Sood

, Ghosh

, Jeong

, et al. MP29-07 Robotic kidney transplantation with regional hypothermia: Results from a prospective two-arm non-randomized controlled trial (Ideal phase 2B) [abstract]. J Urol, 2016; 195:e382.

Competent. Available at: www.dictionary.com/browse/competent?s=t. (Accessed on June 22, 2017 ).

Proficient. Available at: http://dictionary.cambridge.org/dictionary/english/proficient#dataset-american-english. (Accessed on June 22, 2017 ).

10.

Mastery. Available at: http://dictionary.cambridge.org/dictionary/english/mastery. (Accessed on June 22, 2017 ).

11.

Heylen

, Pirenne

, Samuel

, et al. The impact of anastomosis time during kidney transplantation on graft loss: A Eurotransplant cohort study. Am J Transplant, 2017; 17:724–732.

12.

Kukla

, Cholewa

, Chronowska

, et al. Effect of the second warm ischemia time and its components on early and long-term kidney graft function. Transplant Proc, 2016; 48:1365–1369.

13.

Tennankore

, Kim

, Alwayn

, Kiberd

. Prolonged warm ischemia time is associated with graft failure and mortality after kidney transplantation. Kidney Int, 2016; 89:648–658.

14.

Giulianotti

, Gorodner

, Sbrana

, et al. Robotic transabdominal kidney transplantation in a morbidly obese patient. Am J Transplant, 2010; 10:1478–1482.

15.

Oberholzer

, Giulianotti

, Danielson

, et al. Minimally invasive robotic kidney transplantation for obese patients previously denied access to transplantation. Am J Transplant, 2013; 13:721–728.

16.

Modi

, Pal

, Modi

, et al. Retroperitoneoscopic living-donor nephrectomy and laparoscopic kidney transplantation: Experience of initial 72 cases. Transplantation, 2013; 95:100–105.