Four groups of patients (n = 458) treated with minimally invasive adrenalectomy were evaluated retrospectively or prospectively. Data of Group 1 consisting of 212 consecutive laparoscopic retroperitoneal adrenalectomy cases performed by Dr.Wang’s team from January 2010 to December 2014 was retrospectively evaluated. Data of group 2 consisting of 105 consecutive robotic retroperitoneal adrenalectomy cases performed by the same surgical team from February 2015 to June 2016 were prospectively collected and evaluated. Group 3 consisting of 28 consecutive cases of robotic transperitoneal adrenalectomy cases performed by Dr.Eun’s team from January 2014 to April 2016 and Group 4 consisting of 113 laparoscopic retroperitoneal adrenalectomy cases performed by Dr.Deng’s team from January 2014 to December 2016 was retrospectively evaluated. All patients were followed up for 4.5 to 5.5 years. Patient demographics in each group were listed in Table 1 . In summary, 74 (34.91%), 46 (43.81%), 16 (57.14%) and 47 (41.59%) patients were male in Group 1 to 4, respectively. The average age was 47.13 ± 13.03, 48.01 ± 12.20, 60.79 ± 10.64 and 45.50 ± 12.64 yr, respectively. The average body mass index (BMI) was 22.90 ± 3.40, 23.67 ± 3.66, 28.42 ± 6.93 and 23.33 ± 4.08, respectively.

Preoperatively, all patients underwent routine laboratory test for differential diagnosis of adrenal lesions including but not limited to blood potassium, aldosteronism test, urine and plasma cortisol, plasma catecholamines and urine vanillylmandelic acid. The presence and anatomical characteristics of adrenal masses were confirmed by non-contrast- and contrast-enhanced computed tomography (CT).

The ADRENAL score consists of 7 components that describe critical endocrinological, oncological and anatomical features of an adrenal mass as well as the patient body mass index. Each component in the ADRENAL score is designated by an English letter, forming the acronym ADRENAL: (A)ldosterone/cortisol/catecholamine secretion or suspicion of malignancy based on endocrinological and radiological study, (D)imension (tumor size as the maximal diameter), (R)elationship to adjacent organs, (E)nhancement on computerized tomography, (N)earness of the tumor to major vessels, (A)dipose (patient habitus as body mass index), and a (L)aterality descriptor. Of the 7 components, 6 (A.D.R.E.N.A.) are scored on a 0, 1 or 2-point scale. The 7^th descriptor is a suffix which describes the laterality of the mass, either left (l) or right (r). The scoring criteria were described in Table 2 and illustrated in Figures 1 , 2 and Supplemental Figure 1 .

>In authors’ institute, a team including three urologists, one endocrinologist, one radiologist and one anesthesiologist were assigned to the collection and revision of patient data. Two urologist independently recorded “(D)imension”, “(R)elationship to adjacent organs”, “(E)nhancement on enhanced CT scan”, “(N)earness of the tumor to major vessels”, “(A)dipose: BMI” and “(L)aterality” of the tumor which were verified by the 3^rd urologist and an radiologist if any discrepancy noted. The endocrinologist and radiologist were responsible for the evaluation and scoring of “(A)ldosterone/cortisol/catecholamine/malignancy”. Age-adjusted Charlson comorbidity score and ASA (American Society of Anesthesiologists) score of the patients were assessed and reviewed by an urologist with on-line Charlson comorbidity score calculator (http://www.touchcalc.com/calculators/cci_js) and an anesthesiologist, respectively. Tumor characteristics and average ADRENAL score in each group were listed in Table 3 . In summary, the average tumor size was 3.36 ± 1.95, 4.07 ± 2.72, 5.13 ± 3.94 and 2.97 ± 1.67 cm in Group 1 to 4, respectively. 110 (51.89%), 49 (46.67%), 10 (35.71%) and 58 (51.33%) tumors were located on the left side, respectively. The independently assessed ADRENAL score was 2.52 ± 1.89 (1 to 8), 3.26 ± 2.45 (1 to 9), 4.32 ± 2.50 (2 to 9) and 5.16 ± 1.83 (1 to 9), respectively.

Data are shown as the mean plus or minus standard deviation (SD). Statistical difference in ASA score and age-adjusted Charlson comorbidity score among different ADRENAL score groups were analyzed by one-way ANOVA test. The correlation analysis of the ADRENAL score and each component verse operative time (OT) and estimated blood loss (EBL) was performed by Pearson correlation test. The correlation of the ADRENAL score and OT or EBL was tested in linear regression model, in which an R² of greater than 0.5 was considered a significant correlation. Statistical difference in OT and EBL among Grade I, Grade II and Grade III groups were analyzed by Student t test. Pearson chi-square test was used to calculate the odds ratio. All data were analyzed with the Statistical Package for Social Science Software, v.16.0 (SPSS Inc., Chicago, IL, USA) and a p-value of less than 0.05 was considered statistically significant. Graphs were plotted in GraphPad Prism v.7 (GraphPad Software Inc., La Jolla, CA, USA) or Microsoft Excel 2010 (Microsoft Corp., Redmond, WA, USA).

Surgical outcomes of the four groups of patients were listed in Table 4 . In summary, the average operative time was 130.27 ± 50.30, 149.64 ± 72.64, 208.04 ± 120.65 and 106.96 ± 49.16 min in Group 1 to 4, respectively. To note that the operative time for both laparoscopic and robotic surgery was defined as from the first incision to the closure of all incisions. The average estimated blood loss was 134.96 ± 101.71, 85.14 ± 155.60, 261.60 ± 570.47 and 93.85 ± 153.49 ml, respectively. One case in Group 1, another case in Group 2 and 5 cases in Group 4 were converted to open surgery due to significant adhesion or excessive bleeding. No conversion was observed in Group 3. One case in Group 1, another case in Group 2 and 2 cases in Group 3 were reported to have positive surgical margin. Surgical margin was negative for all cases in Group 4. The average hospital stay was 15.70 ± 6.88, 15.34 ± 4.68, 3.46 ± 3.54 and 15.75 ± 8.18 days, respectively. Overall complication rate was 9.91% (n = 21), 3.81% (n = 4), 7.14% (n = 2) and 25.66% (n = 29), respectively. No patient death was observed during perioperative period in all series. Postoperative pathological study ( Table 3 ) revealed cortical adenoma [146 (68.87%), 63 (60.00%), 11 (39.29%) and 81 (71.68%) cases in Group 1 to 4, respectively] as the most common indication, followed by pheochromocytoma [45 (21.23%), 21 (20.00%), 6 (21.43%) and 22 (19.47%), respectively] and other pathologies including cyst, hematoma, angioleiomyolipoma, myelolipoma, and rarely B-cell lymphoma. Malignancy of the adrenal gland was rare [Group 1: 6 (2.83%); Group 2: 2 (1.90%); Group 4: 4 (3.54%)] but accounted for 32.14% (n = 9) in Group 3 due to selection of patients.

The ADRENAL score was applied to all patients according the scoring criteria as demonstrated previously ( Table 2 and Supplemental Table 1 ). Patients were grouped according to ADRENAL score in each series. The variance in age-adjusted Charlson comorbidity score (11) and ASA (American Society of Anesthesiologists) score (16), two reported risk factors of adrenal surgery, was first tested ( Supplemental Figure 2 ). Results showed that the difference in comorbidity and ASA score was statistically significant in Group 3 (p = 0.0119), but not in Group 1 (p = 0.4137), Group 2 (p = 0.2251) or Group 4 (p = 0.7208). A significant difference in ASA score was found in Group 4 (p = 0.0019), but not in Group 1 (p = 0.2721), Group 2 (p = 0.7546) or Group 3 (p = 0.2251). Keeping these in mind, we proceeded to validation of correlation between ADRENAL score and OT and EBL, two quantifiable major surgical outcomes.

We first tested the correlation between each of the 7 components in ADRENAL score verse OT and EBL ( Table 5 ). Based on the scoring criteria of ADRENAL score, a 0, 1 or 2 points was assigned to tumor function and malignancy (A), tumor size (D), relationship with adjacent organs (R), intratumoral enhancement (E), distance to major vessels (N) and BMI (A). Correlation between the scored value of the above 6 components and the actual value of tumor size (D) and BMI (A) verse OT and EBL was evaluated by Pearson correlation test. Results showed that these 6 components seemed to inconsistently correlate with OT and EBL in all 4 groups. The exact p values were listed in Table 5 . No significant difference was found in either OT or EBL between left and right tumors in either group by Student t test. Due to the selection of patients, we did not include tumors with local invasion in the indication of our laparoscopic adrenalectomy (Group 1). The (R) score was therefore 0 for all cases in Group 1 and cannot be analyzed.

Next, we analyzed the correlation between ADRENAL score verse OT and EBL. The correlation between the numeric value of ADRENAL score verse OT and EBL was consistently and statistically significant in all groups by Pearson correlation test ( Table 5 ). To further evaluate the value of ADRENAL score in predicting surgical outcomes, we tested ADRENAL score in linear regression model to see whether higher ADRENAL score correlated with longer OT or greater EBL ( Figure 3 ). Indeed, the numeric value of ADRENAL score verse OT and EBL fitted well in linear regression model as indicated by an R² of greater than 0.5 (R² was 0.8477 for OT and 0.8150 for EBL in Group 1; 0.6013 for OT and 0.7973 for EBL in Group 2; 0.8040 for OT and 0.8359 for EBL in Group 3; 0.7869 for OT in Group 4). However, the R² for ADRENAL score verse EBL in Group 4 was 0.4805.

For more practical use of the ADRENAL score, we proposed a 4-grade classification for adrenal masses based on the ADRENAL score. Grade 0: ADRENAL score = 0; Grade I: ADRENAL score < 4 (1-3); Grade II: ADRENAL score ≥ 4 but ≤ 7 (4-7); Grade III: ADRENAL score > 7 (8-12). Grade 0 adrenal masses or tumors assigned with an ADRENAL score of 0 were not indicated for surgery and managed by watchful waiting in all authors’ institutions, and therefore not included in this study. Percentage of Grade I, Grade II and Grade III tumors in each group were listed in Supplemental Figure 3 . Student t test was performed to determine any statistical difference in OT or EBL between any two grades in each group ( Figure 4 ). Indeed, both OT and EBL in Grade III verse Grade I was consistently statistically significant in all 4 groups (OT: p values for Grade I verse Grade III in Group 1 to Group 4 were 0.0428, 0.0116, 0.037 and 0.0033, respectively; EBL: 0.0322, < 0.0001, 0.0236 and 0.0013, respectively). However, OT or EBL of Grade I verse Grade II or Grade II verse Grade III in the 4 groups was not consistently statistically significant (OT: p values for Grade I verse Grade II in Group 1 to Group 4 were 0.0004, NS, NS and NS, respectively; Grade II verse Grade III: NS, NS, NS and 0.0011, respectively; EBL: p values for Grade I verse Grade II in Group 1 to Group 4 were NS, 0.0003, NS and NS, respectively; Grade II verse Grade III: NS, 0.024, NS and 0.013, respectively; NS: not significant), despite the fact that higher grade correlated with longer OT and larger EBL in all 4 groups ( Supplemental Tables 2 , 3 ).

Surgeries of Group 1 and Group 2 patients were performed by the same surgical team using conventional laparoscopy and da Vinci robotic system, respectively, both via retroperitoneal approach. By comparing the surgical outcomes of these two groups stratified by the 4-grade classification based on the ADRENAL score, which we believe was more reasonable than an overall comparison, we could determine whether robotic surgery was indeed superior to laparoscopic surgery eliminating any bias from different surgical techniques and approaches. We first did a non-stratified analysis of the difference in overall OT and EBL between Group 1 (laparoscopic) and Group 2 (robotic). Robotic surgery seemed to take a slightly longer time than laparoscopic surgery (146.21 ± 64.75 vs 130.89 ± 50.46 min) but with a significant p value of 0.0215, while the blood loss was significantly lower in robotic group than laparoscopic group (101.06 ± 138.88 vs 151.69 ± 94.14 ml, p = 0.0020). Stratified by the 4-grade classification based on the ADRENAL score ( Figure 4 ), no significant difference was found in OT between robotic and laparoscopic group in Grade I, Grade II, nor Grade III tumors. However, robotic surgery seemed to correlate with significantly lower EBL than laparoscopic group in both Grade I and Grade II tumors (p < 0.0001 and p = 0.0129, respectively), but not in Grade III tumors.

In terms of perioperative complications (Group 1: number of complications/number of cases = 21/212, overall complication rate: 9.91%; Group 2: 4/105, 3.81%; Group 3: 2/28, 7.14%; Group 4: 28/113, 24.78%), most were Clavien-Dindo Grade I including fever and vomiting managed with antipyretics and antiemetics, respectively. 9 patients in Group 1, 1 patient in Group 2 and 2 patients in Group 3 had postoperative infection and managed with antibiotics. 1 patient in Group 1, 2 patients in Group 2 and 1 patient in Group 2 had abnormal blood pressure and managed with vasoactive drugs accordingly. 1 patient in Group 2 and 1 patient in Group 3 converted to open surgery due to excessive adhesion and received blood transfusion. No Grade III to Grade V complication was encountered in patients included in this study. Overall, patients with Grade II and Grade III tumor seemed to have a higher, but not statistical significant, risk of perioperative complication rate than those bearing Grade I tumor ( Supplemental Table 4 ).

Conversion to open surgery was only observed in 1 Grade II case in Group 1, 1 Grade III case in Group 2, 1 Grade III case in Group 3, 2 Grade II cases and 3 Grade III cases in Group 4, due to significant adhesion ( Supplemental Table 5 ). All conversion was observed in Grade II or Grade III tumors, while none in Grade I tumors in all 4 groups. 2 Grade II cases in Group 1, 1 Grade III case in Group 2, 1 Grade III case in Group 3, 1 Grade I case, 3 Grade II cases and 3 Grade III cases in Group 4 received intraoperative transfusion due to excessive bleeding ( Supplemental Table 6 ).

Length of hospital stay was listed in Supplemental Table 7 . To note that the length of hospital stay in Group 1, 2 and 4 (Chinese hospitals), which was defined as from the first day of administration to the day of discharge, appeared to be significantly longer than that in Group 3 (US hospital) due to different policies for inpatient management. Results showed that patients bearing Grade III tumor had significantly prolonged hospital stay in Group 2 and Group 4, compared with patients with Grade I tumor in each respective group.

In this study, we included 4 groups of patients treated with laparoscopic or robotic adrenalectomy via retroperitoneal or transperitoneal approach, aiming to achieve an unbiased validation of the ADRENAL score. To improve the quality of the analysis, doctors involved in the data collection and scoring process were all blinded to the final outcome of statistical analysis; and a single or the same group of doctors was assigned to the scoring of a specific component in the ADRENAL score for each group. The rational of including the 7 components in the ADRENAL score was given in Supplemental Materials .

One potential limitation of this study is the lack of stratified evaluation of complications based on Clavien-Dindo classification (22). Further studies are required to establish a possible correlation between the ADRENAL score and the grade of complications. We also noted a significant variance (p = 0.0119) in age-adjusted Charlson comorbidity score among patients assigned with different ADRENAL score from Group 3. As such, the possible interference of patients comorbidity cannot be ruled out in all statistical analysis in Group 3. Another limitation is that we did not include open surgery in our study, which is still the standard of care for adrenal malignancy. Further external validation of the ADRENAL score is required.