Standard pancreatoduodenectomy versus extended pancreatoduodenectomy with modified retroperitoneal nerve resection in patients with pancreatic head cancer: a multicenter randomized controlled trial

2.1 Participating hospitals

Initially, 3 hospitals agreed to participate in this study. Then, 3 other centers showed interest in participating and joined the study one year later. Ultimately, patients were enrolled from 6 tertiary hospitals (all are high-volume centers in pancreatic cancer diagnosis and treatment in China, Supplementary Table S1).

2.2 Patients and study design

This multicenter, randomized, controlled, nonblinded, parallel-group trial compared SPD versus EPD with retroperitoneal nerve resection for pancreatic head cancer. The protocol complied with the Declaration of Helsinki and was approved and overseen by the clinical ethics committee of each participating hospital. This study was registered at the Chinese Clinical Trial Registry (http://www.chictr.org.cn, no: ChiCTR-TRC-12002548). Patients with resectable pancreatic head tumors eligible to undergo pancreatoduodenectomy were enrolled between October 3, 2012, and September 21, 2017. All listed authors had access to the study data and approved the final manuscript. The study sponsors had no role in the design and conduct of the study.

2.3 Inclusion and exclusion criteria

Patients were included if they (1) were 18 to 80 years of age (the upper limit age was changed from 70 to 80 years due to an increase in the number of adults 70 years and older, and pancreatic cancer tends to occur at an older age), regardless of gender; (2) had potentially curable cancer of the pancreatic head (stage I or II according to the American Joint Committee on Cancer [AJCC] 7^th edition), as shown on preoperative imaging examinations (enhanced computed tomography, magnetic resonance imaging/magnetic resonance cholangiopancreatography, endoscopic ultrasound, endoscopic retrograde cholangiopancreatography, positron emission tomography-computed tomography or fine-needle aspiration biopsy); (3) had a Karnofsky performance status score > 70; (4) had Loyer grade A to D type; (5) had no obvious surgical contraindications; and (6) provided written informed consent. Patients were excluded if they (1) had an unresectable condition or metastasis (stage III or IV) found during surgery; (2) had a pathologic diagnosis of a benign tumor of the head of the pancreas or a tumor in the tail of the pancreas; (3) had surgical contraindications; (4) had a history of other malignancies; (5) were pregnant (urine human chorionic gonadotropin (HCG) > 2500 IU/L, diagnosed as early pregnancy), planned to become pregnant or were lactating; (6) had received other specific pancreatic cancer treatments (preoperative chemotherapy or chemoradiation) within three months; (7) had mental disease; (8) participated in other clinical trials 3 months before; or (9) had impaired visceral function (cardiac function 3-4, alanine aminotransferase[ALT] and/or aspartate aminotransferase [AST] exceeding 3 times the upper limit, creatinine [Cr] beyond the upper limit).

2.4 Study treatment

In the SPD group, lymph nodes (LNs) around the gastric pylorus (stations 5, 6), LNs around the pancreatic head (stations 13a, 13b, 17a, and 17b), LNs anterior to the common hepatic artery (station 8a), LNs on the right side of the hepatoduodenal ligament (stations 12b, 12c), and LNs on the right side of the root of the superior mesenteric artery (SMA) (stations 14a, 14b) were dissected without retroperitoneal nerve resection. During EPD, the following nerve tissues at the retroperitoneum and LNs around the pancreatic head were dissected: (I) nerves and soft tissues between the inferior vena cava (including the aortic plexus) and abdominal aorta (including LN stations 16a2 and 16b1); (II) all the nerves and soft tissues around the hepatoduodenal ligament, which were completely dissected and skeletonized (including whole LN station 12); (III) the common hepatic artery, which was isolated, and its surrounding nerves and soft tissues (including LN stations 8a and 8p); (IV) the celiac trunk, which was isolated, and its nerves and soft tissues 270° around the right longitudinal axis (including LN station 9); (V) the root of the SMA, which was dissected to open the vascular sheath, and the proximal uncinate process mesentery, which was removed along its right side, the nerves and soft tissues 270° around the right longitudinal axis (including LN stations 14a and 14b); and (VI) the nerves and soft tissues in the dense postpancreatic connective tissues that fixed the pancreas at the celiac trunk-aorta-SMA artery axis. Both the portal vein (PV) and the superior mesenteric vein (SMV) were resected and reconstructed in patients with PV/SMV involvement to achieve radical resection. Differences in the extent of resection are summarized in Supplementary Table S2 and Figure 1.

First, all participating hospitals were high-volume pancreatic surgery centers (completing more than 100 pancreatoduodenectomy operations each year), and all the surgeons participating in this trial had rich surgical experience (the cumulative number of pancreatoduodenectomies exceeded 300 surgeries). Second, for the homogenization of the surgical technique, the Pancreatic Cancer Committee of the Chinese Anti-Cancer Association conducted multiple unified training sessions and assessments on the operators before the start of the study [8]. Specifically, (1) unified surgical specifications and procedures were formulated; (2) strict assessment of EPD and SPD surgical proficiency from each operator by independent expert evaluation was performed; (3) 6 meetings (April 15, May 20, June 24, July 15, August 12 and September 15, 2012) were held to discuss the surgical videos and unify the surgical procedures before patient enrollment; and (4) all randomized cases were evaluated for heterogeneity through objective data, such as surgical images (Supplementary Figure S1) and videos, by third-party experts to ensure homogeneity of patient enrollment.

2.5 Randomization

Patients were randomly assigned at a 1:1 ratio to undergo SPD or EPD with retroperitoneal nerve resection using a stratified permuted block method. Patients were stratified according to the coordinating hospitals and predefined preoperative cancer antigen 19-9 (CA19-9) (< 200.0 U/mL or ≥ 200.0 U/mL). Computer-generated random assignment lists were created at the School of Public Health, Sun Yat-sen University (Guangzhou, Guangdong, China). The assignments were placed in sealed envelopes, labeled by stratum, and unsealed only after exploratory laparotomy. Investigators at each center enrolled participants and assigned them to their corresponding interventions. Each patient received a unique study number that remained unchanged throughout the trial.

2.6 Study outcomes

The primary endpoint of this study was OS by ITT, calculated from the date of randomization until the date of death from any cause. The secondary endpoints were disease-free survival (DFS), mortality, morbidity, and postoperative pain intensity. DFS was calculated from the date of randomization until the date of the first event of either recurrence or death from any cause. The censored data were defined as patients without events at study termination or those lost to follow-up at any time during the study. Prespecified subgroups were defined for preoperative CA19–9 (< 200.0 U/mL or ≥ 200.0 U/mL based on the predictive value of CA19–9 from previous reports [9-11] and discussions of the researchers), and we performed subgroup analyses with this cutoff for OS and DFS. Moreover, we performed post hoc subgroup analyses of other factors, including age (< 60/≥ 60), gender (male/female), preoperative carcinoembryonic antigen (CEA) (< 5.00/≥ 5.00 ng/mL), PV/SMV resection (yes/no), type of resection (R0/R1), N stage (N0/N1), tumor-node-metastasis (TNM) stage (I/IIA/IIB), histology (well differentiation [WD] and moderate differentiation [MD]/poor differentiation [PD]), perineural invasion (+/-) and adjuvant treatment (yes/no), for OS and DFS in the ITT population. Pain intensity was assessed with a numerical rating scale (NRS) [12]. An 11-point scale from 0 to 10 was used to describe pain (minimum to maximum). The NRS was explained to the patients, and they were asked to circle the number best representing the degree of their pain on the last day of preoperation (baseline, day 0), 1 month postoperation, and 6 months postoperation. The patients were followed up once a month from 1 to 6 months after surgery, every 3 months from 6 months to 3 years after surgery, and every 6 months for 3 years after surgery. The follow-up examinations included routine blood tests, biochemical tests, digestive tract tumor index detection, B-ultrasound, and enhanced computed tomography (CT) scan examinations. The short form 36 (SF-36) was used to assess the patients’ general health status. The SF-36 is composed of eight multi-item scales (physical functioning, physical role, pain, general health, vitality, social function, emotional role, and mental health), with scores for each of these scales (or dimensions) ranging from 0 to 100, whereby a higher score indicates a higher health-related quality of life [13]. The patients completed the SF-36 questionnaire with the assistance of the follower. The SF-36 score and pain score of the patients were assessed preoperatively and every 3 months after surgery. Each follow-up included gathering information on the chemotherapy regimen, adjuvant therapy regimen and adverse reactions. The last date of follow-up was defined as three calendar years since the last enrolled patient underwent surgery.

An interim analysis was conducted on the primary endpoint when 50% of patients were randomized and completed the 6-month follow-up. The interim analysis was performed by an independent statistician blinded to the treatment allocation. The trial was calculated to end using symmetric stopping boundaries at P < 0.001.

2.7 Pathological analysis

The LNs and nerve tissues removed during dissection were marked with the exact location and sent for a pathological examination after resection. Pathology reports contained the primary pathologic diagnosis, the extent of disease, margin status, LN status and overall pathological stage as previously described. The stages of the resected specimens were classified according to the AJCC TNM classification (7^th edition). All participating surgeons observed intraoperative photographs or videos of both groups to ensure consistency of the extent of nerve, tissue and LN dissection.

2.8 Postoperative chemotherapy

Chemotherapy was recommended for all patients except for those with a poor performance status (grade 3-5 by ECOG [Eastern Cooperative Oncology Group] performance status scale) [14] or organ dysfunction and those who refused adjuvant treatment. The first postoperative chemotherapy cycle started within 2 months of surgery. The chemotherapy regimen was as follows: gemcitabine 1000 mg/m², dissolved in 100 mL normal saline, intravenous infusion, completed within 30 minutes; once weekly, after 2 consecutive times, rest for 1 week, every 3 weeks for a course of treatment, for 8 consecutive courses.

2.9 Postoperative pain control strategy

In this study, for postoperative pain control, all our patients followed the principle of pain ladder treatment. For mild pain, we recommended using nonopioid adjuvant analgesics (ibuprofen, aspirin). For moderate pain, we recommended using weak opioids plus or minus nonsteroidal anti-inflammatory drugs (NSAIDs) plus or minus adjuvant pain relievers (tramadol, celebrex, codeine, etc.). For severe pain, we recommended using opioids plus or minus nonsteroidal adjuvant painkillers (morphine, duloxetine, etc.)

2.10 Sample size and statistical analysis

Based on previous RCTs [3, 6, 15] and our pre-experiment analysis, our trial was powered for the superiority of OS data at 3 years according to the operation, assuming that the 3-year OS rate (35%) of patients from the EPD group was 15% higher than that of patients from the SPD group (20%). Our estimation showed that an enrollment of 180 patients would provide 80% power to detect the superiority of a procedure with a 1-sided α = 0.05 and β = 0.2. Taking into account an estimated 10% dropout rate, the sample size was increased to 200 participants per group (n = 400).

The ITT population was defined as all randomly assigned patients. The qualified population was defined as patients enrolled in randomization who underwent appropriate surgery without rule violation for the extent of surgical dissection, with ductal adenocarcinoma as the final pathology, with a complete case report form and without loss to follow-up. Analyses were conducted using the ITT principle, irrespective of any protocol deviations or violations.

Patients who underwent the operation to which they were originally allocated and satisfied the criteria for optimal surgery based on photographs uploaded to our data center were evaluated. Categorical variables are expressed as proportions, whereas continuous variables are expressed as the medians (range, minimum to maximum) or means (standard deviation, SD) where appropriate. Missing data were handled using a Markov chain Monte Carlo multiple imputation approach with the assumption of missing data at random [16]. Continuous variables were reported as the median and range, where appropriate, and compared using Student's t-test (when the data conformed to a normal distribution) or Mann‒Whitney U test (when the data did not conform to a normal distribution). Nominal data were compared using χ² tests or Fisher's exact test. Survival outcomes were calculated using the Kaplan‒Meier method and compared using the log-rank test (stratified for predefined preoperative CA19–9 [< 200.0 U/mL or ≥ 200.0 U/mL]). Variables revealed as statistically significant by the univariate analysis were included in the multivariate analysis, which was performed using a Cox proportional hazards regression model. Both univariate and multivariate analyses were performed for the treatment method. To evaluate the association between levels of CA19–9 and disease-progression survival, a multivariable Cox model with restricted cubic splines (RCS) was built. RCS has been widely described as a valid strategy for analyzing the relationship between survival and independent variables [17, 18]. The analysis was performed with R software version 4.2.1. To minimize the effect of confounding factors and potential bias between the SPD and EPD groups, the propensity score was calculated using logistic regression, and we performed 1:1 patient matching by the nearest-neighbor matching method without replacement. We used a caliper radius equal to 0.2 of the standard deviation to prevent poor matching. The variables included in the matching model were age, gender, preoperative CA19–9 value, PV/SMV resection, N stage and T stage. All statistical analyses were performed using Statistical Package For The Social Sciences (SPSS) version 23.0 (SPSS Inc., Chicago, IL, USA), and a 2-sided P value less than 0.05 was considered statistically significant.

2.11 Data management

After random assignment, the following data were collected from all patients: clinical and pathological information, details of the operative procedure (including photographs of the operation field and a surgeon questionnaire detailing the operative findings), and other relevant information. Adverse events were graded according to the Common Terminology Criteria for Adverse Events (CTCAE, version 4.0). Follow-up was obtained from hospitalization or outpatient records, telephone calls, and assistance from the Chinese public security administration. Assessors were blinded to the treatment groups (i.e., the SPD group vs. the EPD group).

3.1 Patients

Between October 3, 2012, and September 21, 2017, 468 patients from six Chinese centers were screened (Figure 2). After enrollment, 68 patients were excluded for the following reasons: 28 patients refused to participate despite initially agreeing, 32 patients had unresectable or metastatic tumors during the intraoperative exploration, and 8 patients actually received distal pancreatectomy. Thus, 400 eligible patients were enrolled and randomly allocated into two groups (Figure 2). The last follow-up date was expected to be on September 21, 2020, but to further improve the integrity of the data, we extended the follow-up time to November 30, 2020.

Of the remaining 400 patients, 199 were randomized to receive SPD, 201 were randomized to receive EPD with retroperitoneal nerve resection and constituted the ITT population. Of them, 21 were excluded for the following reasons: (1) having undergone inadequate or inappropriate surgery, as revealed by reviewing the images of the surgical field (n = 6); (2) having a nonductal adenocarcinoma (n = 2), such as ampullary adenocarcinoma and duodenal adenocarcinoma; or (3) having inadequate case report forms (n = 9) or withdrawing consent (n = 4, two by the patients and two by the surgeons). Consequently, 379 qualified participants were included, with 187 in the SPD group and 192 in the EPD group. The two groups were well balanced in the following characteristics at baseline: median age, gender, preoperative CEA and CA19–9 levels, preoperative total bilirubin, and resection rate of the PV/SMV (Table 1). The operation time was 45 minutes longer in the EPD group than in the SPD group (P < 0.001), with no difference in either estimated blood loss (EPD vs. SPD, median [range, mL], 300 [20–2,500] vs. 300 [20–6,000], P = 0.113; Table 1) or red blood cell (RBC) transfusion (P = 0.870; Table 1). Overall, 343 patients underwent adjuvant treatment after surgical resection (Table 1).

3.2 Postoperative pathological differences between groups

Table 1 also shows the pathological characteristics of the two groups. The percentage of patients who underwent R1 resection was similar in the SPD group (12.56%) and the EPD group (8.46%; P = 0.195). In addition, there were no differences between the two groups in tumor size, T stage or AJCC TNM stage. As expected, the total number of retrieved LNs was significantly higher in the EPD group (median [range], 20 [7–63]) than in the SPD group (median [range], 15 [7–46]; P < 0.001). Moreover, a significantly higher number of positive LNs was observed in the EPD group than in the SPD group (P = 0.005), but the LN (+) ratio was not significantly different between the two groups (P = 0.101).

3.3 Morbidity and mortality

Although there was a trend for higher morbidity in the EPD group, there was no significant difference between the two groups (EPD vs. SPD, 37.81% vs. 34.17%; P = 0.467; Table 2). Moreover, no significant difference was found between the two groups in postoperative hospital stay (EPD vs. SPD, median [range], 18 [6-105] vs. 18 [9-49], P = 0.232; Supplementary Table S3). The incidence of diarrhea in the EPD group within 3 months of surgery was 7.46% compared with 5.03% in the SPD group; no significant difference was found (P = 0.409), suggesting that dissection of the retroperitoneal plexus 270° on the right side surrounding the celiac trunk and SMA might not significantly affect bowel movement. The inpatient mortality was similar between the two groups (EPD vs. SPD, 0.50% vs. 0.50%, Table 2). Regarding the specific cause of death, 1 patient in the EPD group died of severe sepsis with gastroduodenal artery rupture, and 1 patient in the SPD group died of cirrhotic liver failure.

3.4 Survival and recurrence data

For the ITT analysis, all randomly enrolled patients were included. Kaplan‒Meier analysis of OS showed a better prognostic trend for the EPD group than for the SPD group, but the difference was not significant (Figure 3A). The median OS (ITT population) was 23.0 months in the EPD group versus 20.2 months in the SPD group (hazard ratio (HR), 0.84; 95% CI, 0.68 to 1.04; P = 0.100; Table 2). The OS rates in the EPD and SPD groups were 76.62% and 70.35% at 1 year, 45.77% and 40.20% at 2 years, 29.35% and 22.61% at 3 years, and 9.95% and 9.05% at 5 years, respectively (Table 2). DFS was significantly better in the EPD group by ITT (EPD vs. SPD, 16.1 months vs. 13.2 months; HR, 0.80; 95% CI, 0.66 to 0.98; P = 0.031; Figure 3B and Table 2).

3.5 Subgroup analysis

The predefined subgroup with a preoperative CA19–9 < 200.0 U/mL showed a significantly improved median OS and DFS for EPD (EPD vs. SPD; OS: 30.8 months vs. 20.9 months; HR, 0.68; 95% CI, 0.50 to 0.92; P = 0.009; DFS: 23.4 months vs. 13.5 months; HR, 0.62; 95% CI, 0.47 to 0.83; P < 0.001, Figure 3C-D and Table 3). In the subgroup with a preoperative CA19–9 < 200.0 U/mL, the respective 1-, 2-, 3-, and 5-year OS rates were 82.24%, 59.81%, 40.19% and 16.82% in the EPD arm and 72.55%, 42.16%, 26.47%, and 10.78% in the SPD arm (Table 3). The predefined subgroup of patients with a preoperative CA19–9 ≥ 200.0 U/mL showed no significant difference in OS or DFS (Figure 3E-F and Table 4). Post hoc analysis of subgroups by age (< 60/≥ 60), gender (male/female), PV/SMV resection (yes/no), preoperative CEA (< 5.00/≥ 5.00 ng/mL), resectability (resectable/borderline resectable), type of resection (R0/R1), N stage (N0/N1), TNM stage (I/IIA/IIB), perineural invasion (positive/negative) and postoperative adjuvant treatment (with/without) showed no differences in OS (Supplementary Table S4). In addition, in the subgroup with well and moderate pathological differentiation, the EPD group showed superior oncologic outcomes in terms of median OS and DFS (OS: HR, 0.72; 95% CI, 0.54 to 0.96, P = 0.020, Supplementary Table S4; DFS: HR, 0.69; 95% CI, 0.52 to 0.91, P = 0.006, Supplementary Table S5). In the subgroup analysis (Supplementary Table S5), with the results of some prognostic factors presented in Supplementary Figure S2 and Supplementary Figure S3, survival benefits for DFS were found from EPD treatment in patients with preoperative CEA < 5.00 ng/mL (HR, 0.75; 95% CI, 0.59 to 0.96, P = 0.018, Supplementary Figure S2B), without PV/SMV resection (HR, 0.79; 95% CI, 0.63 to 0.99, P = 0.038, Supplementary Figure S2F), with a resectable stage (HR, 0.74; 95% CI, 0.58 to 0.94, P = 0.013, Supplementary Figure S3H), with R0 resection (HR, 0.77; 95% CI, 0.62 to 0.95, P = 0.014, Supplementary Figure S2H), and with postoperative adjuvant chemotherapy (HR, 0.77; 95% CI, 0.62 to 0.96, P = 0.019, Supplementary Figure S2J).

As mentioned in the methods section, the prespecified threshold of the subgroup for CA19–9 was 200.0 U/mL. The rationale for this threshold was examined through receiver operating characteristic (ROC) and RCS analyses. The ROC analysis identified that the best cutoff value of preoperative CA19–9 was 198.7 U/mL for predicting tumor DFS, a value very close to 200.0 U/mL (Figure 4A-B); in addition, as shown in the RCS model (Figure 4C), the HR values for DFS in cases with CA19–9 ≥ 200.0 U/mL were consistently greater than those for cases with CA19–9 < 200.0 U/mL, both supporting the prespecified threshold of 200.0 U/mL.

To evaluate the benefit of adjuvant chemotherapy according to different CA19–9 levels, we compared the prognosis of patients stratified by adjuvant chemotherapy application in the patients with CA19–9 < 200.0 U/mL and those with CA19–9 ≥ 200.0 U/mL. In the subgroup with preoperative CA19–9 < 200.0 U/mL, there was no difference in OS and DFS between those receiving or not receiving adjuvant chemotherapy. In contrast, in the subgroup with preoperative CA19–9 ≥ 200.0 U/mL, adjuvant chemotherapy improved both OS and DFS (Supplementary Table S6). Since DFS benefits were found from EPD treatment for patients who received adjuvant chemotherapy (Supplementary Table S5), subgroup analyses were further performed to evaluate whether different CA19–9 levels could affect the benefits of adjuvant chemotherapy. In the subgroup with preoperative CA19–9 < 200.0 U/mL, a significant improvement in median OS and DFS was found in the EPD group in patients with adjuvant chemotherapy (Supplementary Table S7). However, in the subgroup with preoperative CA19–9 ≥ 200.0 U/mL, EPD showed no advantage regardless of whether patients received adjuvant chemotherapy (Supplementary Table S8). In addition, we conducted a 1:1 propensity matching analysis and further performed survival and subgroup analyses for the treatments after matching. Consistent with our aforementioned findings, no difference was found in the propensity score matching (PSM) population between the treatment groups in both OS and DFS (Supplementary Table S9), while patients with predefined preoperative CA19–9 < 200.0 U/mL showed a better outcome in the EPD group (Supplementary Table S10). In terms of the subgroup analysis in the PSM population, EPD treatment significantly improved the OS in patients with WD & MD (Supplementary Table S11) and DFS in patients with preoperative CEA < 5.00 ng/mL, with resectable stage, with R0 resection or with WD & MD (Supplementary Table S12).

3.6 Prognostic factors

Preoperative CEA ≥ 5.00 ng/mL, preoperative CA19–9 ≥ 200.0 U/mL, PV/SMV resection, borderline resectable, R1 resection, N1 stage, poor differentiation, and absence of adjuvant treatment were identified as adverse prognostic factors for OS in the univariate analysis (ITT population, Supplementary Table S13). When these factors combined with treatment methods were included in the multivariate Cox proportional hazards model, preoperative CA19–9 ≥ 200.0 U/mL, N1 stage and poor differentiation remained indicators for poor OS (Supplementary Table S13).

In the EPD group, the retroperitoneal nerve plexus was removed during the surgery and marked with the exact location for a separate pathological examination. The univariate analysis of OS (qualified population in the EPD group) showed no significant difference in patients with or without nerve invasion in celiac axis plexus, SMA plexus, hepatoduodenal ligament plexus, common hepatic artery plexus or post-pancreatic plexus. In particular, nerve invasion of the aortic plexus suggested poor OS (HR, 0.58; 95% CI, 0.36 to 0.94; P = 0.006; Supplementary Table S14).

3.7 Recurrence pattern and pain intensity

There were no differences in the total recurrence rate between the EPD and SPD groups (Supplementary Table S15). As expected, the locoregional recurrence rate was significantly lower in the EPD group than in the SPD group (16.48% vs. 35.20%; P < 0.001; Supplementary Table S15). No difference was found in the total systemic recurrence rate between the two groups. However, the recurrence rate of mesenteric LNs was significantly higher in the SPD group (10.06% vs. 3.98%; P = 0.022; Supplementary Table S15), while peritoneal seeding was more frequently detected in the EPD group (17.61% vs. 8.94%; P = 0.029; Supplementary Table S15).

Concerning the pain intensity analysis between the EPD and SPD groups, baseline pain intensity was comparable, and no significant difference was found in changes in either upper abdominal pain or back pain intensity at 1 month postoperation from baseline between the two groups (ITT population, Supplementary Table S16). Similarly, no significant differences were found in changes in upper abdominal pain intensity at 6 months postoperation from baseline between the two groups (ITT population, Supplementary Table S17). In particular, at 6 months postoperation, in the EPD group, back pain intensity decreased from 1.02 (SD, 1.83) at baseline to 0.71 (SD, 1.51), while in the SPD group, back pain intensity increased from 1.07 (SD, 1.73) at baseline to 1.43 (SD, 2.42), with a significant change of -0.82 (95% CI, -1.37 to -0.29; P = 0.003; Supplementary Table S17).