RESULTS
Phase one: Reliability assessment
We analysed the measurements from 15 patients (Group 1), including 30
observations for inter-rater variability and another 30 for intra-rater
variability (Figure 1). The measurements from all three anatomical
locations showed good correlation, agreement, and concordance.
We found a strong linear correlation between the two observers for all
three readings (Figure 2b), and SAD at the umbilicus showed the best
inter-rater concordance (ρc = 0.96, 95% CI 0.94–0.98)
(Table S2). Bland-Altman plots also show that umbilical SAD had the best
agreement (Figure 2c).
The first and second measurements of the observers were nearly
identical. There was a strong linear correlation for all three
measurements (r > 0.98). Concordance and
Bland-Altman plots revealed a high agreement for all sagittal abdominal
diameters (ρc>0.98) (Table S3).
Given the superior reliability of umbilical SAD, we tested its
diagnostic accuracy to predict surgical morbidity in group 2.
Phase two: Evaluation of diagnostic accuracy and subgroup
analysis
Thirty-five patients (20.4%) were excluded because of missing data. A
total of 136 patients was analysed (Figure 1).
We found that umbilical SAD (u-SAD) was normally distributed in our
sample (mean 239 ± 40 mm SD) (Table 1). Patients who encountered
surgical morbidity had a significantly higher u-SAD compared with those
without surgical morbidity (mean 243 ± 39 mm vs. 230 ± 36 mm,p =0.03).
In the univariate analysis, only the waist-hip ratio (WHR) and u-SAD
showed a significant association with the main outcome measure
(p <0.05). Despite these findings, when we assessed
their diagnostic accuracy as a single measurement, neither of them
proved to be clinically useful to predict surgical morbidity
(AUC<0.70).
We found that nearly half of intraoperative complications occurred
during the para-aortic lymph node dissection (6/14). A higher
measurement of u-SAD was significantly associated with an incomplete
lymphadenectomy (p <0.001). We also found a strong
positive correlation between the conversion rate and the u-SAD
measurement, although this was not statistically significant (Figure
S1). The rest of the surgical morbidity variables were not independently
associated with the u-SAD measurement.
The vast majority of patients in group 2 (Table S4) underwent a
hysterectomy, bilateral salpingo-oophorectomy, pelvic and para-aortic
lymphadenectomy (85.3%); only eight patients underwent exclusively
para-aortic lymphadenectomy. The most frequent histopathology was
endometrioid (53.7%) followed by serous (21.3%) neoplasia. Previous
surgery, either open or laparoscopic, was not associated with the
primary end-point. The median follow-up was 31 months (IQR 15–52).
We observed 17 deaths, but only two of them were grade V operative
complications (both within 30 days of the surgery). One of them was
directly associated with the procedure (bowel perforation), and the
other one was a patient who suffered a stroke 5 days after surgery.
Thirteen patients died from disease progression and the remaining three
from other metastatic malignancies (breast cancer and multiple endocrine
neoplasia).
All subgroups were uniformly distributed (Figure 1), even though
patients were originally randomized only to the para-aortic
lymphadenectomy technique.
Intraoperative complications were more frequent in the transperitoneal
vs. extraperitoneal subgroup (9 vs. 5), but this difference was not
statistically significant (p =0.173). We found a significant
negative correlation between u-SAD and aortic lymph node yield in the
transperitoneal group (p <0.001), but no correlation was
observed in the extraperitoneal subgroup (p =0.115) (Figure S2).
In the transperitoneal subgroup, BMI and u-SAD showed a significant
independent association with the primary end-point
(p <0.05). In this same subgroup, u-SAD was
significantly higher in patients who encountered surgical morbidity
(Figure 3a). In contrast, in the extraperitoneal subgroup, the main
outcome was independent of the u-SAD measurement (Figure 3b).
We compared the diagnostic accuracy of u-SAD with the other
anthropometric measurements of obesity in the transperitoneal subgroup,
and u-SAD outperformed the others (Figure 3c). The discriminatory power
was good in the transperitoneal subgroup, but poor in the
extraperitoneal one (Figure 3d). In the subgroup of patients who
underwent transperitoneal para-aortic lymphadenectomy by conventional
laparoscopy, the diagnostic accuracy was higher (n=34, AUC=0.78).
In patients who underwent staging by the extraperitoneal technique,
surgical morbidity was more frequently observed in those operated by
conventional vs. robotic-assisted laparoscopy (52.6% vs. 28%,p =0.05) (Table S4).
Optimal cut-off point estimation
For patients undergoing para-aortic lymphadenectomy by the
transperitoneal technique, we determined that 246 mm was the optimal
cut-off point for u-SAD as a predictor of surgical morbidity. Using this
value, the sensitivity and specificity were 0.56 and 0.80 respectively
(NPV 0.69, PPV 0.70).