METHODS
Study design
We conducted a retrospective analysis of prospectively collected data
from the STELLA-2 randomized controlled trial 23. The
analysis was completed in two phases. In the first phase, we measured
the agreement of sagittal abdominal diameter (SAD) and then selected the
most reliable measurement. In the second phase, we evaluated its
diagnostic accuracy to assess surgical morbidity (Figure 1).
Patients were not involved in the study design (only participating as
study subjects). The study was carried out in three Spanish referral
hospitals: Vall d’Hebron Barcelona Hospital Campus, Hospital
Universitario La Paz, and Hospital General de Valencia. The study was
approved by the Ethics Committee of Hospital Vall d’Hebron (protocol
PR(AMI)168/2015) and by the institutional review boards of the
participating hospitals.
The Guidelines for Reporting Reliability and Agreement Studies (GRRAS)
and the Standards for Reporting Diagnostic accuracy studies (STARD-2015)
were followed in compliance with the Equator Network recommendations.
The present study did not receive any funding.
Subjects
Between 2012 and 2019, 209 patients were enrolled in the STELLA-2 trial,
a randomized multicentre prospective trial comparing the transperitoneal
and extraperitoneal technique for laparoscopic para-aortic
lymphadenectomy in endometrial and early ovarian cancer23. All the subjects for the present study were
selected from that trial.
For phase one, we randomly selected a group of patients with endometrial
cancer to evaluate SAD in MRI images (Group 1). The sample size for this
group was calculated to detect a minimum correlation coefficient of 0.8,
with an α risk of 0.05 and a β risk of 0.05, in a two-sided test.
For phase two we included all patients from the STELLA-2 trial with
high-risk endometrial cancer who underwent comprehensive surgical
staging by minimally invasive surgery (Group 2). High-risk was defined
in the original trial as the presence of any of the following: deep
myometrial invasion (≥50% as elicited by MRI and/or transvaginal
ultrasound) or stromal cervical involvement, grade 3 endometrial
tumours, or non-endometrioid tumours 23. Patients with
missing data were excluded. They were divided into subgroups according
to the para-aortic lymphadenectomy technique and the minimally invasive
approach (Figure 1).
The surgical procedures performed have been previously described24.
Measurements
Preoperative MRI was performed following the European Society of
Gynaecological Oncology (ESGO) and the Spanish society (SEGO)
guidelines, obtaining T1 and T2-weighted 5 mm axial images of the
abdomen and pelvis.
SAD was measured on axial MRI images using the local software available.
Measurements were made manually using the digital callipers in
millimetres (mm).
We defined three anatomical references for SAD measurement (Figure 2a).
Umbilical SAD had been previously described 25. We
chose the left renal vein as a new anatomical landmark since it’s the
superior limit of the para-aortic lymph node dissection3. During this procedure, the inferior mesenteric
artery must also be carefully dissected, so we selected this as another
point of reference. In our experience, these two sites reflect the areas
where the procedure is most challenging and where we encounter the most
complications. Two observers were selected to carry out the SAD
measurements: an experienced radiologist (observer A) and an obstetrics
and gynaecology first-year resident (observer B). They received written
instructions and made two measurements of the three diameters, two weeks
apart.
For inter-rater agreement, we evaluated the concordance between the two
observers’ measurements, whereas for the intra-rater agreement we
compared their first measurements with the ones carried out two weeks
later (Figure 1).
For the second phase, the primary end-point (surgical morbidity) was a
core outcome set defined as the presence of any of the following
criteria: 1) need for blood transfusion, 2) Haematocrit drop
> 90th percentile (>11.8% in our cohort), 3)
Total operative time >90th percentile (>350
min in our cohort), 4) laparoscopic para-aortic lymphadenectomy
operative time >90th percentile (>135 min in
our cohort), 5) Intraoperative surgical complications ≥ grade III26 or during para-aortic lymphadenectomy, 6)
Postoperative surgical complications ≥ grade III 27 or
related to para-aortic lymphadenectomy, 7) uncompleted or converted
laparoscopic para-aortic lymphadenectomy (Table S1).
SAD was measured preoperatively, so observers were unaware of the
outcomes. We performed a multivariate logistic regression analysis
including the following covariates: anthropometric measurements (SAD,
BMI, waist-hip ratio, waist circumference), age-adjusted comorbidity
index 28, tumour characteristics, patients’ age, and
previous surgeries.
The diagnostic accuracy of SAD to predict surgical morbidity was
measured using ROC curves in Group 2 and all subgroups. If the
discriminatory power was adequate, we estimated the optimal cut-off
point (the closest point to the top left corner in the ROC curve) and
calculated sensitivity and specificity, as well as negative and positive
predictive values (NPV and PPV). We also compared the diagnostic
accuracy of SAD with other anthropometric measurements.
Data analysis
In phase one, we calculated Pearson’s correlation coefficient (r )
and traced concordance scatter plots. The agreement was evaluated by
Bland Altman plots 29 and the concordance correlation
coefficient for repeated measurements (ρc)30.
For phase two, a logistic regression analysis was modelled for the
composite outcome and the covariates. DeLong’s test was used to compare
two correlated ROC curves. We computed the area under the curve (AUC)
confidence intervals (95% CI) and considered a clinically appropriate
diagnostic power if AUC was greater than 0.70.
Statistical analysis was carried out using Stata software v13.1
(StataCorp LLC, College Station, TX, USA), R software v. 4.0 (R Core
Team, GNU), and Wizard - Statistics & Analysis v.1.9 (©Evan Miller).
Statistical significance was defined if p <0.05.