2
Tripoloni DE y col. Carta al editor y respuesta. Rev Argent Cir. 2025;117(4):e-cedt
Material and methods
Results
▪
The available literature agrees that the psoas muscle ▪ The authors conclude that “blood albumin level
area should be standardized for sex and age. Crude was associated with anastomotic leak in the
data analysis results in “averaging” female values with colectomy group, although this association was not
male measurements, thereby diluting the differences. statistically significant (OR 0.287; IQR 0.112-0.737).”
Regarding age, higher values typical of young patients We believe that the wording complicates the
would be matched with those of elderly patients. This interpretation of the data, given that:
attenuation of differences reduces the likelihood of c. It would be a conceptual error to use the interval
achieving statistically significant results.
following the odds raꢁo (OR) as an interquarꢁle
range (IQR), as the IQR expresses the dispersion of
numerical data and should not be used with the OR,
which is a measure of associaꢁon or effect.
▪
▪
The authors justify the lack of standardization due
to the absence of anthropometric data necessary
for calculation, but this argument is unacceptable,
as height and weight should have been recorded d. If the two figures (0.112 and 0.737) were a confidence
during the preoperative assessment of nutritional
status.
interval (CI) calculated as part of the logisꢁc analysis
(although with an unclear level of confidence), the
OR would be significant, since the range of values
included in the interval does not contain 1, which is
the “no difference” value (null hypothesis).
It is unlikely that this is a CI, given that the authors
did not report any others. It would therefore be an
interquartile range, but the albumin level range was
recorded in Table 1 as IQR (a single figure obtained
by subtracting the value of the first quartile from the
value of the third quartile), while the paragraph being
interpreted contains two values (presumably those of
the first and third quartiles).
The absence of a defined cutoff point for the
indicators obtained from the computed tomography
images (which, in fact, does not exist for the non-
standardized psoas muscle area) hinders the ability to
determine the proportion of individuals who fall into
the “sarcopenia” category and are at a higher risk of
developing anastomotic leaks.
Since no statistical power analysis was performed,
the probability of a Type II error is unknown. This is
essential for interpreting results without statistical
significance. Although the sample size is not small,
▪
▪
the group of patients with anastomotic leaks is ▪ There is no information on the coefficients or OR
small. This is commendable from a healthcare
perspective, but it reduces the statistical power of the
sample.
calculated by logistic regression analysis for the other
potentially confounding variables.
Insummary,thearꢁclepresentsmethodological
A binary logistic regression analysis is mentioned and staꢁsꢁcal shortcomings that undermine its
among the statistical procedures. Yet, this technique consistency and miss a valuable opportunity to draw
requires a number of events to produce stable valid conclusions from an important healthcare
coefficients, and we have already mentioned that this experience.
was not achieved in the sample by Chinelli et al. Of
the various criteria proposed for calculating sample Daniel E. Tripoloni (ORCID: 0000-0002-8246-5615),
size, the most widely used2 recommends multiplying Dalton A. García López (ORCID: 0000-0002-2269-5777),
the minimum of 10 events (in this case, anastomotic Guadalupe Peón (ORCID: 0009-0004-8150-811X)
leaks) by the number of variables included in the
analysis.
Sanatorio Dr. Julio Mendez. Servicio de Cirugía General
Referencias bibliográficas /References
1
.
Chinelli J, Medina A, Moreira E, Rodríguez G. ¿Es la sarcopenia
un factor de riesgo de dehiscencia anastomóꢁca en cirugía on-
cológica de colon y recto? Rev Argent Cirug. 2025;117(2):e-1833
2. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simula-
ꢁon study of the number of events per variable in logisꢁc regres-
sion analysis. J Clin Epidemiol. 1996;49(12):1373-9. doi: 10.1016/
s0895-4356(96)00236-3. PMID: 8970487
-
hꢂp://dx.doi.org/10.25132/raac.v117.n2.e-1833
Comentario del Director
En relación con la carta dirigida por los Dres.
Tripoloni, García López y Peón, es importante aclarar un
Esꢀmados Dres. Daniel Tripoloni, Dalton García López punto fundamental acerca de las medidas estadísꢁcas
y Guadalupe Peón:
empleadas. No existe un rango intercuarꢀlico para el
odds raꢁo ni para ninguna otra medida del tamaño
Muchas gracias por su lectura del arꢀculo del efecto, tales como la correlación, la d de Cohen,
mencionado y sus interesantes observaciones sobre la el riesgo relaꢁvo, entre otras. Lo que sí se uꢁliza en
metodología estadísꢁca empleada.
estos casos son los intervalos de confianza, que, por