Our results revealed that the stomach position evaluated by CT obtained before the initial procedure was associated with success or failure of the first placement of the post-pyloric EFT by the blind method. The best cut-off value for failure at first attempt of EFT placement, estimated using CT imaging, was the greater curvature lower than spinal level L2–L3.

A previous study [12] used abdominal radiographs after EFT placement to estimate the position of the stomach; however, this method was not clinically effective enough to recognize the difficulty before starting blind placement. Moreover, the position of the stomach estimated using abdominal radiography after EFT placement could have been modified by air injection or stretching of the stomach wall during the placement procedure. The strength of this study is that the effect of stomach position was evaluated using CT images taken before the EFT placement, and these images were not modified by the EFT placement procedure. We demonstrated the association between difficulty at the first attempt of EFT placement and lower stomach position, reaffirmed the hypothesis of a previous study [12].

The clinical application of our findings is influenced by the frequency with which available CT images are present at the time of EFT placement. At the time of screening before collecting data, approximately 90% of patients who underwent EFT had available CT images. Considering the large number of CT scanners per population in our country [16], it may be the lower frequency with available CT images at the EFT placement in our study than in other countries. However, the fact that this study showed an association between CT images and successful post-pyloric EFT placement even though it included CT images acquired well before placement (i.e., not just immediately before) demonstrates the value through wider potential application of our findings.

In patients in whom EFT placement was successfully achieved, the total number of attempts was not substantially influenced by the low stomach position in secondary outcomes (Table 5). Therefore, it remains controversial whether methods other than blind placement of the EFT should be performed on the first attempt when the low stomach position was detected before the first attempt. Of the 328 patients who eventually underwent successful post-pyloric EFT placement, 308 (93.3%) had successful outcomes by the third attempt (Table 3). This information may help make a clinical decision to apply other assistive methods such as fluoroscopy or gastroscopy.

We hypothesized that the length and angle could be alternative indicators of the lowest position of the greater curvature relative to the spinal level. The length and angle had a relatively high correlation with the greater curvature relative to the spine estimated by CT (Table 3). The post-hoc logistic regression analysis using length or angle instead of stomach position revealed that difficulty in post-pyloric EFT placement was associated with length but not angle (see Additional files 3 and 4). Assessing the lowest position of the greater curvature relative to the spinal level as estimated using CT images may be a slightly complicated procedure. Spinal deformities, such as scoliosis of the spine, compression fracture, or lumbosacral transitional vertebra [17], may affect the estimation of stomach position using the spinal level. Therefore, the length, which is easier to measure, may be a more practical and objective indicator for daily clinical use. The angle was more likely to predict success because the steep angle resulting from the caudal extension of the stomach by the EFT placement may be expected to contribute to the difficulty in guiding the EFT tip to the pylorus. However, the angle calculated from the stomach at the time of CT imaging may not have been able to reflect the change in the shape of the stomach due to the EFT placement.

In our study, physician inexperience was associated with the success of EFT placement. The reasons for this result, which was contrary to the generally expected effects, could be speculated as follows. First, supervisors in our institution might have been more responsible than residents for performing the EFT placement on patients with more serious illnesses. These patients often had risk factors for upper gastrointestinal hypomotility and were often admitted at night or on holidays when human resources were insufficient, making it difficult to provide adequate time for EFT placement. Residents tended to be assigned to patients whose general condition was relatively stable or when there was an available time for placement. Second, the nursing care record had a significant number of missing data related to physicians’ experiences. The exact effect of the physician’s experience could not be determined due to the retrospective nature of the study design. A sensitivity analysis of the primary outcome that did not include the physician’s experience as a covariate showed that the effect of stomach position was consistent (see Additional file 5).

Our study revealed that opioid use was associated with difficulty in EFT placement. Opioids are one of the factors that affect gastrointestinal function in critically ill patients. The effects of morphine on upper gastrointestinal motility, including enhanced relaxation of the proximal stomach, increased pyloric tone, and retrograde duodenal contractions [18,19,20]. In enteral nutrition, fentanyl was associated with increased volume of gastric aspiration and upper digestive intolerance [21]. Because opioids are major agents for pain management in the ICU setting and are only one of the many factors involved in gastrointestinal function, the clinical intervention that withholds using opioids solely for EFT placement cannot be easily implemented.

First, this single-center study was performed in an urban educational university hospital in Japan. Whether other countries or institutions can provide similar results is unknown. Furthermore, the effects of unidentified confounders were not taken into account because the study was retrospective. Second, the data of patients who did not undergo CT before ICU admission were not included. This may include selection bias regarding emergency patients who could not afford CT scans or patients with central nervous system disease, stroke, or cerebral hemorrhage, which might have been infrequent for abdominal CT scans. Third, our study could not collect data about the time required for EFT placement, which was one of the indicators of difficulty, because it was not mentioned in the medical records.