Dysphagia is a common complication following stroke and traumatic brain injury (TBI), with reported incidence rates ranging from 27% to 80% [1,2]. Patients with these conditions often require nasogastric tube (NGT) feeding to maintain nutritional supply in acute stage and are frequently encountered in rehabilitation settings. Dysphagia can lead to malnutrition, aspiration pneumonia, prolong hospitalization, and increase the risk of morbidity and mortality [3,4]. NGT feeding may suffice for patients requiring nutritional support for up to 4 weeks [5]. Extended utilization of an NGT can lead to various adverse events, such as nasal wing lesions, chronic sinusitis, gastroesophageal reflux, gastrointestinal bleeding, and aspiration pneumonia [6-8]. Therefore, for extended periods, the placement of a gastrostomy tube should be taken into consideration to avoid the side effects of long-term usage of NGT for nutrition supply [5].

Bedside swallowing assessments can quickly detect dysphagia, but predicting its recovery is challenging. Early identification of dysphagia recovery prognosis would enable prompt administration of appropriate therapies and consideration of suitable feeding methods such as NGT or gastrostomy tube. Preexisting studies have found that patients with high National Institutes of Health Stroke Scale (NIHSS) scores, aspiration during swallowing assessments, advanced age, dysarthria, and intubation are at a higher risk of experiencing prolonged dysphagia lasting for more than 6 weeks [9-11]. However, no work has mentioned the exact percentage of failure of complete oral feeding through the above mentioned parameters. As a result, our work attempts to develop a model capable of predicting the unsuccessful removal of NGT in dysphagia patients who have experienced a stroke or TBI. Additionally, the model may provide the information for the early intervention of gastrostomy tube.

From January 2020 to December 2022, we retrospectively analyzed clinical data of patients admitted to the rehabilitation department in a regional hospital, who experienced stroke or TBI with dysphagia and requiring NGT feeding. A total of 203 patients, comprising 133 males and 70 females, were enrolled in the study, and their NGT removal failure rates were tracked over a six-month period. Baseline demographic characteristic of age, sex, admission diagnosis (ischemic stroke, hemorrhagic stroke, or TBI), initial Glasgow Coma Scale (GCS), body mass index (BMI), Functional Independence Measure (FIM) score, level of activities of daily living (ADLs) dependence were recorded. We define a patient as ADLs total dependent when they are completely unable to perform four aspects: personal hygiene, heavy hygiene (toilet and bathing), eating, and dressing. If the criteria mentioned above are not met, the patient is referred to as ADLs partial dependent. Clinical variables were correlated after one month of swallowing training, including the improvement in Functional Oral Intake Scale (FOIS) levels and variation in the numbers of food type according to International Dysphagia Diet Standardization Initiative (IDDSI). The swallowing training consists of direct method such as modifying food texture, utilizing chin tuck and head turn/tilt compensatory techniques, and indirect method such as applying thermal tactile stimulation, practicing pharyngeal wall contraction exercises (Masako maneuver) and practicing effortful swallowing, according to the evaluation of swallowing condition of the patients. We utilized both videofluoroscopic swallowing study, fiberoptic endoscopic evaluation of swallowing, or bedside clinical swallow assessment by physicians to evaluate patients’ swallowing function. Simultaneously, we integrate the assessment conducted by the speech-language therapist to determine the optimal timing for NGT removal.

This study received approval from the Research Ethics Committee of Hualien Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation (approval number: IRB112-206-B). Due to its retrospective nature, the requirement for informed consent was waived.

SAS Software version 9.4 (SAS Institute) was used for statistical analysis. Exclude patients with missing data and conduct individual subgroup analyses on the available data, resulting in variations in the total number of patients for different variables. The categorical variables were demonstrated as counts and percentages. The continuous variables were presented. Comparisons of these clinical data between each group were made by using the chi-square test for categorical variables and the two sample t-test for continuous variables. A stepwise logistic regression analysis was used to develop a model for predicting failure of NGT removal. A p-value of less than 0.05 was considered statistically significant.

A total of 203 patients were included in the study, of which 150 (73.9%) had their NGT successfully removed, while 53 (26.1%) experienced removal failure after six months of follow-up. Table 1 provides an overview of the subjects’ demographic and clinical variables. Univariate analysis demonstrates a significant correlation between age, admission diagnosis, BMI, FIM score, and ADLs dependence and the failure of NGT removal. The age stratification involves incorporating patients’ age groups into the logistic regression analysis to obtain odds ratios (ORs), and both the age groups of 60–70 and over 70 were found to be significant variables. Using a stepwise regression analysis, we examined the effect of these significant variables. Age between 60–70 years (OR, 3.44; p<0.05), age over 70 years (OR, 6.16; p<0.05), underweight BMI (OR, 5.09; p<0.05), total dependence in ADLs (OR, 9.90; p<0.05), and ischemic stroke (OR, 4.10; p<0.05) were found to be the strongest predictors for failed removal (Table 2). According to the OR values of significant variables, directly rounding to the nearest integer to convert to weight scores, for example: the OR for low BMI was 5.09, then its weight score would be 5. Using the weight scores derived from the predictive model for tube removal failure, the total weight score for tube removal failure in patients is calculated. Analysis of the OR distribution derived from spline regression reveals three distinct risk trends within the graphic representation (Fig. 1). Further categorization stratifies risk into groups scoring 0–6, 7–16, and 17 points or more. Following these groupings, logistic regression is utilized to estimate the ORs prior to the adjustment of other variables (Table 3). The failure rates of NGT removal were 5.1% in the low-risk group, 21.5% in the moderate-risk group, and 73.0% in the high-risk group (Table 3).

After swallowing training in high-risk and low-risk populations, the degree of improvement in FOIS and IDDSI does not impact the risk of failure in NGT removal as much as it does in the moderate-risk population. Therefore, we primarily focused on the moderate-risk population for further analysis. As shown in Table 4, for patients who did not show improvement in FOIS levels after training, the failure rate of NGT removal was 81.2% six months later. For those who improved by one FOIS level, the failure rate was 50.0%, while for those who improved by two FOIS levels, the failure rate was 37.5%. For patients whose IDDSI food types did not change after training, the failure rate of tube removal was 85.7%. However, for those who improved by one type of food, the failure rate was 10.0%, and for those who improved by two types of food, the failure rate was 21.4%. In the moderate-risk group, improvements in FOIS levels and an increase in IDDSI food types after one month of swallowing training may be used to predict the likelihood of NGT removal failure after six months of follow-up observation. We amplified our model sample using bootstrapping by 1,000 times for internal validation and obtained the area under the receiver operating characteristic (ROC) curve. The area under the ROC curve is 0.829. The sensitivity and specificity were 75.5% and 23.3%, respectively. In the Fig. 2, the blue line represents our model, while the dashed line represents the results obtained through bootstrapping. The Fig. 2 shows that the two lines almost overlap (DeLong test p-value=0.9831).

In this work, we identified several strong predictors for failed remove of NGT after six months of follow-up, including age between 60–70 years, age over 70 years, underweight BMI, total dependence in ADLs, and ischemic stroke. We also utilized the weighted scores of the ORs to create a formula to forecast the risk of NGT removal failure and found the risk score greater than 17 have 73.0% of failure of removing NGT.

For patients requiring nutritional support for up to 30 days, NGT feeding alone may be sufficient, but for longer durations, consideration should be given to percutaneous endoscopic gastrostomy (PEG) tube placement [5]. Compared to NGT, PEG offers several advantages, including a lower risk of intervention failure, fewer incidences of pneumonia among PEG recipients, improved quality of life, and higher levels of albumin [8]. Currently, PEG has become a recognized treatment option for temporary or permanent dysphagia resulting from neurological disorders and oncological conditions [12]. Implementing continuous enteral nutrition at an earlier stage holds the potential to significantly enhance prognosis [12]. Additionally, with early gastrostomy placement, the prevalence of low albumin levels and higher comorbidity, both of which are risk factors for worse outcomes, would be reduced in patients chosen for this intervention [12]. Nevertheless, the benefits of PEG feeding remain uncertain in patient with advanced dementia, and elderly patients over 80 years old [5].

With this result, we propose the following recommendations for dysphagia patients with brain insult who require tube feeding for nutrition support: those identified as high risk based on the research results could early decide whether to undergo or not. Patients with moderate risk should evaluate improvements in FOIS and IDDSI scores after the first month of swallowing training. If there is no improvement observed, these patients may benefit from early decision of further management, considering the high failure rates of 81.2% and 85.7% for the FOIS and IDDSI non-improvement groups, respectively, as indicated in Table 4. Additionally, when suggesting PEG placement, it is advisable to take into account factors such as the presence of advanced dementia and whether the patient is over the age of 80, as per our model’s recommendations.

In our study, we observed that 26.1% of the patients failure of removing the NGT after six months follow up. This finding aligns with previous research, which reported removal failure rates ranging from 26% to 69% [13-15]. In spite of antecedent studies investigating the correlation between various risk factors and NGT removal in stroke patients [16,17], our study provides processional data for predicting the risk of failure of removing. Moreover, preceding research has highlighted that swallowing difficulties resulting from TBI are comparable to those observed in stroke patients [1]. Therefore, we can do early decision of alternative management of NGT feeding for the patients with dysphagia such as insertion of gastrostomy, specifically focused on patients with stroke or TBI.

The parameters in our study consistent with preexisting research findings are age, sex, initial GCS and ADLs dependence. Former research has shown that older patients and those with lower FIM scores have a reduced likelihood of NGT removal [14,15]. A lower FIM score indicates that subject has a higher degree of dependency in their daily activities. In our investigation, we also noticed significant differences in age and ADLs dependence between the nonweaned group and the weaned group. The non-weaned group comprised individuals who are older and a larger number of individuals who are completely dependent on ADLs.

Previous studies demonstrated methodological variations and reported inconsistent associations between sex and oral intake recovery [3,16,18]. Certain studies [16,18] excluded sex as a variable in their multivariable analyses due to the absence of significant differences in bivariate analysis. One study indicated that female sex was associated with better outcomes [19]. In our study, there was no significant difference in sex between the NGT-weaned group and the NGT non-weaned group. Further research is needed to explore its impact on the recovery of swallowing disorders.

In individuals with TBI, a low initial GCS consistently indicated compromised oral intake [20]. Additionally, Morgan et al. [21] noted that low GCS scores were predictive of dysphagia in pediatric TBI patients. Another study found that among stroke patients, the deterioration in verbal response, assessed using the GCS, demonstrated a notably stronger association with swallowing recovery failure [22]. However, the statistical significance of the GCS variable in the present study may be influenced by the sample size. Numerous studies [9,10,11,23] in this domain have demonstrated a link between NIHSS scores and unfavorable recovery outcomes. One study found that higher NIHSS scores were associated with delayed removal times of NGT [14]. Nonetheless, while the NIHSS is mainly employed for prognostic assessment in post-stroke patients, the GCS has broader applications, encompassing prognostic evaluation in brain injury patients and even the assessment of recovery in swallowing disorders, in addition to its use with stroke patients.

A study proposed in 2021 that there is no correlation between body weight and NGT removal [17], which differs from our research findings. Our study highlighted the significant impact of being underweight on the failure of NGT removal. A BMI below 20 is widely acknowledged as an indicator of malnutrition, and a low BMI is correlated with a higher probability of probable sarcopenia [24,25].

Previous studies have also indicated a link between malnutrition and dysphagia [26], and there is evidence of a relationship between sarcopenia and decreased swallowing function, particularly among older adults in Japan [27]. This finding may inferred that malnutrition and sarcopenia may be risk factors for the poor recovery of swallowing function in individuals who have experienced a stroke or brain injury. The decline in overall muscle mass and strength that accompanies being underweight may contribute to the weakening of the muscles involved in the swallowing process [28]. This weakened musculature could explain the increased likelihood of NGT removal failure observed in underweight individuals in our study. Furthermore, a separate study demonstrated that stroke survivors with lower BMI values, particularly those in the lowest quartiles, exhibited more severe stroke symptoms and experienced poorer functional recovery compared to individuals with higher BMI values [29]. These studies further support the association between underweight and greater severity of stroke, poorer functional recovery, and even swallowing function.

There are limited study mentioning the correlation between stroke types and the failure rate of NGT removal [17,30], and only one study identified intracerebral hemorrhage as a predictive factor for NGT removal [31]. Our study revealed a significant association between ischemic stroke and the failure rate of tube removal. Furthermore, another study published in Stroke provided additional insights by demonstrating that patients with cerebral infarction had poorer functional and neurological recovery compared to those with hemorrhagic stroke [32]. The relationship between stroke type and removal of NGT needs further study.

A retrospective study revealed that for each increase of one FOIS level, there was a 3.7-fold higher likelihood of tube removal [33]. Consistently, our study yielded similar results, indicating that the extent of FOIS level improvement after one month of swallowing training could predict the likelihood of NGT removal failure six months later. Patients who exhibited no change in FOIS levels following the initial month of swallowing training after experiencing a brain insult had the highest rate of tube removal failure, whereas those who demonstrated greater improvements had lower failure rates. Additionally, another study found that higher FOIS levels were correlated with a higher probability of returning to complete oral intake post-stroke [34]. Furthermore, our study identified that the greater improvement in the ability to consume various food items (according to IDDSI) after one month of swallowing training, the lower the likelihood of NGT removal failure six months after the brain insult. Although one study mentioned that FOIS at admission could predict the time to return to unrestricted oral intake [35], no previous study has evaluated the probability of NGT removal failure six months after brain insult using changes in FOIS and IDDSI following one month of swallowing training, which is a relatively unique feature of our predictive model.

Using our predictive model, patients with brain insults can be initially categorized into risk groups for swallowing disorders. Subsequently, advanced swallowing interventions can be implemented for high-risk patients who encounter difficulties in NGT removal. These interventions may include the combined use of repetitive transcranial magnetic stimulation or VitalStim, long-term education on NGT care, or alternative options for enteral feeding. For individuals at moderate risk, a progress assessment of NGT removal rate using FOIS and IDDSI can be conducted one month later. If there is no improvement, rehabilitation training similar to that for high-risk cases should be followed, while others can undergo regular swallowing rehabilitation training.

One study utilized FIM-motor, FIM-cognition, days after onset, and age to construct the equation [15]; the other utilized the Barthel index at admission, lip-closing status, ability to answer simple questions, and functional independence before stroke to construct the equation [17]. In contrast, ours utilizes patient age, stroke history, BMI, and total dependence in ADLs to form a risk prediction model, and provides early intervention recommendations. Our risk factors are relatively simple and applicable for predicting NGT removal in both stroke and TBI patients.

There are certain limitations in our study that need to be taken into account when interpreting these findings. Firstly, it is important to note that this study is retrospective in nature, which means we are unable to control for participant consistency. Secondly, due to the retrospective nature of this research, there may be instances of missing and incomplete clinical data. Thirdly, we faced the challenge of not being able to control the intervention methods applied to participants. Not all individuals in the study received the same swallowing training, leading to variations in the methods used. Fourthly, we did not actually have the patients undergo nutritional assessments or tests for sarcopenia. These constraints highlight the need for future research with more controlled settings and comprehensive data collection to further validate and refine our findings.

In conclusion, our study identified factors predicting NGT removal failure, including older age, underweight, complete ADLs dependence, and ischemic stroke. We also created a predictive model to assess this risk, aiding healthcare professionals in identifying high-risk patients and providing suitable interventions and training.