The diagnosis and burden of gastroparesis has exponentially increased over the last decade. A United States Medicare-based data collection on hospitalizations from 1995 to 2004 showed a 158% increase in hospitalizations with gastroparesis as the primary diagnosis over this time period [1]. The increased rate of hospitalizations may be attributed to the introduction of gastric electrical stimulation as treatment for gastroparesis [2]. Increased diagnosis of gastroparesis may also be related to better recognition of this complication in patients with diabetes [3,4]. The true population prevalence of gastroparesis is largely unknown. Most studies have come from large tertiary academic centers or have focused solely on diabetic gastroparesis. These studies estimate the prevalence to be 50-65% among diabetic patients [5-7]. The first large population community-based study was done in Olmsted County, Minnesota. This study estimated impaired gastric emptying to affect 24.2 per 100,000 persons [7].

The outcome and natural history of gastroparesis is also not well understood due to few studies with small numbers and short follow up [1]. One study showed no correlation with delayed gastric emptying and increased mortality after adjustment for comorbidities [5]. Other studies have shown an increase in morbidity, mortality, emergency room use, doctor’s visits and hospitalizations in patients with documented gastroparesis as compared to the general population [7-9]. Patients with uncontrolled gastroparesis are known to have significant impairment in their quality of life as well as significant health care costs [10,11]. Using a validated quality of life questionnaire, a reduction in scores across all domains (physical, emotional, vitality, mental, social, bodily pain, and general health) has been found in patients with gastroparesis [10]. Another study showed poor quality of life was independent of other comorbid factors including age, gender, smoking, alcohol use and type of diabetes [11].

The documentation of delayed gastric emptying can be accomplished by several tests but the gold standard is a scintigraphic gastric emptying (GE) study that demonstrates gastric retention of solids [12,13]. Patient factors including medications, tobacco use, presence of pain, immobility, hyperglycemia, and gender can all affect study outcomes [6,13]. Two broad classes of medication are known to affect gastric motility; prokinetic agents accelerate gastric transit whereas another group of medications slow gastric transit times. Known prokinetic medications include metoclopramide, domperidone, erythromycin and cholinergic medications such as bethanechol [14-20]. Medications known to slow gastric emptying include opioid narcotics, tramadol, tricyclic antidepressants, calcium channel blockers, dopamine agonists, octreotide, anticholinergics, clonidine, and phenothiazines [14,21-33]. Nicotine has also been shown to slow gastric emptying [13,34,35]. Acute changes in blood sugar impact gastric motility. Gastric emptying is slower with hyperglycemia and accelerated during hypoglycemia [6,36-38]. Painful stimuli may also slow gastric emptying as the human body releases noradrenaline in response to painful stimuli which inhibits gastric emptying and tone [34,39].

GE studies are currently performed in both the outpatient and inpatient setting; however, there is no data on the prevalence of positive and negative testing between these two populations. Inpatients are more likely to be immobile, experiencing painful stimuli, and, most importantly, receive medications known to alter gastric motility. Therefore, we hypothesize that inpatient studies will have a higher rate of positive tests. Our aim was to identify the incidence and determinants of abnormal GE studies in the outpatient and inpatient setting.

After obtaining approval from the Institution Review Board of Saint Louis University a retrospective chart review was performed. All scintigraphic GE studies done between June 1, 2012 and April 11, 2014 at Saint Louis University hospital were identified through our radiological database and via CPT codes. The start date was chosen as it coincided with the first day our electronic health record was initiated. Data collected included demographics, details of the gastric emptying study, medications and past medical history. Labs and imaging performed within one year prior to the gastric emptying study date were also collected. Descriptive statistics were used to report findings with categorical variables analyzed using Pearson Chi-Square test and continuous variables analyzed with independent sample t-test. All statistical analyses were performed using IBM SPSS Statistics version 21.0 software [40].

A total of 122 GE studies were identified; 107 (34 inpatient, 73 outpatient) were analyzed. Thirteen studies were excluded because the subjects were less than eighteen years of age and two because the studies were not performed at our institution. Six other studies were deemed incomplete and excluded from the analysis (Two patients vomited prior to completing the study. One patient refused to finish the study. Two studies included either duodenal or esophageal activity. One study was done with tube feeds and only T1/2 was recorded).

42 (22 inpatient, 20 outpatient) tests were positive and 59 (8 inpatient, 51 outpatient) tests were negative for impaired gastric emptying. Inpatient GE testing was abnormal in 73% of the cases as compared to outpatient GE testing at 28% (p<0.01). The study flow is shown in Figure 1. Nearly all, 97%, inpatients received a medication known to alter gastric motility within 72 hours of testing. In comparison, 70% of outpatients were prescribed a medication known to alter gastric motility. 67% of inpatients received opioid narcotics within 72 hours of testing compared to 28% prescribed as an outpatient (p<0.001). 23% of inpatients received tramadol within 72 hours of testing compared to 13% of outpatients (Table 1).

Figure 1: Study flow of inpatient and outpatient gastric emptying testing.

Table 1: Medications – Inpatient vs. Outpatient

Mean patient age was 51 years (64% female, 60% Caucasian). There were more women in the outpatient population (p=0.04). Inpatient GE studies were performed more often in African Americans, while the outpatient population was mostly Caucasian (p=0.002). The mean BMI for inpatients was 25.90 compared to 29.49 for the outpatients (p=0.02). Table 2 highlights demographics of the study. However, when comparing age, sex, race, ethnicity and BMI there was no statistically significant difference between positive and negative tests (data not shown).

Table 2: Demographics – Inpatient vs. Outpatient.

The indications most commonly documented by providers for testing were nausea and/or vomiting (44%) followed by abdominal pain (25%) and early satiety (8%). Similarly, patients reported nausea and/or vomiting (84%), abdominal pain (68%), and gastroesophageal reflux disease (GERD) symptoms (51%) prior to testing. These are patient reported symptoms as documented in admission history or the clinic progress note. Nausea and vomiting were associated with positive tests (p=0.03) and the presence of classic GERD symptoms was associated with negative tests (p=0.01) across both inpatient and outpatient populations (Table 3).

^a Fisher’s Exact test reported due to small cell sizes.
*Significant at p<0.05

Table 3: Patient reported symptoms – Positive vs. Negative Test.

When comparing negative versus positive tests there was no statistical difference between patients with a prior cholecystectomy or total abdominal hysterectomy, however 41% of all patients undergoing GE studies had a hysterectomy prior to testing and 40% of patients had a cholecystectomy.

Patients with diabetes (type 1 or 2) were 2.28 (95% CI 1.00-5.17), times more likely to have a positive test than patients without diabetes. Half of the abnormal tests were found in patients with diabetes.

There was no association between duration of diabetes and increasing prevalence of positive tests. The frequency of all other disease states was too infrequent to assess any association with positive vs. negative tests (Table 4).

^a Fisher’s Exact test reported due to small cell sizes.
*Significant at p<0.05

Table 4: PMH and PSH – Positive vs. Negative Test.

The majority of diabetic subjects (72%) had a hemoglobin A1c recorded and the average hemoglobin A1c was 7.9%. There was no association between increased A1c levels and increased prevalence of positive tests. Nearly all the inpatients (93%) had a morning fasting glucose documented on the day of testing. The mean glucose was 131 with a maximum of 267 and a minimum of 55. Subjects with a positive GE test had statistically significant lower mean albumin level (3.38) compared to those with a negative test (3.81), p=0.007.

The impact admission status has on GE studies had not been reported in the literature previously. Our retrospective study found a statistically significant association between an increase in positive tests and inpatient status. This may be due to the effect of medications received as an inpatient as nearly all inpatients (97%) received medications known to alter gastric motility and a majority received opioid narcotics within 72 hours of testing. In addition, there was a statistically significant difference in inpatient (66.7%) versus outpatient (28.2%) opioid use (p<0.001). Given opioids strong anti-motility properties this may play a key role in the increased rate of positive tests in the inpatient setting.

Pro-motility agents were given to 56.7% of inpatients prior to testing whereas clinical records indicated just 16.9% of outpatients had previously been prescribed such medication (p<0.001). Metoclopramide was the most commonly prescribed pro-motility medication (88% of the inpatients and 83% of the outpatients). The most common indication for admission was nausea and/or vomiting and metoclopramide is often used for its anti-emetic properties in the inpatient setting as well as empirically to treat presumed gastroparesis. Despite the high number of patients receiving such medications there was still a very high number of abnormal gastric emptying studies in the inpatient population. A possible explanation for this could be that the combination of opioid medications, immobility and general pain/discomfort these inpatients are experiencing is a stronger factor in modifying gastric motility then metoclopramide. In the outpatient setting the use of metoclopramide as a therapeutic agent for impaired gastric motility might be successful, therefore decreasing the number of positive tests seen in the outpatient setting.

The correlation between subjective symptoms and objective findings of impaired gastric motility is poor and symptoms are generally poor markers of gastroparesis [41,42]. However, in our study there was a statistically significant association between an increased risk of a positive test and the symptoms of nausea and/or vomiting (p=0.03). Nausea and vomiting symptoms were also the most common indication for testing (44%) and the most common patient reported symptoms (84%).

Current guidelines recommend that patient’s with classic symptoms of GERD undergo a trial of proton pump inhibitors as an initial diagnostic and therapeutic intervention [43]. In our study, classic GERD symptoms were associated with a negative test (p=0.01). This finding seems to validate current guidelines, thus illustrating the need for thorough history taking and thoughtful decisions on differential diagnosis and further testing.

A high percentage of patients had either a prior cholecystectomy (40%) or total abdominal hysterectomy (41%) prior to testing for gastroparesis. Given the relative difficulty in making a diagnosis of gastroparesis due to poor correlation between clinical symptoms and documented gastroparesis, the high prevalence of abdominal surgeries may be related to misdiagnosis. This can be seen in other abdominal pain syndromes. For example, the numbers of abdominal and pelvic surgeries are disproportionately high despite normal pathology and histology in surgically resected tissue in IBS patients [44]. In the setting of visceral pain syndromes, patients may be inappropriately exposed to procedures that place unnecessary financial burdens, with increased morbidity and mortality.

Not surprisingly, patients with diabetes mellitus were more likely to have a positive test than patients without diabetes. In addition, the majority of the abnormal tests were found in patients with diabetes. Longer duration of disease and poorer glycemic control are potential risk factors for developing gastroparesis [12]. However, in our study these associations were not seen. Because of the retrospective study, we had limited data on glycemic control and duration of diabetes mellitus.

As stated earlier, elevated blood glucose at the time of testing can slow gastric motility. In our study nearly all the inpatients (93%) had a fasting morning glucose documented on the day of testing. They were all fairly well controlled as the mean glucose was 131 with a maximum of 267 and a minimum of 55. Given this data, it is unlikely that elevated glucose levels are the driving force behind the increased number of positive tests seen in the inpatient population.

As albumin is a marker of general nutrition, it is not surprising that patients with delayed gastric emptying have significantly lower markers of nutrition. In fact, one prospective study of 146 patients with gastroparesis showed that as many as 21% required nutritional support with either a feeding tube or parenteral nutrition [8].

This study had several limitations inherent to a retrospective chart review. Data was collected using electronic medical records retrospectively through charts subject to incomplete records and missing data. Although there was a statistically significant increase in positive studies done as an inpatient and many of these patients received medication known to alter gastric motility, this only establishes association and not causality. In addition, our patient population was comprised of African Americans and Caucasians which limits the ability to generalize our findings to the general population.

Our documentation of outpatient medications was limited to clinic visits closest to the gastric emptying study as our radiology department does not collect medication history at the time of the study. Although this is a limitation of our study, it also provides an avenue for improving patient care and physician practices. Guidelines suggest that ordering providers counsel patients on the discontinuation of appropriate medications prior to testing [45]. In addition, documentation that these medications have been stopped prior to testing regardless of the study location (inpatient or outpatient) is imperative to ensuring the validity of the GE study as symptoms of gastroparesis are non-specific and may correlate poorly with documented gastroparesis. However, providers may forget to counsel patients and/or patients may forget to discontinue medications. Creating an intake form to be completed by the patient prior to testing will allow the physicians to more accurately interpret the study and will provide valuable information to the physician when deciding on a diagnosis and treatment plan.

Optimizing diagnosis and management of patients with gastroparesis is further heightened as it appears that the prevalence of gastroparesis is increasing with significant impact on morbidity, mortality, quality of life and an increased burden on health care systems [1,46]. It is well documented that certain patient factors can greatly affect gastric motility and many of these factors are different between inpatient and outpatient populations. Our study demonstrates a statistically significant increased proportion of abnormal GE studies in inpatient studies when comparing inpatient to outpatient studies. However, a majority of these inpatient studies are being done with patients who have received medications known to alter gastric motility and therefore may be at an increased risk for false positives. Data from this study may help to improve the sensitivity and specificity of GE studies.