Baseline features and differences in 48 week clinical outcomes in patients with gastroparesis and type 1 vs type 2 diabetes.

Diabetic gastroparesis is associated with nausea, vomiting, fullness, bloating, early satiety, and epigastric discomfort/pain and is diagnosed by documenting delayed gastric emptying ( 1 , 2 , 3 , 4 , 5 ). However, emptying delays correlate poorly with symptoms, suggesting other pathogenic factors influence symptoms. These factors include: (i) chronic hyperglycemia, which acutely impairs gastric neuromuscular function; (ii) gastric factors ranging from impaired fundic accommodation and gastric electrical dysrhythmias; and (iii) psychological dysfunction, which is prevalent in diabetic gastroparesis ( 6 , 7 , 8 , 9 , 10 , 11 , 12 ). Gastroparesis is thought to contribute to poor glycemic control which results in ketoacidosis and other complications that increase hospitalizations and outpatient visits and costs ( 13 , 14 ). Longitudinal studies suggest diabetic gastroparesis follows an indolent course with stable gastrointestinal (GI) symptoms and emptying rates over 25 years, although increased mortality has been reported ( 15 , 16 , 17 ). Furthermore, a recently published study observed no differences in overall symptom improvements over 48 weeks in patients with diabetic versus idiopathic gastroparesis ( 18 ). Type 1 diabetes (T1DM) from failed insulin production is distinct from Type 2 disease (T2DM) which is due to insulin resistance and variable insulin release deficits ( 19 , 20 ). T1DM requires insulin therapy, while T2DM is managed with diet and oral medications in milder cases and insulin in more severe cases. Gastroparesis is reported in 27–58% of T1DM patients versus 20–40% with T2DM; the 10-year incidence of gastroparesis is five times higher with T1DM (5.2% vs. 1.0%)( 21 , 22 , 23 , 24 ). Gastroparesis is associated with increased hemoglobin A1c (HbA1c) levels and diabetic complications (retinopathy, neuropathy) in T1DM, while obesity status has been associated with symptoms in T2DM with gastroparesis ( 25 , 26 ). Comprehensive comparisons of clinical profiles, comorbidities, disease severity, resource utilization, psychological dysfunction, quality of life, and clinical courses in patients with gastroparesis and T1DM versus T2DM have not been performed. Our aim was to compare the clinical features of patients with gastroparesis and T1DM and T2DM at baseline enrollment into the NIDDK Gastroparesis Registry and after 48 weeks of follow-up during which time the patients’ GI symptoms were managed by gastroenterologists at tertiary centers. We hypothesized that patients with T1DM gastroparesis at baseline have (i) more severe GI symptoms, (ii) more severely delayed gastric emptying, (iii) poorer glycemic control, (iv) more peripheral neuropathy, (v) more healthcare utilization, and (vi) more impaired psychological dysfunction and quality of life compared with patients with T2DM and gastroparesis. We further hypothesized that symptoms, psychological function, and quality of life would show similar longitudinal changes in both subtypes after 48 weeks of management.

Demographic and clinical factors and comorbidities for the T1DM and T2DM patients with gastroparesis at baseline are shown in Table 1 . T2DM patients with gastroparesis had several expected differences compared with T1DM patients. These T2DM patients were older at enrollment and at the onset of GI symptoms, had higher BMIs, and were more often overweight, obese, or postmenopausal (P<0.001). T1DM patients reported longer durations of diabetes prior to the onset of gastroparesis (P=0.005). On average, HbA1c levels were greater in T1DM patients by 0.9% (P=0.003); T1DM patients comprised larger proportions with HbA1c levels ≥8% (37/53, 69.8%) versus <8% (38/81, 46.9%)(P=0.009). Almost all T1DM (98.7%) and T2DM (98.3%) patients reported ≥1 comorbidity, but numbers of comorbidities were higher with T2DM (5.5±3.4 vs. 4.0±2.8)(P=0.005). Coronary and cerebrovascular disease and interstitial cystitis were significantly more common in T2DM patients (P≤0.05). T2DM patients more often underwent hysterectomies (P<0.001). Similar percentages of T1DM versus T2DM patients reported peripheral neuropathy (43.6% vs. 37.3%, P=0.46). The most severe symptoms reported by both T1DM and T2DM patients with gastroparesis were nausea and postprandial fullness. Patients with T1DM and gastroparesis were more often assigned by the investigator to the severe gastroparesis category (49% vs. 39%) and less to the mild category (6% vs. 16%) compared with T2DM patients (P=0.05)( Table 2A ). However, patient-rated overall GCSI scores were similar in T1DM and T2DM patients (2.8±1.1 vs. 3.0±1.0, P=0.28). Individual GI symptoms were also similar, except for higher bloating symptoms in T2DM patients (P=0.04). More T1DM patients had delayed emptying at two hours (P=0.006) and four hours (P35% 4 hour retention) compared with T2DM patients (54% vs. 32%, P=0.001). Medication use from prokinetics to opiates was similar in the two subtypes, except T2DM patients had more metformin use (P<0.001). Nineteen of 59 T2DM gastroparetics (32%) used metformin on enrollment; none were on other antihyperglycemic agents (exenatide, liraglutide, pramlintide) that cause nausea. Overall, GI symptoms were similar in the T2DM patients who were taking metformin versus those patients not taking metformin (P=0.91), although vomiting scores on average were lower in the group receiving metformin at baseline by 1.0 point (P=0.04)( Supplemental Table 1 ). T1DM patients reported more hospitalizations in the year before enrollment solely for gastroparesis (5.1±6.4 vs. 3.2±6.6, P=0.003), and were hospitalized more often for nausea and vomiting (P=0.001), abdominal pain (P=0.003), and dehydration (P=0.01) compared with T2DM patients ( Table 2A ). TPN use at baseline was similar in the T1DM and T2DM patients (P=0.78). More T1DM patients underwent GES implantation before enrollment in the GpCRC Registry (15% vs. 3%, P=0.02). More T1DM patients with gastroparesis reported severe state anxiety (Y1 score ≥50) (P=0.04) and though not significant, more severe trait anxiety (Y2 score ≥50)(P=0.06) compared with T2DM patients. Other psychological survey, overall PAGI-QOL, individual PAGI-QOL domain, and SF-36v2 scores were similar in the two diabetes subtypes ( Table 2A ). Table 2B shows investigator-rated severity of gastroparesis and patient-scored GCSI results in the two subgroups according to baseline HbA1c values <8% vs. ≥8% and the presence or absence of peripheral neuropathy. HbA1c groupings had no relationship to investigator ratings of gastroparesis severity, overall GCSI scores and individual GI symptom scores. Delays in gastric emptying at two or four hours were not related to HbA1c status (P=0.96 or P=0.79). GI symptom severities (except for postprandial fullness) were similar whether the T1DM and T2DM patients did or did not report peripheral neuropathy. Investigator-rated gastroparesis severity and delays in gastric emptying were similar regardless of peripheral neuropathy status. Forty-six baseline predictors were used in regression analyses to determine clinical characteristics that distinguished the T1DM and T2DM patients. Few baseline characteristics discriminated the subtypes ( Table 3 ). Compared with T2DM gastroparesis patients, T1DM patients were about one-third less likely to have more severe bloating (OR=0.62, P=0.02) and almost twice as likely to have GERD symptoms (OR=1.70, P=0.02). T1DM patients were younger (OR Age≥50 yrs =0.07, P<0.001), had more peripheral neuropathy (OR=3.81, P=0.02), had more than 9 times the odds of normal or underweight status (OR=0.11, P<0.001), and reported approximately 25% fewer numbers of comorbidities (OR=0.76, P=0.02) than T2DM patients with gastroparesis. Ninety of 137 enrolled patients (66%) completed the 48 week visit: 44 patients with T1DM (56% of baseline cohort) and 46 patients with T2DM (79% of the baseline cohort). Compared to those with only enrollment data, patients completing follow-up were more often male, white race and overweight and less likely to have GES surgery (P≤0.05); diabetic subgroup was not associated with completing follow-up with adjustment for all other characteristics (OR T1DM vs. T2DM=0.35, P=0.09)( Supplemental Table 2 ). BMI did not change significantly over the 48 weeks in these T1DM and T2DM patients as shown in Table 4 . HbA1c levels increased similarly, but not significantly, compared with baseline over the 48 weeks in both T1DM and T2DM patients (P=0.51). Three patients with T1DM and one with T2DM died during the 48-week period. GI symptom severity did not decrease at 48 weeks in the patients with T1DM as measured by GCSI and individual scores ( Figure 1A ). In contrast, overall GCSI scores and all individual symptoms (except postprandial fullness and visible distention) decreased significantly at 48 weeks in the T2DM patients ( Figure 1B ). Figure 1C shows the changes in patient-reported symptoms (±95% CI) at 48 weeks for both subtypes. Investigator-rated gastroparesis severity ratings showed similar reductions from baseline within T1DM (mean change=-0.33, P=0.009) and T2DM (mean change=-0.30, P=0.02) patients; however, these changes were not different between the subtypes (P T1DM vs T2DM =0.23)( Table 4 ). Increased use of prokinetic drugs (+15.9%), proton pump inhibitor/other GI agents (+13.6%), and anxiolytic drugs (+25.0%) was recorded in T1DM patients (P≤0.03), whereas increased use of opiates (+17.4%) was documented in T2DM patients (P=0.04) at 48 weeks compared with baseline ( Table 4 ). Percentages of patients hospitalized for gastroparesis during the 48-week follow-up decreased 15.1% in the T1DM patients (P=0.04), but did not significantly change for T2DM patients; however, decreases in hospitalizations for T1DM versus T2DM patients were not significantly different (P=0.26). Numbers of ED visits and changes in TPN use over 48 weeks were not different between diabetes subtypes. Implantation of GES devices increased 20.5% over 48 weeks in patients with T1DM (P=0.01) and increased 10.9% in T2DM patients (P=0.06). Including those who were implanted before enrollment, more T1DM patients were receiving GES after 48 weeks of follow-up compared with patients with T2DM and gastroparesis (31.9% vs. 10.9%)(P=0.02). No changes in any psychological or quality of life parameter in either subtype or between subtypes were observed at 48 weeks of follow-up ( Table 4 ). Multiple regression analyses were used to assess the relationship of eight clinical factors to outcomes at 48 weeks in the T1DM and T2DM groups with gastroparesis ( Table 5 ). T1DM patients were less likely to report decreased vomiting (OR =0.21, 95% CI: 0.05–0.87; P=0.03), but more likely to have reductions in loss of appetite scores (OR=4.25, 95% CI: 1.07–16.92; P=0.04) compared to T2DM patients. When data from both diabetic subtypes were pooled, no reduction in any parameter of gastroparesis severity was related to initial HbA1c levels or presence of peripheral neuropathy on enrollment ( Supplemental Table 3 ). Except for decreases in abdominal pain scores in T2DM patients whose HbA1c increased over 48 weeks (P=0.02), changes in symptom severity over 48 weeks were similar in T1DM and T2DM patients whose HbA1c levels either worsened or decreased ( Supplemental Table 4 ). Reduction or no reduction in HbA1c levels did not vary significantly between diabetic subgroups over the 48 week period (data not shown).

Our findings delineate many clinical similarities in patients with T1DM and T2DM and gastroparesis and confirm several demographic differences. GI symptoms rated at baseline were remarkably similar in intensity between diabetic subtypes, including nausea and stomach fullness, with only greater bloating in T2DM patients and increased GERD in T1DM patients being significantly different. Our results also showed the HbA1c levels and the severity of gastric emptying delay did not correlate with the symptoms associated with gastroparesis in either patients with T1DM or T2DM; even though gastric retention severity was higher in T1DM, symptoms were not correspondingly increased at enrollment. The poor relation of symptom severity to gastric emptying in T1DM versus T2DM patients is consistent with recent literature, and suggests other pathophysiologic abnormalities mediate GI symptom genesis ( 8 , 44 ). Factors such as poor fundic accommodation, heightened sensitivity to gastric distention, gastric dysrhythmias, and pyloric dysfunction warrant study as potential causes of GI symptoms associated with diabetic gastroparesis ( 45 , 46 , 47 ). Despite reporting similar GI symptom intensity as T2DM patients, hospitalizations for gastroparesis and for GES implantations were higher in T1DM patients at baseline. It is likely that factors other than gastrointestinal symptoms such as poor glycemic control, as well as dehydration and electrolyte disturbances brought on by acute vomiting may be more relevant drivers of hospitalizations in T1DM patients. Despite this greater resource use in T1DM, overall medication use profiles and quality of life scores were similar to T2DM gastroparetics at baseline. The clinical perception that patients with T1DM and gastroparesis are frequently underweight is not supported by our findings. We found almost half of T1DM patients with gastroparesis were overweight or obese and only 3% were underweight, while T2DM patients with gastroparesis were even heavier as in prior reports ( 19 , 20 ). Baseline TPN use was noted by less than 10% of patients in both subgroups reflecting the ability of most patients to sustain intake by oral or enteral routes. However, these findings do not rule out significant nutritional impairments. Our group previously reported mean daily caloric intakes of less than 1200 calories with deficiencies in essential nutrients including vitamin B 6 , vitamin K, and iron in patients with gastroparesis ( 48 ). An infectious prodrome was noted in 14% of T1DM and T2DM patients. A similar incidence of infectious prodrome has been observed with idiopathic gastroparesis, suggestive of a potential viral etiology in these non-diabetic patients ( 8 ). The role of infections as cofactors in triggering the onset of diabetic gastroparesis could be the focus of additional study. A new finding of this investigation is the difference in gastrointestinal symptoms in the two groups at the 48 week follow up visits. Symptom scores decreased only in the T2DM patients while symptom severity was mostly unchanged in those with T1DM. It is possible the lack of reduction in GI symptom scores at 48 weeks in T1DM patients may reflect irreversible diabetes-related damage to the stomach wall. However, in ultrastructural studies from full thickness gastric biopsies, no differences were observed in the loss of enteric neurons, depletion of interstitial cells of Cajal, or in myenteric immune cell infiltration in specimens from T1DM versus T2DM patients ( 49 ). This differential outcome in GI symptoms in T1DM versus T2DM patients occurred despite aggressive management over the 48 weeks of follow-up. Patients with T1DM more often were prescribed prokinetic agents, proton pump inhibitors, and anxiolytics, and had more GES implants compared with T2DM patients. These interventions had little positive impact on symptoms in the T1DM subgroup, suggesting these patients had a refractory and end-stage condition. However, investigator ratings of gastroparesis severity improved in both diabetic groups at 48 weeks; it is possible this divergence of clinician and patient ratings stemmed from decreases in hospitalizations observed in the T1DM patients. Nevertheless, future studies of investigational therapies of diabetic gastroparesis may need to consider differential responses in the two subtypes. Our results showed no relationship between HbA1c levels and patient-reported GCSI scores or investigator-rated gastroparesis severity at 48 weeks. Furthermore, the decrease in GI symptoms reported by the T2DM patients occurred even though HbA1c values increased slightly over the 48 weeks. Thus, chronic glycemic control did not appear to influence GI symptoms in either group of diabetic patients with gastroparesis. Efforts for tighter glycemic control in patients with long-standing diabetes are important for many reasons, but these findings suggest that symptom reductions (particularly in T2DM patients) can occur without improved glucose control. Ongoing studies employing intensive insulin therapy will more rigorously determine if improved glycemia has additional symptom benefits. Although the presence of peripheral neuropathy was a discriminator of gastroparesis in T1DM versus T2DM on regression analysis, neuropathy did not relate to GCSI scores suggesting that peripheral and visceral complications of diabetes may not necessarily be linked. This finding also raises the possibility that gastroparesis in diabetes is not primarily neuropathic in origin, as suggested by histopathologic investigations performed by the GpCRC ( 26 ). Medication use over 48 weeks differed in the diabetic subtypes. Nearly half of the patients in the subgroups were receiving opioid agents at baseline, but T2DM patients more often were given new opiate prescriptions over 48 weeks of follow up. Abdominal pain is the predominant symptom in 20% of gastroparesis patients, irrespective of etiology ( 4 ). The mechanisms for abdominal pain and the reasons for starting narcotics in T2DM patients are likely to be multifactorial; these could not be discerned from our analyses. In general, opiates slow gastric emptying and may worsen symptoms associated with gastroparesis. However, the reductions in GI symptom scores in the T2DM patients suggest that narcotics did not adversely affect these patients from an overall perspective. Many oral antidiabetic drugs such as metformin can cause nausea and vomiting ( 25 ). Unexpectedly, metformin use was actually associated with less vomiting in T2DM patients. Nevertheless, metformin intake should be considered among the causes of unexplained GI symptoms in T2DM patients. Psychological dysfunction and quality of life are poor in patients with gastroparesis ( 11 ). These measures were equally poor in T1DM and T2DM patients at baseline, although more T1DM patients had severe anxiety. Psychological and quality of life parameters remained unchanged at 48 weeks of follow-up despite decreases in GI symptom scores in gastroparesis patients with T2DM. These findings suggest that GI symptom severity is not the only factor influencing either psychological distress or poor quality of life, at least in T2DM patients. GES therapy was employed more often for gastroparesis in T1DM patients compared to T2DM patients, but the T1DM group did not report a decrease in GI symptoms. Although differences did not reach significance, GES use also was higher over 48 weeks in the T2DM patients (change=11%, P=0.06). GES therapy decreases nausea and vomiting in some but not all studies of diabetic gastroparesis ( 50 ). Future controlled investigations assessing GES efficacy in diabetic gastroparesis should be performed to contrast benefits in T1DM versus T2DM patients. Our study had some limitations. First, determination of diabetic subtype was dependent on subject report and review of the medical records by the investigator. Second, referral bias may have influenced the findings because the patients were referred to the tertiary motility centers of the GpCRC. Thus, our patients may not reflect typical patients managed in the community and they may have had clinical features that were unfavorable for symptom reductions over 48 weeks of follow up. However, given the similarities in baseline GI symptoms in the T1DM and T2DM patients, it is likely referral bias was similar for both diabetic subtypes. Nevertheless, more T1DM patients had GES therapy which probably reflects the refractory nature of symptoms in this group. Third, assessments of healthcare utilization did not include costs or address length of stay, outpatient visits, and missed work. Fourth, 48 weeks may be an inadequate time period to detect symptom score reductions in T1DM patients or differentiate resource utilization. The numbers of patients available for study at 48 weeks may have precluded some smaller differences not being detected; however, our study had 80% power to detect a minimal difference of 0.6 SD units in GCSI between the two subtypes. Finally, we had some concerns about non-significant trends to higher dropouts over 48 weeks of follow-up in the T1DM patients. Although several minor differences were observed in patients who did versus did not attend their 48 study visits, the lack of relation of 48 week visit attendance and GCSI scores, diabetes subtype, and hospitalizations confirmed that the different outcomes of T1DM and T2DM patients were not due to differential study compliance. We believe these limitations are countered by the significant strengths of the study including the large numbers of patients with gastroparesis and T1DM and T2DM, use of standardized tests and protocols and comprehensive collection of clinical, psychological, quality of life, and healthcare usage data. In conclusion, our findings challenge several clinical axioms about gastroparesis in the two diabetic subtypes. First, baseline gastrointestinal symptoms associated with gastroparesis were remarkably similar in T1DM versus T2DM patients, even though T1DM patients had more severe gastric emptying delays and higher hemoglobin A1c values. These observations suggest the presence of other gastric or extragastric pathogenic factors may mediate gastroparesis symptom severity. Second, symptoms associated with gastroparesis in both diabetic subtypes did not correlate with HbA1c levels or severity of gastroparesis; and last, after 48 weeks of follow-up, most GI symptom scores decreased only in T2DM patients even though T1DM patients showed increased medical and surgical treatment utilization. These similarities and differences in patients with T1DM and T2DM form a basis for further research to improve clinical outcomes with novel drugs, gastric stimulation parameters, and insulin dosing regimens for symptoms associated with gastroparesis.

Seventy-eight patients with T1DM and 59 patients with T2DM and gastroparesis in the Gastroparesis Clinical Research Consortium (GpCRC) Registry were identified. Each patient completed validated surveys and underwent examinations and blood testing on enrollment and at 48-week follow-up visits from January 2007 to May 2011 (ClinicalTrials.gov Identifier: NCT00398801 ). All subjects reported symptoms associated with gastroparesis for at least 12 weeks duration (not necessarily contiguous weeks) and had gastroparesis defined by scintigraphy (>60% retention at 2 hours and/or >10% retention at 4 hours) within 6 months of enrollment ( 5 ). Prokinetics, opiates, anticholinergics, and other agents that affect gut transit were stopped at least 72 hours before gastric emptying testing. Upper endoscopy performed within 1 year of Registry enrollment showed no evidence of organic causes of symptoms. Patients with ulcers, malignancy, mechanical obstruction, active inflammatory bowel disease, eosinophilic gastroenteritis, neurologic disease, hepatic or renal disease, other metabolic disease, or prior gastroesophageal surgery were excluded. The determination of T1DM versus T2DM status and the diagnosis of diabetic gastroparesis were made by each site investigator based upon patient reports and review of records. Studies were approved by Institutional Review Boards at each Clinical Center and Data Coordinating Center. Patients provided written informed consent. Survey completion, examinations, and local laboratory blood testing were performed on enrollment and 48-week follow up visits. Demographic and medical information was collected on Registration and Baseline Medical History forms ( Supplemental Methods ), including self-reported clinician-diagnosed peripheral neuropathy. Body mass index (BMI) was calculated at both times from physical examination data; numbers and percentages who were overweight or obese (≥25 kg/m 2 ) were calculated. Numbers and percentages with any comorbidity and numbers of comorbidities were determined on enrollment ( Supplemental Methods ). As inflammatory activation has been identified in some cases of gastroparesis, C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) were measured on enrollment as non-specific markers of inflammation ( 27 ); numbers and percentages with elevated CRP (>0.8 mg/dl) and ESR values (>20 mm/hr) were determined; any inflammation was defined as either an elevated CRP and/or elevated ESR. Hemoglobin A1c (HbA1c) was quantified at both visits; numbers and percentages of patients with HbA1c values <8% vs. ≥8% were defined. Gastroparesis severity was quantified in four ways: 1) investigator-rated gastroparesis severity was assessed on enrollment and at 48 weeks by each principal investigator using an expert consensus stratification ( Supplemental Methods )( 2 ); 2) Patient Assessment of Upper Gastrointestinal Disorders Symptoms (PAGI-SYM) questionnaires were used to quantify 20 individual symptoms that the patient scored from 0 (none) to 5 (most severe)( 28 ); 3) overall symptom severity was determined by total scores from the Gastroparesis Cardinal Symptom Index (GCSI)( Supplemental Methods )( 29 ); and 4) percentages of test meal retained at four hours from pre-enrollment scintigraphy studies were used to stratify results into mild (11–20%), moderate (21–35%), and severe (>35% retained) gastric emptying delays ( 30 ). Medication use was queried on enrollment and at 48 weeks ( Supplemental Methods ). Numbers and percentages of T2DM patients taking antidiabetic medications known to cause nausea and vomiting were determined ( 25 ). Symptoms were compared in patients who were taking versus not taking these agents on enrollment. Health utilization parameters were determined. On enrollment, patients reported how many times they were hospitalized over the prior year and for what reasons they were hospitalized. At 48 weeks, they were asked how many times they had required emergency department (ED) evaluation or hospitalization solely for gastroparesis since enrollment (excluding gastric electrical stimulator [GES] implantation). Numbers and percentages of patients on total parenteral nutrition (TPN) and who had undergone GES implantation were determined at baseline and 48 weeks. Measures of psychological dysfunction and quality of life were quantified. Depression and anxiety were enumerated by the Beck Depression Inventory (BDI) and State and Trait Anxiety Inventory (STAI)( Supplemental Methods )( 31 , 32 ). Numbers and percentages with severe depression (BDI score >28), state anxiety (Y1 score ≥50), and trait anxiety (Y2 score ≥50) were calculated. Disease-specific and generic quality of life was assessed by Patient Assessment of Upper Gastrointestinal Disorders Quality of Life (PAGI-QOL) and Short Form-36v2 (SF-36v2) surveys, respectively ( Supplemental Methods )( 33 , 34 ). Enrollment (baseline) symptom scores were subtracted from 48-week values to calculate changes in all measures. Baseline BMI, HbA1c, hospitalizations for gastroparesis, patients on TPN or undergoing GES, BDI, Y1, Y2, and quality of life scores were subtracted from 48-week levels to quantify changes. Numbers and percentages of patients on different medications at baseline were subtracted from 48-week values to estimate changes. Number and percentages or means ± SD were reported for enrollment categorical or continuous characteristics. P values were determined from Pearson chi-square or Fisher’s exact tests for categorical characteristics and Kruskal-Wallis tests to account for non-normality of continuous distributions ( 35 ). Baseline discriminators of diabetes subtype were determined from backward stepwise multiple logistic modeling regressing diabetes subtype on the 46 baseline characteristics, forcing age at enrollment, sex, and white race into the model, with P for exclusion=0.05 ( 36 ). Total number rather than individual comorbidities was included; GCSI was excluded due to collinearity. Hosmer-Lemeshow testing revealed adequate fit for the model (P=0.71). Differences between patients completing 48-week follow up versus patients with only enrollment data were assessed using multiple logistic regression of 48-week completion on baseline characteristics (diabetes subtype, demographics, BMI, severity, medications [prokinetics, opiates, antidepressants], healthcare utilization, psychological function, quality of life). Mean changes ± SD at 48 weeks versus enrollment were computed for all characteristics except ED visits (not queried on enrollment). For continuous characteristics, P values were determined using one sample t-tests of the null hypothesis of no difference in means at both visits within diabetes subtype comparisons. For binary characteristics and medication changes, exact McNemar’s tests for paired proportions were used to determine P and 95% confidence intervals (CI) which were computed using continuity corrections ( 37 , 38 ). Multiple regression models of 48-week changes in continuous characteristics, adjusting for enrollment values, assessed changes between diabetes subtype (except for ED visits)( 39 ). Negative binomial regressions (to account for overdispersion) of ED visits over 48 weeks on diabetes subtypes were used. Wald’s tests using conditional logistic regression tested if 48-week changes in hospitalizations for gastroparesis or medication use varied by subtype ( 36 ). Unconditional exact logistic regression assessed TPN use and GES changes with T2DM. Relative odds of changes in 48-week outcomes were derived from logistic regression models of each indicator at 48 weeks in relation to subtype and enrollment value of the indicator. Models included propensity scores to adjust diabetes subtype effects for probabilities of being T1DM based on age, sex, and race ( 40 ). Outcome indicators defined by 48-week changes from enrollment included any symptom score decrease, no change or decreased BMI, any HbA1c decrease, ≥5 point BDI decrease, any STAI decrease, and any QOL increase. Healthcare utilization reductions were defined as no hospitalizations or ED visits for gastroparesis over 48 weeks. Given the exploratory nature of our study, P values were two-sided and nominal with significance at the P=0.05 level, a priori . Because a goal of these exploratory analyses was to generate new hypotheses to be tested in future confirmatory studies, correction for multiple comparisons was not performed. Such adjustments reduce the power of an investigation to define important differences, are unnecessary if exploratory research questions are unrelated, and are only required for studies which aim to offer decisive proof of a predefined hypothesis to endorse decision-making protocols ( 41 , 42 , 43 ). Stata (Stata Statistical Software, Release v12; StataCorp LP, College Station, TX) and SAS (version 9.3, SAS Institute, Inc., Cary, NC) software were employed.