Outcomes and Factors Associated With Reduced Symptoms in Patients with Gastroparesis

Abstract

Background & Aims

Gastroparesis is a chronic clinical syndrome characterized by delayed gastric emptying. However, little is known about patient outcomes or factors associated with reduction of symptoms.

Methods

We studied adult patients with gastroparesis (of diabetic or idiopathic type) enrolled in the NIDDK Gastroparesis Clinical Research Consortium Gastroparesis Registry, seen every 16 weeks and treated according to the standard of care with prescribed medications or other therapies at 7 tertiary care centers. Characteristics associated with reduced symptoms, based on a decrease ≥1 in the gastroparesis cardinal symptom index (GCSI) score after 48 weeks of care, were determined from logistic regression models. Data were collected from patients for up to 4y (median, 2.1 y).

Results

Of 262 patients, 28% had reductions in GCSI scores ≥1 at 48 weeks. However, there were no changes in GCSI score from weeks 48 through 192. Factors independently associated with reduced symptoms at 48 weeks included male sex, age ≥50 y, initial infectious prodrome, antidepressant use, and 4 hr gastric retention>20%. Factors associated with no reduction in symptoms included overweight or obesity, a history of smoking, use of pain modulators, moderate to severe abdominal pain, a severe gastroesophageal reflex, and moderate to severe depression.

Conclusions

Over a median follow-up period of 2.1 y, 28% of patients treated for gastroparesis at centers of expertise had reductions in GCSI scores ≥1, regardless of diabetes. These findings indicate the chronic nature of gastroparesis. We identified factors associated with reduced symptoms that might be used to guide treatment. ClinicalTrials.gov no: NCT00398801

Introduction

Gastroparesis is a syndrome characterized by delayed gastric emptying, in the absence of an obvious structural abnormality. Symptoms include chronic nausea and vomiting, early satiety, postprandial fullness and abdominal distention. Gastroparesis is usually either idiopathic or secondary to diabetes in the vast majority of cases. Little is known about the longitudinal course of patients with gastroparesis and the clinical or pathophysiological characteristics, if any, that may predict it. Trials of therapy, even when rarely associated with a positive outcome, have generally been very limited in duration.^{1, 2} Other reports on outcomes have been retrospective in nature with metrics that have not been standardized.³

The large multi-center Gastroparesis Registry (GpR) study organized by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) funded Gastroparesis Clinical Research Consortium (GpCRC) has provided the opportunity to study these patients in a more comprehensive, prospective and systematic manner (ClinicalTrials.gov Identifier: NCT00398801). Previous studies based on this Registry have contributed to our knowledge about the baseline clinical, demographic and pathophysiological features of these patients.^4–12

In this study, our aim was to describe changes in patient clinical outcomes and determine what baseline attributes, if any, predict substantial improvement in symptoms using follow-up data available at 48 weeks from patients followed prospectively for up to four years (median=2.1 years). A secondary aim was to describe the proportion of symptomatic improvement in these patients over the study follow-up.

Methods

Patients

Adult patients were recruited through the GpR. Patients eligible for GpR had to have symptoms of at least 12-weeks in duration and no evidence of mechanical gastric obstruction by endoscopy. In addition, a detailed history and physical exam were required to exclude any other plausible explanation for symptoms in these patients before enrollment, and appropriate tests (e.g. biliary imaging or gastrointestinal contrast studies) were ordered as indicated. Once enrolled, patients were followed every 16 weeks and prescribed medications or other therapies as per the judgment of the treating physician. At clinic visits, SES, anthropometric, medical history, symptom severity, psychological function and quality of life data were collected via surveys and physical exams. Blood for routine biomarkers (glycated hemoglobin, HbA1c; C-reactive protein, CRP; erythrocyte sedimentation rate, ESR) was collected; inflammation was defined as either CRP>1.0 mm/dL and/or ESR>20 mm/hr. For this study, patients had to have either diabetic or idiopathic gastroparesis, and at least 1 complete clinical report form of interest at week 48.

Human Subjects Approval

The study was approved by the Institutional Review Board of all the participating institutions as well as the consortium’s Data and Safety Monitoring Board.

Assessments

Medical history and physical exams

Patient self-reported medical histories were collected at enrollment and at each follow-up visit from face-to-face interviews using itemized check lists of physician-diagnosed comorbidities, medications, and supplemental therapies and procedures. The number of hospitalizations during the past year at enrollment or the numbers of hospitalizations and emergency department visits (ED) incident since last visit at follow-up were queried. The study physician used his or her best judgment to determine the gastroparesis etiology using the patient’s history, medical records and laboratory tests. History and current status of drinking and smoking were collected; a regular smoker was defined as a patient who had ever smoked at least 100 cigarettes. Physical examinations included duplicate height, body weight, waist and hip measurements, which were averaged for analyses. Patients were classified as either overweight or obese (BMI≥25 kg/m²) or as normal or underweight (BMI<25 kg/m²).

Symptom Scoring and Grading of Severity

Symptoms were scored using the PAGI-SYM (Patient Assessment of Upper Gastrointestinal Disorders Symptom Severity Index) questionnaire, which also contains the Gastroparesis Cardinal Symptom Index (GCSI), a validated scale of severity (from 0 to 5) in patients that utilizes three symptom clusters (nausea/vomiting/retching, post-prandial fullness/satiety, and bloating/stomach distention) measured over the preceding two weeks.^{13, 14} The GERD severity subscale utilizes the 7 items querying the heartburn/regurgitation cluster.¹⁵ Patients also recorded the predominant symptom, which was then grouped into one of three categories: nausea cluster, abdominal pain, or any other symptom. Moderate to severe abdominal pain was defined as either upper abdominal pain or discomfort score of 3 or greater.¹⁶ The clinical severity of gastroparesis was graded globally into three levels (1= mild, 2 =compensated and 3 = severe, gastric failure) by the investigators as previously described.¹⁷

Gastric Emptying

A consensus protocol was used to assess gastric emptying by scintigraphy using a low-fat, egg white meal labeled with radioactive technetium with imaging at 0, 1, 2, 4 hours after meal ingestion.¹⁸ As per this consensus report, delayed gastric emptying was defined as greater than 60% retention at two hours and/or 10% at 4 hours. For purposes of this study, moderate to severe delay was defined as >20% retention at 4 hours.

Quality of Life (QOL)

The Patient Assessment of Upper Gastrointestinal Disorders-Quality of Life (PAGI-QOL) is a 30-item instrument that has been validated as a reliable and sensitive measure of quality of life in patients suffering from dyspepsia, GERD or gastroparesis in the prior two weeks.^{19, 20} Evaluation of a patient’s general quality of life in the prior four weeks was determined using the physical and mental health component subscores of the Medical Outcomes Study 36-Item Short-Form Health Survey V2 (SF-36v2) standard recall, which has been normalized to the 1998 U.S. general population with a mean (±SD) of 50 (±10). A higher score on either instrument reflects higher QOL or better health outcome.

Psychological Measurements

The Beck Depression Inventory (BDI)²¹ and State-Trait Anxiety Inventory (STAI) were used to measure depression in the past two weeks and current and general anxiety, respectively.²² Moderate to severe depression was defined as a BDI score greater than 20, and severe state or trait anxiety was defined as a score of 50 or more.

Statistical Analysis

Changes in global patient outcomes were analyzed by computing mean change at 48 weeks from baseline of each outcome, and computing one sample t-tests of the null hypothesis of no difference in follow-up and baseline means; an exact McNemar’s test for paired proportions was used to assess TPN.²³ Determination of differences in mean change in each outcome between diabetic and idiopathic subgroups was computed using multiple linear regression models of each continuous outcome, including the subgroup indicator and controlling for the outcome’s baseline value. Rates of total hospitalizations in the past year were compared using a mixed effects negative binomial regression model with counts of hospitalizations as the outcome, an indicator variable for etiology, and controlling for the total hospitalization rate for the year prior to enrollment. Each patient’s TPN use at 48 weeks was compared to TPN use at enrollment within each subgroup; a conditional logistic regression model was used to assess change in use by etiology.

Patient symptomatic improvement was defined as a ≥1 point decrease in GCSI at 48 weeks from enrollment. Supplemental Table 1 presents the 35 baseline predictors in the candidate set. Baseline predictors were presented as percentages or means (±SD), mean changes from baseline, and mean baseline GCSI and 48 week GCSI change for each category of each predictor. Significance of the differences between the improved versus not improved patients were determined from either a t-test or ANOVA for continuous characteristics and from chi-square tests or Fisher exact tests for categorical characteristics. The final model of independent baseline predictors of 48 week symptom improvement was determined from multiple logistic regression of improvement in relation to all possible subsets of the candidate set of 35 baseline indicators that resulted in the model with the lowest Akaike Information Criteria (AIC).^{24, 25} Model fit was determined adequate using the Hosmer-Lemeshow test.²⁶ To account for over-dispersion in the final model, standard errors were scaled using square root of the deviance-based dispersion. Cross-validated Area Under the Receiver Operator Characteristic (AUROC) curve was computed from jackknifed regression.²⁷ The AUROCs for the comprehensive model including the full candidate set of predictors and the final parsimonious model were not different (P=.07) (Supplemental Table 1). Logistic regression using the final model and 20 multiple imputation data sets for the observations missing baseline data assessed no differences for model estimates from those reported.

Patterns of improvement in GCSI scores, defined as the difference between the GCSI at a visit timepoint being 1 or more points lower than the baseline GCSI, were shown graphically for all patients and comparing etiology subgroups. A test for trend was used to assess whether improvement differed across timepoints for all patients and between etiology subgroups. Determination of differences between etiology subgroups at selected timepoints was computed from multiple logistic regression models controlling for the baseline GCSI; the models across all timepoints used robust variance estimation and a method to account for repeated subject observations; to test for a trend of symptom improvement over time, the P value was determined from a Wald test of time and time by subgroup interactions. All reported P values were 2-sided and nominal. All statistical analyses were performed using SAS Statistical Software version 9.3 (SAS Institute, Inc., Cary, NC) and Stata (Release v13; StataCorp LP, College Station, TX).

All authors had access to the study data and reviewed and approved the final manuscript.

Results

Patients

Of a total of 394 patients with gastroparesis (of either diabetic or idiopathic type) enrolled in the Registry study over a three-year period, 262 met the entry criteria for inclusion in the analyses of the improvement outcomes at 48 weeks. There were 132 (34%) patients not included in the analyses due to not having completed a PAGI-SYM at 48 weeks. A total of 47 (18%) males and 215 (82%) females, with an average age of 44 years (SD 13.5 years) were in our study; the vast majority of patients (89%) were white. Overall, the physician reported the etiology for 177 patients (68%) as idiopathic, 40 (15%) as Type 1 diabetes and 45 (17%) as Type 2 diabetes.

Analyses of patients with and without the 48 week visit indicated that those patients missing a GCSI score at 48 weeks were almost twice as likely to have severe trait anxiety than those with a symptom improvement score at 48 weeks (OR=2.20, CI: 1.24–3.93, P=.007), but otherwise did not differ on demographics, symptom severity, clinical characteristics and depression (Table 1).

Table 1.

Association of Demographics and Baseline Characteristics in Relation to Patients without a 48 week PAGI-SYM Compared to Patients Completing the 48 week Follow-up

Baseline Characteristics	No Vs Completed 48 Week PAGI-SYM (N=394)
	OR	95% CI	P*
Demographics:
Male (vs female)	0.56	0.28–1.09	.09
Race: White (vs non-white)	0.66	0.33–1.32	.24
Age 50+ (vs <50 yrs)	0.81	0.47–1.38	.43
Education: college degree (yes vs no)	0.60	0.35–1.05	.07
Income (≥ $50,000 vs < $50,000)	1.12	0.69–1.82	.64
Symptom Severity:
Overall GCSI score	0.98	0.59, 1.64	.95
Nausea subscale	0.99	0.78, 1.25	.91
Post prandial fullness subscale	0.87	0.61, 1.25	.46
Abdominal pain moderate/severe (vs mild/none)	1.15	0.64, 2.08	.64
GERD subscale	0.95	0.78, 1.16	.64
Clinical:
Retention >20% (vs ≤ 20%)	1.29	0.79–2.12	.32
Etiology:			.24
Idiopathic	1.00
Diabetes Type 2	0.79	0.35–1.82	.58
Diabetes Type 1	1.65	0.73–3.72	.23
BMI overweight (≥25kg/m2) vs (not)	0.67	0.41–1.10	.11
Smoked regularly (vs not)	1.49	0.89–2.48	.13
Initial infectious prodrome (vs none)	0.67	0.36–1.26	.22
Hba1c (%)	1.04	0.64–1.69	.88
Psychological and QOL:
BDI moderate to severe (vs less)	0.92	0.51–1.66	.77
Trait anxiety ≥ 50 (vs not severe)	2.20	1.24–3.93	.007
Intercept	1.06	0.22–5.05	.94

Note: Of 394 patients with gastric emptying, 132 (34%) did not have 48 week visit data, 260 had complete 48 week visit data for all characteristics in the model, and 2 had a 48 week PAGI-SYM but were missing data on a baseline characteristic.

^*ORs, 95% CI, and P determined from a multiple logistic regression model of no follow-up PAGI-SYM data available at 48 weeks compared to having 48 week PAGI-SYM data in relation to 18 predictors.

Change in global outcomes over the first 48 weeks

Patients with gastroparesis as a group, showed a significant decline, as compared to baseline, of 0.42 in the GCSI score and a decline of 0.26 in the investigator-rated gastric symptom severity with an average weight gain of 1.64 kg (all P<.001). Mild improvements in quality of life as measured by PAGI-QOL (mean gain of 0.29) and both the physical and mental components of the SF-36v2 (mean gains of 1.77 and 1.89, respectively)(all P≤.003). However, no significant changes were seen for depression or anxiety scores. Total hospitalizations declined significantly by 0.73 from a baseline rate of 2.3 per year (P=.003). A more robust reduction (nearly 50%) was seen in the proportion of patients on TPN, decreasing from 9.7% at baseline to 4.9% after 48 weeks (P=.03).

Diabetic and idiopathic patients with gastroparesis showed no significant differences in improvement in any of the measures, except for the difference in the average number of ED visits in the first year (3.6±6.0 vs. 2.0±4.6, P=.009)(see Table 2).

Table 2.

Patient Outcomes after 48 Weeks in Diabetic and Idiopathic Gastroparesis Patients: Symptoms, Anthropometry, Clinical Factors, Depression, and Quality of Life

Outcome characteristics	Total Patients*			P‡	Diabetic		Idiopathic		Diabetic vs Idiopathic P§
	Baseline	Mean Δ†	95% C.I.		Baseline	Mean Δ†	Baseline	Mean Δ†
Symptom Severity:
Investigator-rated gastric symptom severity	2.2	−0.26	−0.35, −0.17	<.001	2.3	−0.31	2.08	−0.23	.38
Patient-rated overall GCSI score (PAGI-SYM)	2.98	−0.42	−0.56, −0.29	<.001	2.98	−0.35	2.91	−0.46	.58
Anthropometric:
BMI (kg/m2)	27.9	0.61	0.30,0.93	<.001	30.7	0.61	26.2	0.62	.88
Weight (kg)	76.1	1.64	0.78,2.50	<.001	85.8	1.52	70.7	1.69	.87
Clinical factors:¶
Total hospitalizations (no.)ǁ	2.3	−0.73	−1.02, −0.26	.003	2.5	−1.14	1.07	−0.52	.10
On TPN (%)	9.7%	−4.8%	−9.2, −0.5%	.03	6.7%	−3.3%	11.2%	−5.6%	.98
Depression & QOL:
Depression (BDI)	18.8	−0.54	−1.76,0.69	.39	20.2	−0.28	17.9	−0.66	.50
State Anxiety total score	44.6	0.04	−1.72,1.81	.96	44.4	0.74	43.6	−0.31	.61
Trait Anxiety total score	43.6	0.41	−0.95,1.78	.55	44.4	1.34	42.7	−0.05	.26
PAGI-QOL total score	2.5	0.29	0.17,0.41	<.001	2.5	0.19	2.5	0.34	.33
SF-36v2 physical component	32.9	1.77	0.61,2.92	.003	31.3	1.56	34.8	1.87	.34
SF-36v2 mental component	37.3	1.89	0.32,3.45	.002	37.1	1.08	38.8	2.28	.21

^*Total N determined by value for outcome being available at baseline and at 48 weeks; total N varies between 251 and 269; for patients with diabetes, total N varies between 82 and 90; and for idiopathics, between 168 and 179.

^†Mean baseline and mean change of outcome (48 weeks – baseline) presented for all patients and separately by etiology subgroup; 95% confidence interval (C.I.) for the mean change presented for all patients.

^‡For continuous characteristics, P value determined using one sample t-test of the null hypothesis of no difference in means at follow-up and baseline. For the binary outcome, TPN, an exact McNemar’s test for paired proportions was used to determine P, and 95% confidence intervals (C.I.) determined using a continuity correction.

^§For continuous characteristics, P determined using multiple regression of each characteristic of change at 48 weeks from baseline in relation to indicator of etiology (diabetic vs idiopathic) with adjustment for the baseline value of the characteristic. P for the total hospitalizations in past year determined from a negative binomial regression of total hospitalizations in relation to etiology with adjustment for the total hospitalizations in year prior to enrollment. P for the binary characteristic, TPN, derived from a Wald test to assess whether change in TPN use varied by etiology using conditional logistic regression.

^¶ED visits were not reported at enrollment. The mean number of ED visits from enrollment to 48 weeks was 2.55±5.16 for all gastroparesis patients; comparing patients with diabetes to those with idiopathic gastroparesis it was 3.6±6.0, and 2.0±4.6, respectively, P=.009 (determined using negative binomial regression).

^ǁTotal hospitalizations since baseline excluded gastric stimulator placement.

Range of improvement

The range of any improvement in GCSI from baseline at 48 weeks was approximately normally distributed. We further examined the number and percent of patients with symptomatic improvement using different definitions of improvement based on the absolute change in GCSI as presented in Table 3. Two-thirds of the patients (66%) registered a decline of any magnitude in their GCSI over a 48-week period. A decline of at least 1.0 point was observed in 28% of all patients, while a decline of 0.75 and 0.50 was seen in 34% and 41% of patients, respectively. Only 15% of patients reported an improvement of 50% or more in their GCSI score. There were no differences in the two etiological groups by any definition of symptom improvement.

Table 3.

Number and Percent of Patients with Improvement in Gastroparesis Symptoms over 48 Weeks using Different Definitions of Improvement

Definition Of Improvement	All Patients (N=262) No. Improve (%)	Diabetic (DM) (N=85) No. Improve (%)	Idiopathic (N=177) No. Improve (%)	P*DM vs Idiopathic
No. points decrease from baseline at 48 weeks
> 0.00	175 (67%)	52 (61%)	123 (69%)	.21
0.25	135 (51%)	38 (45%)	97 (55%)	.14
0.50	108 (41%)	34 (40%)	74 (42%)	.87
0.75	89 (34%)	25 (29%)	64 (36%)	.31
1.00	73 (28%)	21 (25%)	52 (29%)	.47
1.25	54 (21%)	17 (20%)	37 (21%)	.92
1.50	42 (16%)	13 (15%)	29 (16%)	.88
1.75	29 (11%)	11 (13%)	18 (10%)	.41
2.00	16 (6%)	6 (7%)	10 (6%)	.58
% change at 48 weeks
50% over baseline	38 (15%)	14 (16%)	24 (14%)	.56
Mean change:
Mean GCSI change	−0.42 ± 1.11	−0.35 ± 1.18	−0.46 ± 1.07	.58

^*To compute the difference in mean GCSI change between patients with diabetic gastroparesis and idiopathic, multiple linear regression was used; all other differences between the 2 groups were assessed from multiple logistic models. Each regression model of symptomatic improvement included an indicator variable for the subgroup and controlled for the baseline value of the GCSI score.

Characteristics of patients who improved by 48 weeks versus those who did not

Table 4 presents the baseline characteristics of patients in relationship to symptomatic improvement at 48 weeks, along with mean baseline and absolute changes in GCSI score for each of these characteristics, with additional details presented in Supplemental Table 2. We found that improved gastroparesis patients were distributed no differently than unimproved ones with respect to several demographic variables such as gender, race and SES, but a greater proportion of patients 50 years or older were improved at 48 weeks as compared with younger patients (36% vs 24%, P=.04).

Table 4.

Proportion of Patients with Improvement in GCSI (≥1 point on GCSI) at 48 Weeks for Baseline Predictors and Corresponding Mean Change from Baseline in GCSI Scores

Baseline Characteristics	Total No. (N=262)	Improved At 48 Wk N (%)	P*	Mean Baseline GCSI	P†	Mean Change in GCSI	P†
Demographics:
Sex			.69		.01		.19
Male	47	12 (26%)		2.64		−0.23
Female	215	61 (28%)		3.01		−0.47
Race			.21		.96		.58
White	234	68 (29%)		2.98		−0.44
Non-white	28	5 (18%)		2.99		−0.31
Age			.04		.01		.16
≥ 50 years	83	30 (36%)		2.74		−0.58
<50 years	179	43 (24%)		3.09		−0.35
Education:			.22		.01		.16
College degree or higher	83	19 (23%)		2.75		−0.28
Some college or less	179	54 (30%)		3.10		−0.49
Income:			.25		.02		.55
$50,000 or higher	143	44 (31%)		2.85		−0.39
< $50,000	119	29 (24%)		3.15		−0.47
Symptoms: Investigator- and Patient-rated
Global symptom severity (investigator-rated)			.48		<.001		.48
Mild	36	11 (31%)		2.40		−0.44
Moderate	136	34 (25%)		2.99		−0.36
Gastric failure	87	28 (32%)		3.24		−0.55
Nausea/vomiting moderate/severe‡			.004		<.001		<.001
Yes	114	42 (37%)		3.58		−0.75
No	148	31 (37%)		2.53		−0.17
Fullness moderate/severe‡			.002		<.001		<.001
Yes	185	62 (34%)		3.40		−0.62
No	77	11 (14%)		1.98		0.06
Bloating moderate/severe‡			.17		<.001		<.001
Yes	162	50 (31%)		3.52		−0.61
No	100	23 (23%)		2.11		−0.13
Abdominal pain mod/severe‡			.81		<.001		.008
Yes	191	54 (28%)		3.28		−0.53
No	71	19 (27%)		2.18		−0.13
GERD moderate/severe‡			.95		<.001		.10
Yes	65	18 (28%)		3.75		−0.62
No	196	55 (28%)		2.73		−0.36
Predominant symptom‡:			.03		<.001		.003
Other PAGI-SYM	104	25 (24%)		2.63		−0.26
Nausea/vomiting	111	40 (36%)		3.33		−0.66
Upper abdominal pain	47	8 (17%)		2.95		−0.23
Gastric emptying scintigraphy:
Gastric Retention at 4 hours:			.04		.13		.23
≤ 20%	109	23 (21%)		2.97		−0.34
>21%	153	50 (33%)		3.15		−0.59
Medical history:
Etiology:			.63		.94		.45
Idiopathic	177	52 (29%)		3.00		−0.46
Diabetes Type 2	45	11 (28%)		2.96		−0.49
Diabetes Type 1	40	10 (22%)		2.92		−0.20
BMI category‡			.006		.47		.05
Overweight or obese	147	31 (21%)		2.94		−0.30
Not overweight	115	42 (37%)		3.04		−0.58
Smoked (ever regularly)			.03		.41		.003
Yes	84	16 (19%)		2.91		−0.13
No	178	57 (32%)		3.02		−0.56
Acute onset of symptoms			.41		.20		.59
Acute	139	42 (30%)		2.91		−0.46
Insidious	121	31 (26%)		3.08		−0.46
Initial infectious prodrome			.02		.71		.38
Yes	48	20 (42%)		2.93		−0.55
No	214	53 (25%)		3.00		−0.39
Inflammation‡			.41		.82		.71
Yes	111	28 (25%)		3.00		−0.39
No	151	45 (30%)		2.97		−0.45
HbA1c (%):			.49		.34		.33
< 7 %	208	60 (29%)		3.01		−0.46
≥ 7 %	54	13 (24%)		2.86		−0.29
Current medication use:
Narcotics			.09		.08		.49
Yes	115	26 (23%)		3.11		−0.37
No	147	47 (32%)		2.88		−0.46
Protein pump inhibitors			.28		.98		.42
Yes	212	56 (26%)		2.98		−0.40
No	50	17 (34%)		2.98		−0.54
Prokinetics			.24		.88		.98
Yes	148	37 (25%)		2.99		−0.42
No	114	36 (32%)		2.97		−0.42
Antiemetics			.89		<.001		.23
Yes	167	47 (28%)		3.17		−0.48
No	95	26 (27%)		2.66		−0.31
Antidepressants			.44		.70		.40
Yes	105	32 (30%)		3.02		−0.33
No	157	41 (26%)		2.97		−0.46
Anxiolytics			.02		.21		.03
Yes	50	7 (14%)		3.15		−0.11
No	212	66 (31%)		2.94		−0.50
Pain modulators			.07		.47		.55
Yes	63	12 (19%)		3.07		−0.34
No	199	61 (31%)		2.96		−0.45
TPN			.42		.04		.20
On	26	9 (35%)		3.38		−0.69
Not on	236	64 (27%)		2.94		−0.39
GES implantation			.99		.07		.94
Yes	18	5 (28%)		3.40		−0.40
No	244	68 (28%)		2.95		−0.42
Psychological & QOL
Depression (BDI)			.07		<.001		.68
Moderate/severe(> 20)	109	24 (22%)		3.31		−0.39
Mild to none (≤ 20)	153	49 (32%)		2.75		−0.45
State-Trait Anxiety Inventory:
State anxiety ≥ 50			.25		.01		.24
Yes	86	20 (23%)		3.21		−0.31
No	176	53 (30%)		2.87		−0.48
Trait anxiety ≥ 50			.49		.07		.57
Yes	80	20 (25%)		3.16		−0.36
No	182	53 (29%)		2.91		−0.45
PAGI-QOL score			.82		<.001		.88
Medium to high (≥ 3)	87	25 (29%)		2.29		−0.41
Low (score < 3)	175	48 (27%)		3.33		−0.43
SF-36v2 Overall Health:
Physical health:			.42		<.001		.28
Score ≥ 35	101	31 (31%)		2.59		−0.33
Score < 35	161	42 (26%)		3.23		−0.48
Mental health:			.33		.002		.73
Score ≥ 35	149	45 (30%)		2.81		−0.44
Score < 35	113	28 (25%)		3.21		−0.40

^*Improved defined as the 48 week PAGI-SYM GCSI being 1 or more points lower than the baseline (BS) GCSI. P (2-sided) determined from either a Pearson’s Chi-Square test.

^†P (2-sided) determined from t-test or ANOVA for outcomes with 3 categories.

^‡Definitions: Nausea/vomiting, fullness (post-prandial), bloating, and GERD symptoms were each defined as moderate to severe if the subscale was 3 or higher; abdominal pain was defined as moderate to severe if any of the components of the subscale was 3 or higher; Predominant symptom is the patient’s selection for the most severe among any symptom on the PAGI-SYM questionnaire in the previous 2 weeks, and the other category includes post-prandial fullness, bloating, lower abdominal pain, GERD, constipation, or diarrhea; Overweight or obese defined as BMI≥25kg/m², not overweight is BMI <25kg/m²; Inflammation defined as an elevated C-reactive protein (CRP) level or erythrocyte sedimentation rate (ESR).

Several measures of symptom severity at baseline were also examined. The investigator-rated severity profiles did not differ amongst gastroparesis patients who improved and those who did not. However, total GCSI scores as well as several subscales (nausea/vomiting, postprandial fullness and bloating) were higher in the improved group. This is also reflected in the finding that patients who improved identified nausea/vomiting/retching as their predominant symptom in the previous 2 weeks more often than upper abdominal pain or discomfort. However, quality of life measures at baseline were no different in the improved and unimproved groups. A greater proportion of patients with moderate to severe delay in gastric emptying at baseline were improved at 48 weeks as compared with the unimproved group (68% versus 54%, P=.04).

Idiopathic and diabetic gastroparesis patients (both type 1 and 2) were equally represented in the improved and unimproved groups of gastroparetics. HbA1c levels were similar as well. However, an important difference at baseline was in the average BMI, with unimproved patients being significantly heavier (28.4 ± 7.6 versus 26.2 ± 7.9 kg/m²) and only 21% of overweight or obese patients improved after 48 weeks as compared with 37% of normal or underweight patients (P=.006). There was also a significant difference in the proportion of smokers with 19% showing improvement at 48 weeks as compared with 32% of non-smokers (P= 0.03). Improvement was proportionately greater also in patients whose illness appeared to be associated with an infectious prodrome at the onset as compared to those without such a prodrome (42% versus 25%; P=.02). However, an acute onset of symptoms by itself was not associated with improvement.

We also examined the relationship between medication use (at the time of enrollment) and symptomatic improvement at 48 weeks and found no difference amongst users of proton pump inhibitors, prokinetics, antiemetics, and antidepressants. However, gastroparesis patients on anxiolytics were less likely to show an improvement at 48 weeks as compared with those not on these medications (14.0% versus 31.1%, P=.02). Further, borderline associations were seen in the use of other pain modulators with non-improvement. Neither the use of TPN or presence of gastric electric stimulation was associated with improvement.

Amongst the psychometric measures used in the study, only a borderline significant improvement was seen in association with moderate to severe depression and no association was found with anxiety state or trait measures.

Predictors of 48 week symptomatic improvement using baseline patient characteristics

Independent baseline predictors of 48 week symptomatic improvement using a parsimonious model are shown in Table 5. Older age (≥50 years) was associated with the best outcome with an odds ratio of improvement of 3.35 (CI:1.62–6.91, P=.001). Other positive predictors were the overall GCSI score at baseline (with higher scores associated with a more favorable response; OR=2.87, CI: 1.57–5.23, P=.001), gastric retention of >20% at four hours (OR=2.22, CI: 1.12–4.40, P=.02) and antidepressant use (OR=2.27, CI: 1.13–4.58, P=.02). Though not significant, male gender was associated with improvement (OR=2.30, P=.07), as was an initial infectious prodrome (OR=2.22, P=.05). Negative predictors of improvement included being overweight or obese (OR=0.43, P=.01), having more severe GERD (OR=0.66, P=.006), moderate to severe depression (OR=0.45, P=.03), the use of anxiolytics and pain modulators (OR=0.28, P=.02 and 0.34, P=.01, respectively), smoking (OR=0.46, P=.04) and moderate to severe abdominal pain (OR=0.40, P=.04).

Table 5.

Baseline Predictors of Improvement in Gastroparesis Symptoms after 48 Weeks using Logistic Regression Analyses in Patients with Gastroparesis

Baseline Predictors Selected by Minimum AIC	Symptoms Improved Vs Not Improved (N=260; 73 Improved)
	OR*	CI	P†
Demographics:
Sex (male vs female)	2.30	0.93–5.67	.07
Age (≥ 50 vs <50 years)	3.35	1.62–6.91	.001
Education: college degree or higher (yes vs < no)	0.58	0.27–1.22	.15
Gastroparesis Symptoms:
Overall GCSI score	2.87	1.57–5.23	.001
Post-prandial fullness severity subscale	1.53	0.92–2.57	.10
Abdominal pain moderate/severe‡ (vs not)	0.40	0.17–0.96	.04
GERD severity subscale	0.66	0.50–0.89	.006
Gastric Emptying Scintigraphy:
Retention > 20% at 4 hours (vs ≤ 20%)	2.22	1.12–4.40	.02
Clinical factors:
Overweight or obese (BMI ≥25 vs < 25 kg/m2)	0.43	0.22–0.84	.01
Smoked (ever vs never regularly)	0.46	0.22–0.98	.04
Initial infectious prodrome (vs not)	2.22	1.01–4.87	.05
Prokinetics use (vs none)	0.57	0.29–1.10	.09
Antidepressants use (vs none)	2.27	1.13–4.58	.02
Anti-anxiety (anxiolytics) use (vs none)	0.28	0.10–0.79	.02
Pain modulator medication use (vs none)	0.34	0.14–0.81	.01
Psychological:
Depression‡: Moderate to severe (BDI > 20 vs ≤ 20)	0.45	0.22–0.93	.03
Intercept	0.02	0.00–0.11	<.001
AUROC	0.77§	0.71–0.84

^*Odds of improvement; improvement defined as GCSI score at 48 weeks compared to the baseline score is 1 point or more lower (number patients with improvement=73, without improvement=187, with missing data for a predictor=2).

^†

P values were determined from a multiple logistic regression analyses of improvement using the 35 baseline predictors in the candidate set listed in Table 4, including baseline GCSI score. To account for over-dispersion of the model, standard errors were scaled using square root of the deviance-based dispersion. Nausea, post-prandial, and GERD severity subscales and the SF-36 v2 physical and mental health components were analyzed as continuous rather than categorical variables. The bloating subscale was excluded from the candidate set due to collinearity with other baseline predictors. The model presented was the model selected having the minimum Akaike Information Criterion (AIC) amongst all of the possible subsets of predictors.

• Goodness of fit: Hosmer-Lemeshow χ² (d.f.=6)=10.0, P=.12
• Cross-validated AUROC using jackknifed regression method is presented in the table.
• AUROC without cross-validation is: 0.84 (0.78–0.89).

^‡Moderate/severe abdominal pain defined as either upper abdominal pain or discomfort PAGI-SYM symptom score≥3; Beck Depression Index (BDI) score>20 indicates a moderate to severe level of depression; Trait Anxiety score ≥50 indicates severe anxiety

^§Prediction equation (b(SE)) for probability(improvement) = odds/(1+odds), where odds=EXP[−3.981(0.932) + 0.834(0.480) if male + 1.209(0.385) if age≥50 (yrs) −0.548(0.398)^*if college degree or higher + 1.053(0.321)^*GCSI total score + 0.428(0.273)^*post-prandial subscale score −0.907(0.461) if abdominal pain moderate/severe −0.409(0.156)^*GERD subscale score + 0.795(0.365) if 4 hr GES retention >20% −0.842(0.356) if BMI ≥25 (kg/m²) −0.770(0.400) if smoked regularly + 0.798(0.418) if initial infectious prodrome −0.564(0.350) if use prokinetics + 0.820(0.374) if use anti-depressants −1.264(0.545) if use anxiolytics −1.086(0.463) if use pain modulators −0.788(0.381) if BDI>20]

Extension of Follow-up

In additional analyses we examined symptomatic improvement of patients beyond the first 48 weeks. Although progressively fewer patients were available over time, we were able to draw some inferences over a total of 192 weeks. Figure 1A displays the improvement from baseline at each follow-up visit for patients with gastroparesis. No difference was seen in the percent of patients improving between 48 and 192 weeks; however, patients’ symptoms improved significantly from 16 to 48 weeks (19% vs 28%, P=.01). Analyses of symptomatic improvement across follow-up comparing diabetic and idiopathic patients indicated no differences between subgroups at annual time points or across all follow-up. Neither diabetics nor idiopathics had differences in improvement between the first year to the study end; however, idiopathics had increased improvement from 16 to 48 weeks (18% vs 29%, P=.001) (Figure 1B).

Figure 1A.

Figure 1B.

DISCUSSION

This study represents the largest prospective survey of patients with gastroparesis and provides important information on this clinically challenging syndrome. Although a third of patients improve modestly, the majority of patients show little, if any, substantial improvement, as assessed by a validated measure, the GCSI.

This study used the GCSI to monitor symptoms at baseline and in follow-up. In the original description of the GCSI, a decrease in symptom severity of 0.75 points in the mean GCSI total score was considered an improvement in clinical status by physicians as well as patients.¹⁴ However, subsequent studies using larger numbers of patients and an interventional arm suggest that this number may be too low. Thus, in a randomized controlled trial of a ghrelin agonist for gastroparesis, the anchor-based determination of the minimal clinically important difference (MCID) was determined to be 0.94 at the group level between active versus placebo arms.²⁸ In addition to these results, mathematical models comparing these suggested cut-points in our study patients selected the 1.0 decline to be the best model by AIC selection criterion, as well as having the highest ability to discriminate the group with improvement.

In this study, we used a decline of 1.0 to classify patients as symptomatically improved. As compared with less robust MCIDs, this resulted in only a marginal difference in terms of the proportion of patients improving. These results speak of a considerably bleaker outcome than previous case series have reported.^{3, 29} However, even though those reports were from tertiary academic centers, they were retrospective in nature, did not use comprehensive and validated instruments for collection of information and did not have a standardized definition for clinical improvement.

Given these results, it is not surprising to see that patients continued to carry a significant burden of disease with attendant socio-economic consequences. Patients, particularly those that were diabetic, often visited the emergency department and were repeatedly hospitalized, up to 3 or more times a year. However, it should be noted that these numbers were based on self-reporting and may be subject to error. Nevertheless, many other measures attest to the lack of substantial improvement. Quality of life remained impaired, as assessed by two measures. The first, PAGI-QOL, is more disease specific and created from interviews of patients with GERD, dyspepsia and gastroparesis from six countries including the USA. Our study patients averaged 2.5, which corresponds to a patient-rated assessment of severe/very severe, according to the original validation report.²⁰. Further, although there was a statistically significant improvement in the PAGI-QOL scores after 48 weeks, the mean change was 0.29, which is considerably less than MCID (0.4) when comparing pre-treatment and post-treatment PAGI-QOL total scores.²⁰. These results paralleled those obtained by the more general SF-36 QOL instrument, with both physical and mental component considerably lower than the general population norm of 50 and also somewhat lower than what has been reported in patients with inflammatory bowel disease including Crohn’s disease (between 40–50).^{30, 31}

An important aim of this study was to determine important predictors of improvement over a year. The 48 week data from these patients were used, primarily due to much smaller numbers of patients available with 96 or more weeks of follow-up. Though adequate to provide an overall look at improvement in the cohort and between the etiology subgroups, there are not enough patients with improvement in order to develop predictive models. Regression analysis of baseline patient characteristics revealed the following independent predictors of symptomatic improvement at 48 weeks: male gender, age ≥50 years, initial infectious prodrome, antidepressants use, 4-hour gastric retention >20%, never smoked, BMI<25kg/m², no pain modulators use, no to mild abdominal pain, low GERD severity, and no to mild depression.

Although causality cannot be established, some of these factors may have plausible explanations. Thus, a similar association with an improved outcome with age has been noted in some other medical disorders as well, with the speculation that this may relate to better coping/adapting skills acquired with age, a tendency to view adverse events more positively by elderly patients and a smaller gap between their health status and that of their peer comparison group.³² Male gender was associated with a more than two-fold improvement over females. Although it is well known that gastroparesis predominantly affects women, this study shows for the first time that they are also less likely to improve over time. The underlying mechanisms remain unknown, although experimental evidence suggests that females are more vulnerable to disturbances in nitrinergic signaling.³³

A history of infectious prodrome was also associated with improvement, which is consistent with historic reports. In this study, 18% of patients with gastroparesis reported an infectious prodrome to their illness, which is similar to what we had previously described for idiopathic gastroparesis.⁷ Small retrospective studies have suggested that a presumed post-viral association can be found in about 7–8% of patients and that these patients tend to have a much better prognosis, with several showing complete resolution within a few within a few years.^{34, 35}

Depression was an important negative factor, with moderate to severe depression resulting in over a fifty percent reduction in the probability of improvement. We have previously shown in this cohort that higher depression and anxiety scores are associated with gastroparesis severity on investigator- and patient-reported assessments.⁹ The relationship between depression and chronic illness is well established but complex and not fully understood even in classic inflammatory conditions such as rheumatoid arthritis.³⁶ Depression can induce or exacerbate physical derangement directly such as by aberrant activation of the immune system, while chronic physical symptoms can in turn generate dysphoria and feelings of helplessness. Further, the presence of concomitant depression may significantly alter coping mechanisms as well as contribute to increased sensory perception with intensification of symptoms such as pain and nausea. Finally, despite these intuitively appealing theoretical considerations, most studies have not consistently shown an improvement in somatic illness by treating depression or vice versa.³⁷ Nevertheless, it is intriguing that antidepressant use was associated with an increase of nearly two-fold in the odds of improving.

In our study, GERD symptom severity also correlated with a lower likelihood of improvement.³⁸ It is known that delayed gastric emptying increases the total number of reflux events in patients. Further, in some patients with gastroparesis, heartburn and/or regurgitation may be the predominant symptom.³⁹ On the other hand, nausea and/or vomiting can occur in 15–30% of patients with reflux-induced esophagitis⁴⁰ and in some cases, GERD can present as intractable nausea.⁴¹ It should also be noted however, that we did not document reflux by objective measures and further, the use of proton pump inhibitors was not important in predicting improvement. Thus it is possible that some of the GERD-like symptoms may reflect abnormalities in sensation rather than true reflux.

Body weight was an important and relatively novel association with the likelihood of improvement. We have previously shown that nearly half of patients with idiopathic gastroparesis are either overweight or obese, a finding that may be counter-intuitive, given that the clinical picture is dominated by complaints of nausea, vomiting and early satiety.⁷ In the present study, 56% of the cohort was overweight, and these patients were much less likely to improve than patients whose BMI was less than 25 kg/m². The relationship between obesity and gastric function is still incomplete, but there is no obvious correlation between body habitus and gastric emptying or other parameters such as accommodation.⁴² However, in a recent report of patients with type 2 diabetes, obesity was found to be a significant predictor of symptoms suggestive of gastroparesis and a significantly greater prevalence of early satiety, fullness, bloating and abdominal distention was noted in obese patients.⁴³ A high fat diet leading to obesity can also potentially play a role in the pathogenesis of enteric neuropathy and gastroparesis, as suggested by studies in rodents.^44–48

Patients with gastroparesis with a history of smoking were less likely to improve. Although this is not surprising, given the overall negative effects of smoking on health, it is more difficult to speculate on the specific mechanisms. There are many constituents of smoking, such as carbon monoxide and even nicotine itself that can influence gastrointestinal motility and acutely, nicotine can induce emesis;^{49, 50} however, paradoxically, smokers are at lower risk of post-operative nausea and vomiting.^{51, 52} Abdominal pain, independent of narcotic use, was another negative predictor. Although pain is not a component of the GCSI, it is increasingly being recognized as an important symptom in gastroparesis^{16, 53} and has also been shown to be more important than other gastrointestinal manifestations in determining quality of life in conditions such as irritable bowel syndrome.⁵⁴

In conclusion, we have shown that less than a third of patients with gastroparesis improve over time and the disease burden remains high. We have also identified several patient characteristics that are associated with symptomatic improvement. These results may help us triage patients with gastroparesis (at least for those refractory patients presenting to tertiary medical centers) for more focused or aggressive treatment strategies. We also hope that these results will provide the basis for validation studies in the larger community setting in the future.

Author Contributions and Conflict of Interest Statement

Pankaj J. Pasricha: study concept and design, acquisition of data, writing of manuscript

Katherine P. Yates: statistical analysis and presentation of data, writing and revision of manuscript

Linda Nguyen: acquisition of data; critical revision of the manuscript for important intellectual content

John Clarke: critical revision of the manuscript for important intellectual content

William L. Hasler: study concept and design, acquisition of data, critical revision of the manuscript for important intellectual content

Thomas L. Abell: study concept and design, acquisition of data, critical revision of the manuscript for important intellectual content

Gianrico Farrugia: study concept and design, critical revision of the manuscript for important intellectual content

Kenneth L. Koch: study concept and design, acquisition of data, critical revision of the manuscript for important intellectual content

Henry P. Parkman: study concept and design, acquisition of data, critical revision of the manuscript for important intellectual content

William J. Snape: acquisition of data, critical revision of the manuscript for important intellectual content

Richard W McCallum: acquisition of data, critical revision of the manuscript for important intellectual content

Irene Sarosiek: acquisition of data, critical revision of the manuscript for important intellectual content

Linda Lee: technical support, safety issues.

James Tonascia: study concept and design, statistical analysis, critical revision of the manuscript for important intellectual content

Frank Hamilton: study supervision, critical revision of the manuscript for important intellectual content

Laura Miriel: critical revision of the manuscript for important intellectual content

None of the authors report any conflict of interest with respect to this manuscript or the data in

Author names in bold designate shared co-first authors.