White Paper AGA: Gastroparesis: Clinical and Regulatory Insights for Clinical Trials

Abstract

Gastroparesis continues to represent a large unmet clinical need and a major opportunity for new drug development. This has led to increasing interest by federal funding agencies, regulatory bodies, and industry. This article summarizes the proceedings of the gastroparesis section of the “Drug Development Conference: Clinical Endpoints in Upper GI Disorders” organized by the American Gastroenterological Association in Washington, DC, on October 27–28, 2016. The presentation, diagnosis, and current therapeutic strategies are briefly reviewed, followed by a detailed discussion of the regulatory strategy, recommended endpoints, and future directions.

Presentation and Diagnosis

Gastroparesis is a motility disorder of the stomach characterized by delayed gastric emptying without evidence of obstruction. The predominant symptoms of gastroparesis are nausea, vomiting, early satiety, postprandial fullness, abdominal pain, and bloating all of which can significantly impact quality of life.¹ Patients with gastroparesis, at least those referred to tertiary medical centers, seem to have a chronic and relatively unrelenting course; over a median follow-up period of nearly 2 years, only about a quarter of patients showed improvement, regardless of the presence of diabetes.² Mortality also seems to be significantly higher in patients with gastroparesis compared with age-matched control subjects.³

At tertiary medical centers at least, most forms of gastroparesis are idiopathic in nature, with a significant minority possibly related to a poorly defined infectious prodrome.² Among known causes, diabetes (type 1 and type 2) accounts for about 30% of patients, with the rest associated with gastric surgery, systemic disorders (eg, chronic renal failure, Parkinson disease, scleroderma) or induced by drugs (eg, opioids, anticholinergics).

The most commonly used method for the diagnosis of gastroparesis in the United States is a formal measurement of emptying of a radionuclide-labeled solid meal. The 2 best accepted protocols define normal emptying as more than 40% and 90% at 2 and 4 hours, respectively, with a 2% fat 255-kcal meal,⁴ or >25% and >75% emptied at 2 and 4 hours with a 30% fat 320-kcal meal.⁵ This type of standardized radioscintigraphic test is not routinely performed in the community, rendering it difficult to accurately measure the incidence and prevalence of gastroparesis defined as such.⁶ These efforts are further complicated by relatively acute changes in emptying with glycemic levels in diabetics. A population-based survey of gastroparesis has estimated a low incidence and prevalence of “definite” gastroparesis (6.3 and 24.2 per 100,000 inhabitants, with about twice that number if “probable and possible” cases were included).³ “Definite” was based on documented delay in gastric emptying in the participants’ medical records, based on the 320-kcal, 30% fat meal. This may be lower than the true prevalence rate, based on predictive models, which estimate that 1.8% of the subjects may have gastroparesis, so-called “hidden” gastroparesis.⁷ The “risk” of developing gastroparesis over a 10-year time period was 5.2% in type 1 diabetes mellitus, 1.0% in type 2 diabetes mellitus, and 0.2% in control subjects.⁸ Regardless of the true epidemiology, hospitalizations for gastroparesis seem to be increasing markedly in adults and children.^9,10

Other measures of gastric emptying have been used in clinical settings and include breath testing and the wireless motility capsule (WMC). The nonradioactive ¹³C-labeled spirulina platensis breath test was approved by the Food and Drug Administration (FDA) in 2015 and relies on rate-limiting gastric emptying of a ¹³C-labelled meal that is digested in the small bowel where ¹³CO₂ is liberated, diffuses across the intestinal mucosa, and is transported in the circulation to the lungs where it is exhaled in time-dependent fashion.¹¹ Quantification of indigestible solid gastric emptying also is accomplished using WMC methodologies that measure pyloric transit by detecting characteristic >2–3 pH unit increases as the device passes from the stomach into the duodenum. One prospective study defined delayed gastric emptying as prolonged WMC retention in the stomach >5 hours after ingestion.¹² When directly compared, gastric emptying rates from both breath tests and WMC studies show correlation with 4-hour scintigraphy values but as yet the ¹³C-spirulina breath test has recently become available for use in clinical practice. The WMC is also available for point-of-care testing, but capsule emptying occurs with return of the fasting phase III migrating motor complex signifying and therefore reflects a different phase of gastric motility than the fed state, which may account for the relatively low concordance rate between the 2 tests (53% when tests are performed on separate days with different meals).¹³ Currently in the United States, meal-based scintigraphy is considered the gold standard for the initial diagnosis of gastroparesis, although newer technologies offer promising alternatives that may have advantages in some clinical settings.

Gastroparesis: Current Therapeutic Strategies

The current treatments for gastroparesis are less than satisfactory and this represents a large unmet need and an opportunity for the pharmaceutical industry. In the United States, the only drug approved for this indication is metoclopramide, although the FDA suggests treatment for up to 12 weeks. Metoclopramide may exert its benefits by both a central antinauseant effect and improving gastric emptying. However, its use is largely limited because of neurologic side effects including extrapyramidal reactions (eg, tardive dyskinesia, dystonia, Parkinson-like movements) that have led to a black-box warning. Domperidone, a related drug, does not cross the blood-brain barrier as much as metoclopramide and its use is not associated with the same neurologic risks. However, it is not approved in the United States outside of an investigational new drug protocol as recommended by the FDA. Furthermore, even in countries where it is approved, the drug is coming under increasing scrutiny because of the risk of sudden death and cardiac arrhythmias associated with inhibition of hERG channel activity, which inhibits the rapid component of the cardiac delayed rectifier K(+) current (I(Kr)).^14,15 Macrolide antibiotics, particularly erythromycin, are also commonly used “off-label” to accelerate gastric emptying in patients with gastroparesis based on their ability to act as agonists on the motilin receptor. Clinical trials in small numbers of patients suggest improvement in symptoms and emptying, although long-term benefit may be limited by tachyphylaxis.¹⁶ Other agents have been used in the past including 5HT₄ receptor agonists. Initial agents in this drug class had cardiac side effects; newer 5-HT₄ agonists with little to no cardiac side effects are being evaluated currently.

Although a variety of antiemetic drugs are widely used for symptomatic treatment of nausea in patients with gastroparesis, none have been rigorously evaluated or approved for this indication. Most agents, such as 5-HT₃ receptor antagonists (eg, ondansetron and granisetron) and NK₁ receptor antagonists (eg, aprepitant), have been adopted from other fields of study including chemotherapy-induced and postoperative nausea and vomiting. Most recently, a 4-week randomized placebo controlled trial of aprepitant for the treatment of chronic nausea in patients with and without delayed emptying has shown promise.¹⁷

Clinical guidelines and a summary of evidence on these and other therapies have been published in recent years and are summarized in Table 1.¹⁸ This table predated the results of the NORIG trial, which failed to show a beneficial effect of low-dose nortriptyline in patients with idiopathic gastroparesis.¹⁹ This trial was, however, remarkable for the low placebo response rate (23%) belying the general impression that a high placebo rate is a barrier for proving efficacy in this condition.

Table 1.

Current Approaches for the Treatment of Gastroparesis

Therapy	Recommendation	Level of evidence
Prokinetics to accelerate gastric emptying and reduce symptoms	Strong	Moderate
Metoclopramide	Moderate	Moderate
Domperidone	Moderate	Moderate
Erythromycin	Strong	Moderate
Antiemetics	Conditional	Moderate
Tricyclics	Conditional	Low
Botulinum toxin	Strong, not recommended	High
Gastric electrical stimulation	Conditional	Moderate

From Camilleri et al¹⁸

Gastroparesis: Trial Design and Endpoints

The FDA Guidance for Industry on Gastroparesis was issued for commentary in July 2015. Several important areas in this document are mentioned here for phase III studies for drug approval. Studies should have a 1-to-2-week baseline period with a 12-week treatment period, followed by a 2-to-4-week randomized withdrawal period. A long-term safety study of 12-months duration is recommended. Idiopathic gastroparesis and diabetic gastroparesis should be studied in separate clinical trials. If adequate safety and efficacy are demonstrated for both indications, 1 trial in idiopathic gastroparesis can cross support another trial in diabetic gastroparesis and result in approval for both indications. The primary endpoint should be a change in symptoms from baseline. Gastric emptying need not be studied as an endpoint; if it is, it should be a secondary endpoint.

These guidelines are welcome and necessary given the increasing interest of pharmacologic companies in evaluating new drugs for gastroparesis. In this regard, the question of the appropriate endpoints and evaluable outcome measures has assumed great importance. The first truly validated instrument developed specifically for evaluating gastroparesis was the Gastroparesis Cardinal Symptom Index (GCSI). Before that several scoring systems had been used in clinical trials but none had been subject to rigorous psychometric testing. The key psychometric attributes of a good outcomes measure include the use of an equal-interval scale, such as Likert; reliability (most easily assessed by internal consistency, the extent to which all the items in the instrument measure the same concept or construct [ie, the interrelatedness of the items]); and responsiveness. Most importantly the measure must have construct validity, that is assessed by the extent to which an instrument measures the concept or construct it is intended to measure.

The Gastroparesis Cardinal Symptom Index

Given its widespread use, a thorough understanding of the development and validation of the GCSI is important. The GCSI itself was derived from the Patient Assessment of Upper Gastrointestinal Disorders–Symptom Severity Index (PAGI-SYM), a self-reported instrument for subjects with gastroparesis, dyspepsia, and gastroesophageal reflux disease.²⁰ The 2 key contributors to the construction of this instrument were semi-structured interviews with patients, and interviews and meetings with experienced gastroenterologists. For the former, 122 subjects from 6 countries (France, Germany, Italy, Poland, Sweden [all face-to-face interviews], and the United States [telephone interviews]) were asked both open- and closed-ended questions about their gastrointestinal (GI) symptoms and how this affected their functioning and everyday activities. For the latter, 9 gastroenterologists in the 6 countries were interviewed to specify symptoms by GI disorder. Subsequently, the most relevant symptom domains for the 3 conditions were identified, created in US English and pilot-tested in 13 subjects with gastroesophageal reflux disease symptoms. The draft PAGI-SYM contained 37 items of which 17 were deleted on testing because of redundancy, poor item characteristics, or duplicate loadings on the factor analysis, leaving 20 final symptoms. A 6-point Likert response scale (0 = none; 1 = very mild; 2 = mild; 3 = moderate; 4 = severe; 5 = very severe) was used to rate these over a 2-week recall period. Subscale scores are calculated by taking the mean of the items in each subscale. The total score is calculated by taking the mean of the subscales. Details of the psychometric performance and validation of the PAGI-SYM and PAGI-QOL (the health-related quality of life instrument that was developed in parallel) have been published elsewhere.²¹

The GCSI consists of 3 subscales of the PAGI-SYM instrument, selected to measure important symptoms related to gastroparesis: nausea/vomiting (3 items), postprandial fullness/early satiety (4 items), and bloating/distention (2 items). The psychometric qualities of the GCSI were subsequently examined in 169 patients with gastroparesis (requiring both clinical and emptying criteria to be met), recruited from 7 clinical centers in the United States.^20,22 In addition to the GCSI, patients were asked to complete the Short Form-36 Health Survey and questions about disability days at baseline and after 8 weeks. Symptoms and change in symptoms were also rated independently by clinicians at these visits. Reliability of the GCSI was proven by high internal consistency and test-retest reproducibility (the GCSI total score varied by only 0.02 points over 2 weeks in stable patients). Construct validity was shown by correlation with clinician-rated symptoms, Short Form-36 summary scores, and disability days. Responsiveness was demonstrated by comparing changes in GCSI with changes in overall gastroparesis symptoms as assessed by clinicians and patients.

Another important aspect of a useful outcomes measure is its ability to identify a “minimum clinically important difference” or minimum important difference” (MID), which is the smallest change in the measure that a patient would consider as personally meaningful or important, above and beyond the statistical significance of the change in score. In the original description of the GCSI, a decrease in symptom severity of 0.75 points in the mean GCSI total score (out of total maximum score of 5 derived from 9 symptoms) was considered an improvement using an anchor-based method (ie, corresponding to a change in overall clinical status assessed by physicians and patients).²⁰ When the GCSI was subsequently used in a randomized controlled trial of a novel treatment, the anchor-based determination of the minimum clinically important difference was higher (0.94 at the group level between active versus placebo arms).²³ Finally, a large outcome study of patients with gastroparesis using mathematical models to compare minimum clinically important differences selected a 1.0 decline in GCSI to have a high ability to discriminate the group with improvement.²

The Gastroparesis Cardinal Symptom Index–Daily Diary Score

As originally described, GCSI scores were based on 2-week recall by patients, rendering it vulnerable to concerns about the ability to capture more momentary or short-term fluctuations in symptoms, as recommended by regulatory bodies, such as the FDA.²⁴ A daily diary version of the GCSI (GCSI-DD) was therefore developed to address the potential for recall bias in the original GCSI and perceived lacunae in content, such as items dealing with abdominal pain, an important symptom in at least a subset of patients with gastroparesis. Content validity was first described in a small cohort of 12 patients with gastroparesis who underwent cognitive debriefing after being asked about their experience with gastroparesis symptoms in general and to describe those symptoms.¹ They were also requested to complete the GCSI-DD and asked whether they considered the GCSI symptoms to be important and their descriptions relevant to their experience. All patients understood diary instructions and item content and reported that the diary captured their gastroparesis symptom experience; 83% considered response scales adequate. Although there was significant daily variability in GCSI-DD scores, mean GCSI-DD subscale and total scores over 2 weeks correlated strongly (all r >0.90) with standard 2-week recall GCSI scores, suggesting that, in fact, recall bias was a minor issue in the small sample of patients tested. No ceiling effects were observed for any of the items but floor effects were seen for retching and vomiting, indicating that these items cannot be scored accurately for severity and leading to the use of actual number of episodes as a more robust endpoint, rather than an adjectival, Likert, or numerical score of severity.

Psychometric qualities of the GCSI-DD were reported in a subsequent study of 69 patients with idiopathic, diabetic, and postfundoplication gastroparesis.²⁵ Test-retest reproducibility and internal consistency were rated high and construct validity was demonstrated by significant correlations between GCSI-DD scores and clinician ratings of symptom severity. Responsiveness was tested by categorizing patients over time (4 and 8 weeks) as responders or nonresponders, as defined by the patient-reported and clinician-reported Overall Treatment Effect Scale, with any improvement >1 (“a little better”) indicating a response and any other score (ie, ≤1) indicating the opposite. Based on this, effect size (a standardized indicator for the magnitude of observed change with effect sizes of 0.5–0.8 change considered moderate and >0.8 change considered large) for the major symptoms were calculated. Effect sizes for individual symptoms varied: they were highest for postprandial fullness (0.83), stomach fullness (0.78), early satiety (0.53), loss of appetite (0.51), nausea (0.42), and bloating (0.34); and lowest for upper abdominal pain (0.29), vomiting (0.22), and retching (0.09).

MIDs were also calculated based on the difference in scores associated with small improvement on the overall treatment effects scales. MIDs were 0.55 for nausea, 0.97 for excessive fullness, 0.63 for bloating, and 0.77 for postprandial fullness. In this study, the investigators also explored alternative composite scores based on different combinations of what were believed to be the most relevant/important symptoms, as reported by the patients with gastroparesis. A composite score of 4 symptoms (Composite 1: nausea, bloating, excessive fullness, postprandial fullness) had a moderate effect size of 0.61 and MID of 0.73. Part of the suggestion for this composite is based on the argument that this set of symptoms is clinically relevant but also most consistent with abnormalities in gastric emptying.

The most recent version of the GCSI-DD developed in collaboration with the American Neurogastroenterology and Motility Society has been proposed as a patient-reported outcome endpoint for gastroparesis clinical trials, adheres to recommendations provided in the FDA guidance for gastroparesis, and is currently undergoing evaluation by the FDA. The American Neurogastroenterology and Motility Society GCSI-DD has the patient rate severity of nausea, early satiety, postprandial fullness, upper abdominal pain, and overall symptoms over the last 24 hours. Patients rate their symptom severity on a scale of 0 (no symptom), 1 (mild), 2 (moderate), 3 (severe), and 4 (very severe). In addition, patients record the number of vomiting episodes per day. A daily composite score was calculated as the average of the 5 symptom scores (with a cap of the score for vomiting episodes of 4, analogous to the scoring for the symptom severity).

Other Instruments

In a recent trial of a ghrelin agonist, relamorelin, the investigators used a patient-reported outcome instrument, the Diabetic Gastroparesis Symptom Severity Diary, which was reported to have been developed in a manner consistent with the FDA guidance document.²⁶ This 24-hour recall questionnaire is administered via an interactive voice response system (ensuring timing of responses and standardization) and evaluated bloating, abdominal pain, and nausea items rated from 0 (no symptom) to 10 (worst possible symptom). In addition, the early satiety item is rated on the following scale: only 1 or 2 bites = 10; a small snack or portion of a normalsized meal = 7.5; about half of a normal-sized meal = 5.0; most of a normal-sized meal = 2.5; and all of a normal-sized meal = 0. At the present time, no details are available on the psychometric performance of this instrument.

The Patient-Reported Outcomes Measurement Information System is an initiative of the National Institutes of Health to develop a set of patient-reported outcomes that capture the patients’ illness experience in a structured format using highly efficient, computer-based, and short questionnaires. In 2014, the first National Institutes of Health Patient-Reported Outcomes Measurement Information System G1 symptom measures were published. Using psychometric analyses in 865 patients with a variety of G1 disorders and 1177 participants from the general population, the investigators found 8 major symptom complexes: gastroesophageal reflux, disrupted swallowing, diarrhea, bowel incontinence/soiling, nausea and vomiting, constipation, belly pain, and gas/bloating.²⁷ These complexes are not designed to be disease specific but were envisioned as being useful for assessing anyone experiencing a GI symptom, whether patients or the general population. Some of these items (in italics) are also relevant for patients with gastroparesis but their responsiveness has yet to be tested, even if they were to be used for evaluating individual symptom complexes.

Emerging Biomarkers, Targets, and Future Directions

It is clear that an ideal drug for gastroparesis is one that improves symptoms and accelerates gastric emptying. In terms of drug discovery, however, what is more controversial is how closely these 2 concepts are tied together. In other words, would accelerating gastric emptying inevitably lead to resolution/improvement of symptoms? The current state of knowledge does not allow us to answer this question confidently. First, there is relatively weak correlation between the degree of gastric delay and severity of symptoms.²⁸ However, it has to be acknowledged that until recently, analysis of the published literature is limited by the lack of standardization of the methods and meals used for measurements of gastric emptying, and the lack of standardized appraisal of symptoms. Second, trials of prokinetics have shown either no improvement of symptoms despite improvement of gastric emptying or if symptomatic improvement does occur, there was no relation of symptom reduction to acceleration of emptying.¹⁶ Research on this issue has been hampered by the absence of efficacious medications that act solely by stimulating gastric emptying. This does not mean that improving gastric emptying is not a valid target or that gastroparesis and related conditions are not disorders of motility. Rather, it reinforces the concept that the pathophysiological basis of symptoms is multifaceted, arising from different regional and temporal disturbances affecting fundal accommodation, antral contractions, pyloric relaxation, and so forth, with a variable and somewhat unpredictable net impact on a global parameter, such as emptying.

An understanding of the neuropharmacologic basis for these differences represents an opportunity for more targeted approaches aimed at distinct components of the gastroparesis symptom complex. Such a strategy can expect to receive regulatory support given the recent FDA draft guidance that states “Because gastroparesis manifests as more than 1 core sign or symptom, the effect of new drugs intended to treat gastroparesis on each core sign and/or symptom should be assessed” provided that such drugs “do not worsen the remaining signs/symptoms of gastroparesis.”²⁴

A “dispersive” approach to symptom improvement in gastroparesis has several theoretical advantages. By relieving the pressure to find a “magic bullet” that results in improvement of complex global outcomes, there is an opportunity to develop and validate therapeutic approaches including drugs or devices that either already exist based on symptom efficacy (eg, postoperative or chemotherapy-induced nausea and vomiting). Furthermore, by parsing the symptom complex, it is potentially easier to focus on specific molecular targets. An emerging candidate platform for such an approach is NK₁ receptor antagonism, as demonstrated by encouraging results with aprepitant in the recently completed APRON trial.¹⁷ As another example, the 5-HT₁ agonist buspirone can significantly increase gastric accommodation and reduce symptoms of postprandial fullness, early satiation, and upper abdominal bloating, without changes in gastric emptying.²⁹ Another approach might be the development of rational combination therapies that can be tailored to individual symptom profiles or clinical acuity. Such approaches have been successfully used in other therapeutic areas, such as inflammatory bowel disease or hepatitis. This will provide much needed symptomatic relief to patients while the search for the ideal single drug continues or disease-modifying targets are discovered and validated. Consistent with the FDA guidance, it should also be possible to pursue development of therapeutics for gastroparesis that address composite endpoints, particularly when there is validation that the composite endpoint closely reflects the symptoms of greatest significance to patients with gastroparesis, as shown by the APRON study with aprepitant using GCSI-DD²² and the relamorelin studies using Diabetic Gastroparesis Symptom Severity Diary.^17,26

Several additional questions remain to be fully answered, such as whether patients with diabetic gastroparesis respond differently to pharmacologic treatment than those with the idiopathic form of the disorder. Such a difference in response has been suggested in patients treated with gastric electrical stimulation.³⁰ The prevalence of idiopathic gastroparesis is significantly greater than diabetic gastroparesis. However, most trials of novel drugs have focused on patients with diabetes exclusively, possibly because of concern that the idiopathic group may be heterogeneous in terms of underlying mechanisms, leading to variability in response. However, patients with diabetes may also have potential confounding factors, such as fluctuations in glycemic control, which may be an independent contributor to symptoms in patients with diabetes regardless of the treatment. This may lead to a higher placebo response rate in patients with diabetic gastroparesis because active and placebo groups tend to improve their glycemic levels in the controlled setting of a clinical trial. Although other factors, such as choice of endpoint, may also have contributed, this could explain the lower than usual placebo response (23%) in the most recent trial in idiopathic gastroparesis (NORIG).¹⁹ Given these considerations, especially the much greater prevalence of idiopathic gastroparesis, it seems that proof of concept and phase II studies should start with idiopathic or combined idiopathic and diabetic patient population, especially given the documented evidence of similar pathophysiology of these conditions.³¹ Regardless, these considerations emphasize the need for rigorous and careful characterization and screening of patients before and after enrollment.

Summary

This is the start of an exciting new phase in the treatment of disorders of gastric motility, with renewed interest from the pharmaceutical and device industry, helpful FDA guidance, development and validation of meaningful endpoints, and several promising candidates in the pipeline. Along with continuing gains in knowledge about the pathologic and pathophysiological basis of gastroparesis, the future for meaningful advances in the treatment of gastroparesis is brighter than it has ever been.

Abbreviations used in this paper

FDA

Food and Drug Administration

GCSI

Gastroparesis Cardinal Symptom Index

GCSI-DD

Gastroparesis Cardinal Symptom Index Daily Diary

GI

gastrointestinal

MID

minimal important difference

PAGI-SYM

Patient Assessment of Gastrointestinal Disorders–Symptom Severity Index

WMC

wireless motility capsule