Utility of a ¹³C-Spirulina Stable Isotope Gastric Emptying Breath Test in Diabetes Mellitus

Abstract

Background:

The carbon-13 spirulina gastric emptying breath test (GEBT) is approved to identify delayed, but not accelerated, gastric emptying (GE). We compared the utility of the GEBT to scintigraphy for diagnosing abnormal GE in patients with diabetes mellitus.

Methods:

Twenty-eight patients with diabetes ate a 230-kcal test meal labeled with technetium 99m and ¹³C-spirulina, after which 10 scintigraphic images and breath samples (baseline, 15, 30, 45, 60, 90, 120, 150, 180, and 240 minutes) were collected on 2 occasions 1 week apart. We assessed the accuracy of ¹³C-spirulina GEBT excretion rate (percent dose multiplied by 1,000 [kPCD] min⁻¹) values to predict scintigraphic half-life and distinguish between normal, delayed, and accelerated GE and the intraindividual reproducibility of the GEBT.

Key Results:

Scintigraphy revealed normal, delayed, and accelerated GE, respectively, in 17 (30%), 29 (52%), and 10 (18%) test results. GE T½ values measured with scintigraphy and GEBT were highly concordant within individuals; the intraindividual reproducibility was 34% (scintigraphy) and 15% (GEBT). Compared to current criteria, the kPCD150 (150 minutes) and kPCD180 values provided equally sensitive (90%) and more specific (81% vs 67%) approach for distinguishing between delayed vs normal/accelerated GE. A new metric (kPCD60-kPCD15 minutes) was 90% sensitive and 83% specific for distinguishing between accelerated vs normal/delayed GE. These findings were used to create nomograms and an algorithm for interpreting GEBT results.

Conclusions & Inferences:

Among patients with poorly controlled diabetes, the ¹³C-spirulina GEBT can accurately and precisely assess GE and effectively distinguish between normal, delayed, and accelerated GE.

Graphical Abstract

We assessed gastric emptying on two occasions simultaneously with the ¹³C-spirulina gastric emptying breath test (GEBT) and scintigraphy in 28 patients with diabetes mellitus. Scintigraphy revealed normal, delayed, and accelerated GE, respectively, in 17 (30%), 29 (52%), and 10 (18%) test results. The ¹³C-spirulina GEBT accurately and precisely assessed GE and effectively distinguished between normal, delayed, and accelerated GE.

Introduction

Even among patients who are not taking GLP-1 receptor agonists, up to 50% of patients with poorly controlled diabetes mellitus have delayed gastric emptying (GE), which is associated with gastrointestinal symptoms and affects control of glycemia.^1–9 In the Epidemiology of Diabetes Interventions and Complications Study, participants with delayed GE had more office visits than those with normal GE.¹ Also highlighting the relationship between GE and DM, the GLP-1 receptor agonists cause upper GI symptoms and delay GE.¹⁰ Scintigraphy is the gold standard for assessing GE.^11,12 However, few hospitals offer GE scintigraphy in the United States. Between 2018 and 2022, only 118 laboratories in the United States sought accreditation for this test.¹³ Moreover, on average, these laboratories complied with only 8 of 14 variables deemed necessary for high quality GE scintigraphy studies by the American Neurogastroenterology and Motility Society and the Society of Nuclear Medicine and Molecular Imaging.¹³ Hence, alternative options for assessing GE are necessary.

The carbon-13 (¹³C)-spirulina gastric emptying breath test (GEBT) is approved by the United States Food and Drug Administration (FDA) and the Centers for Medicare & Medicaid Services for diagnosing delayed GE, is endorsed by professional societies for this purpose,^11,14,15 and is widely used to evaluate GE in clinical trials for gastroparesis and in clinical practice.^2,16–24 In contrast to scintigraphy, this GEBT (1) can be performed at home or in community-based practices that lack a nuclear medicine facility (2) is regulated by the FDA, hence performed in a standardized manner, (3) does not entail radiation exposure, and (4) is safe for use in breastfeeding women and children.

After the ¹³C-spirulina is digested and moves into the duodenum, the cyanobacteria cell wall dissolves and releases ¹³C-labeled compounds that are absorbed, metabolized, and subsequently exhaled as ¹³C-labeled carbon dioxide (¹³CO₂). GE is the rate-limiting step for ¹³CO₂ excretion.²⁵ There are 4 considerations regarding the utility of the GEBT. First, in the pivotal validation study in which GE was simultaneously evaluated with scintigraphy and the GEBT, only 11 of 129 patients, had diabetes.²⁶ Because ¹³C-spirulina is digested before it is absorbed, diseases such as small intestinal enteropathy, bacterial overgrowth, pancreatic or biliary insufficiency that are associated with diabetes may interfere with the test.²⁷ Second, among patients with diabetes, the day-to-day reproducibility of the GEBT, which is an important consideration,^12,28,29 is unknown.²⁶ Third, while GE scintigraphy tests quantify actual GE over time,¹² the clinical GEBT reports only categorize GE as normal/delayed. Finally, while the GEBT can also identify accelerated GE,^26,30–32 it is not FDA-approved for this purpose.

The aims of this study were to compare GE measured with GEBT vs scintigraphy, to evaluate the sensitivity and specificity of diagnosing delayed and accelerated GE with GEBT vs scintigraphy, and to assess the day-to-day reproducibility of GE measured with scintigraphy and the GEBT in patients with diabetes.

Methods

This prospective study was approved by the Mayo Clinic Institutional Review Board. Towards the aims of investigating the relationship between glycemia and GE, we simultaneously evaluated GE with the ¹³C-spirulina GEBT and with scintigraphy in 28 patients with type 1 and type 2 diabetes.

Eligibility Criteria

Patients were required to have diabetes for 5 years or longer and a HbA1c greater than 7.5% but were not required to have gastrointestinal symptoms. The exclusion criteria include presence of diseases that might interfere with the study, severe nausea, vomiting or a history of malabsorption, abdominal surgery other than appendectomy, cholecystectomy, tubal ligation or hysterectomy, medications that alter GI motility (eg, opioid and GLP1 agonists), allergies to eggs, wheat, or milk or patients likely to undergo major procedures, e.g. pancreas transplantation, dialysis or kidney transplantation within the next 3 months. However, patients taking dipeptidyl peptidase inhibitors (eg, sitagliptin and vidagliptin) were permitted to participate because such drugs do not delay GE.^33,34 Patients with severe nausea and vomiting were excluded because such patients are arguably more likely to vomit during the study. When the entire meal is not ingested and/or is partly vomited during the study, the estimated GE may be inaccurate.

Procedures

Participants completed the Patient Assessment of Gastrointestinal Disorders-Symptom Severity Index and Hospital Anxiety and Depression Questionnaires,^35,36 then underwent 2 GE studies 1 week apart. The participants fasted overnight, then consumed the 230 kcal GE test meal that was labelled with ¹³C-spirulina and technetium-99m–sulphur colloid in 10 minutes (Supplemental Materials).

Scintigraphic images and breath samples were obtained simultaneously at baseline, 15, 30, 45, 60, 90, 120, 150, 180, and 240 minutes after the meal.^26,30 We collected samples at 15, 30, and 60 minutes, which are not in the FDA-approved GEBT kit, because we sought to distinguish between normal and accelerated GE.³⁰ The carbon-13 labeled carbon dioxide (¹³CO₂):¹²CO₂ ratio in breath samples were analyzed with gas isotope ratio mass spectrometry in a central laboratory (Cairn Diagnostics, Brentwood, Tennessee). Using established formulae, the rate of ¹³CO₂ excretion was calculated by assessing the change in this ratio over time vs baseline (Supplemental Methods).³⁷ Established linear regression equations were used to predict the proportionate GE at each time point from the corresponding ¹³CO₂ excretion rate (percent dose multiplied by 1,000 [kPCD] min⁻¹) (Supplemental Table 1).²⁶ The GE T½ was estimated with linear interpolation during the GEBT and scintigraphy.^{8, 13}

Statistical Analysis

Three approaches were used to analyze GEBT data: GE T½, individual kPCD (percent dose multiplied by 1,000) values or derivations thereof, and the FDA-approved interpretative guideline criteria for delayed GE. The GE T½ and estimated GE values were interpreted relative to 10th to 90th percentile values in 50 healthy persons who underwent combined GEBT-scintigraphy with this meal. For the GE T½, the normal values ranged from 51.1 to 82.6 minutes. The kPCD values at each time point were interpreted relative to 10th to 90th percentile values in the 50 same healthy persons mentioned above who underwent both scintigraphy and GEBT in the GEBT calibration study or relative to FDA-approved reference range cut-off points.²⁵ (Supplemental Table 2). By cutting off the top and bottom 10% of data, the 10–90th range is less impacted by extreme values that could skew the interpretation of “normal” if they were included in a wider 5–95th range.

The performance characteristics for GEBT vs scintigraphy were evaluated with Spearman’s correlation coefficient and Bland Altman test for continuous variables ³⁸ and with Cohen’s κ statistic for categorical variables (ie, normal, delayed, or accelerated GE). Logistic regression models were used to estimate the performance characteristics for distinguishing between delayed and normal/accelerated GE and between accelerated and normal/slow GE. Intraindividual coefficient of variation (CV) was calculated as the root mean square within patients, divided by the mean across patients. The interindividual CV was calculated as the square root of the estimated component of variance between patients divided by the mean. The component of variance between patients was estimated as ½ (mean squared between – mean squared within).

Results

Demographic and Clinical Characteristics

Eighteen of 28 patients (17 women, 11 men) had type 1 diabetes. The mean (SD) age and BMI were 63(25)y and 29.4(5.4) kg/m². Thirteen (46%) had neuropathy, 17 (61%) had retinopathy, and 11 (39%) had nephropathy. All patients had poorly controlled glycemia with a mean (SD) HbA1c of 9.1 (0.81)%. Patient Assessment for Gastrointestinal Disorders-Symptom Severity Index and Patient Assessment for Gastrointestinal Disorders-Quality of Life questionnaire scores were 1.2 (1.1) and 3.8 (1.0), respectively. Only 1 patient had a fasting blood glucose ≥275 mg/dL before the GE study.

GE T½

Detailed performance characteristics for GE measured with scintigraphy and estimated with GEBT at various timepoints are in Table 1. There was strong correlation between scintigraphy and GEBT GE T½ (Table 1, Figure 1). At visit 1, the mean (SD) GE T½ was 103 (60) minutes for scintigraphy and 109 (40) minutes for GEBT (Table 1). At visit 2, the mean (SD) GE T½ were scintigraphy, 90 (49) minutes and GEBT, 106 (39) minutes. In the Bland Altman plot, observe that the T½ difference (scintigraphy – GEBT) was negative for T½ values less than 52.5 but positive for greater values (P=.001) (Figure 1B). The mean difference (scintigraphy – GEBT) in T½ was −11.2 minutes. Figure 1E shows representative curves for normal, accelerated, and delayed GE.

Table 1.

Comparison of GE Measured With GEBT and Scintigraphy

GE, min	Scintigraphy GE %, Mean (SD)		GEBT GE %, Mean (SD)a		GEBT vs scintigraphy, Spearman’s ρ		Intraindividual CV, %		Interindividual CV, %
	Visit 1	Visit 2	Visit 1	Visit 2	Visit 1	Visit 2	Scintigraphy	GEBT	Scintigraphy	GEBT
45	31 (15)	33 (13)	24 (9)	24 (8)	0.70b	0.66b	38	22	23	25
90	51 (21)	55 (16)	46 (17)	46 (16)	0.82b	0.77b	22	15	28	31
120	62 (21)	67 (18)	59 (19)	60 (18)	0.78b	0.77b	15	11	27	29
150	70 (22)	76 (18)	69 (19)	71 (19)	0.72b	0.70b	14	9	24	25
180	78 (22)	82 (18)	78 (19)	80 (19)	0.65b	0.62b	13	8	22	22
240	88 (17)	89 (19)	89 (14)	90 (14)	0.54	0.55	10	6	18	15
T½,	103 (60)	90 (49)	109 (40)	106 (39)	0.85b	0.83b	34	15	46	34

Figure 1.

Comparison of GE T½ Assessed with Scintigraphy and GEBT. Observe strong correlation between T½ assessed with both methods (A) and moderate correlation between visits 1 and 2 (C and D). B, The Bland Altman test was significant (P=.001). In parts A, C, and D the rectangles (red dotted lines) show normal T½ values (51.1 to 82.6 minutes). Open and closed circles show patients in whom the GE diagnosis similar and different between the x and y axis. E, Representative GE curves. GE, gastric emptying; GE T½, gastric emptying half time; GEBT, gastric emptying breath test; kPCD, percent dose multiplied by 1,000.

Diagnosis of Delayed GE

Compared with scintigraphy, the GEBT T½ was 93% sensitive and 63% specific while the FDA-approved interpretative guideline criteria were 90% sensitive and 67% specific for distinguishing between delayed and normal or accelerated GE diagnosed with scintigraphy (Table 2). For the newer, kPCD-based criteria, the sensitivity and specificity ranged from 72% to 93% and from 74% to 96%, respectively (Table 2). Among these, the kPCD150 plus kPCD180 model provides excellent sensitivity (90%) without compromising specificity (81%). But this sensitivity and specificity apply to a specific combination of kPCD150 (ie, 58.2 min⁻¹) and kPCD180 values (67.2 min⁻¹). Equation 1 is a logistic regression model which provides the probability of delayed GE for all combinations of kPCD150 and kPCD180 values.

$$\begin{gathered} {\mathrm{Probability}}_{\text{Delayed}}=\exp \left({\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{s}}\right)/\left[1+\exp \left({\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{s}}\right)\right], \\ \mathrm{where}{\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{s}}=4.279-{0.433}^{\mathrm{*}}\mathrm{kPCD}150+{0.313}^{\mathrm{*}}\mathrm{kPCD}180 \end{gathered}$$ Equation 1

Table 2.

Performance Characteristics of GEBT Criteria for Distinguishing Delayed from Normal or Accelerated GE

	GE T½	FDA approved interpretative guideline criteria b	kPCD90 and kPCD 120c	kPCD90, kPCD120 and kPCD 150c	kPCD45, kPCD150 and kPCD180c	kPCD150 and kPCD180b	kPCD180–1.4 * kPCD150c	kPCD120, kPCD150, and kPCD180c
True positive, No.a	27	26	21	27	26	26	26	23
True negative, No.a	17	18	25	20	22	22	22	26
False positive, No.	10	9	2	7	5	5	5	1
False negative, No.	2	3	8	2	3	3	3	6
Total, No.	56	56	56	56	56	56	56	56
Negative, %	34	38	59	39	45	45	45	57
Positive, %	66	62	41	61	55	55	55	43
Performance statistics
Sensitivity, %	93	90	72	93	90	90	90	79
Specificity, %	63	67	93	74	81	81	81	96
Positive predictive value, %	73	74	91	79	84	84	84	96
Negative predictive value, %	89	86	76	61	88	88	88	81
Area under receiver operating characteristic curve	0.92	0.78	0.89	0.90	0.91	0.91	0.91	0.91

All metrics in this table are derived from one or more kPCD values (ie, at 45, 90, 120, 150, 180, and 240 minutes) that are currently available in the commercial, FDA-approved GEBT.

^aTrue positives had delayed GE by GEBT and scintigraphy. True negatives did not have delayed GE by GEBT and scintigraphy.

^bDelayed if kPCD90 or kPCD120 or kPCD150 are less than the delayed cut-off point (Supplementary Table 2) and/or if the kPCD240 is the maximum excretion rate.

^cBased on logistic regression models.

The term X’Bs is the linear combination of predicted variables weighted by their coefficients. Substituting the X’B_S term from Equation 1 gives us Equation 2.

$${\mathrm{Probability}}_{\text{Delayed}}=\exp (4.279-0.433\mathrm{x}+0.313\mathrm{y}/(1+\exp (4.279-0.433\mathrm{x}+0.313\mathrm{y})$$ Equation 2

The actual kPCD150 and kPCD180 data can be used to compute the contrast term (0.433*kPCD150 + 0.313*kPCD180) in Equation 1. Values ≥ −14.28 and < −14.28 suggest delayed and normal or accelerated GE. Separately, Figures 2A and 2B provide the nomogram, which was derived from Equation 2, that estimates the probability of delayed GE for every combination of kPCD150 and kPCD180 values. The algorithm was based on a threshold value which was derived from kPCD150 and kPCD180 values, optimizing sensitivity and specificity for distinguishing between delayed and normal/accelerated GE. This threshold was derived from an equation which incorporates the ratio of the kPCD150 and kPCD180 terms in this Equation 1, i.e., −0.433/0.313, which is −1.4 (Supplemental Methods). Similar to the model shown in Equation 1, the term (i.e., kPCD180 – 1.4*kPCD150) was also 90% sensitive and 81% specific for distinguishing between delayed vs normal/accelerated GE (Figure 3).

Figure 2.

Nomograms to Predict Probability of Delayed (A and B) and Accelerated GE (C). B, Magnified portion of A. All nomograms show 3 lines for predicted probability of delayed or accelerated GE. The red dotted lines (A and B) show the lower limit of normal for kPCD150 (x axis) and kPCD180 (y axis). The scintigraphy T½ values are adjacent to each point.

Figure 3.

Diagnosis of GE Disturbances With GEBT. The ROC curves depict excellent performance characteristics for distinguishing delayed from normal/accelerated GE (A) and accelerated from normal/delayed GE diagnosed with scintigraphy (B). The algorithm (C) applies the accelerated then the delayed GE nomograms. D, The distribution of scintigraphy GE T½ vs GEBT algorithm-based categories. The red dotted lines are the 10th and 90th percentile values for scintigraphy GE T½. AUC indicates area under the curve; ROC, receiver operating characteristic.

The probability of delayed GE for the total of the 56 test results was ≤53% in 26 test results, between 53% and 79% in 10 test results, and ≥80% in 20 test results (Figure 2). Among test results in which the probability of delayed GE was ≤53%, 53% to 79%, and ≥ 80%, respectively, scintigraphy disclosed delayed GE in 15% (4/26), 60% (6/10), and 95% (19/20) test results (P< .001). In the ≤53%, 53% to 79%, and ≥80% groups, the actual scintigraphy GE T½ only among test results in which GE was delayed was 94 (9) (n=4), 120 (45) (n=6), and 148 (53) (n=19) minutes, respectively (P=.10).

The delayed GE nomogram (Figures 2A and 2B) can be divided into 4 quadrants by the FDA-approved cut-off points (Supplemental Table 2) that are used to identify delayed GE. Of note, even relatively small differences in the kPCD values can be associated with considerable differences in the probability of delayed GE. For example, in 2 test results with identical kPCD180 values (38 min⁻¹), the kPCD150 values were 36.2 min⁻¹ and 31.6 min^-1. Although the difference between these values was only 4.6 units, the nomogram suggested that the probability of delayed GE for kPCD150 values of 36.2 min⁻¹ and 31.6 min⁻¹ were 61% vs 92%, respectively (Figure 2). Indeed, the scintigraphy T½ values were 73 and 123 minutes in these 2 test results with a predicted probability of 61% and 92%.

Performance Characteristics for Diagnosis of Accelerated GE

Increased kPCD45, kPCD90, or kPCD120 values, which are samples that are collected in the FDA-approved GEBT, were insensitive (3%) but highly specific (92%) for distinguishing accelerated from normal/delayed GE (Table 3). By comparison, the new metrics devised in this study were more sensitive but slightly less specific. The difference (kPCD60 - kPCD15) has the best performance characteristics (90% sensitivity and 83% specificity) for distinguishing between accelerated vs normal/delayed GE. For this model, the probability of accelerated GE is as follows:

$$\begin{gathered} {\text{Probability}}_{\mathrm{Accelerated}}=\exp \left({\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{F}}\right)/\left[1+\exp \left({\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{F}}\right)\right], \\ \mathrm{where}{\mathrm{X}}^{\mathrm{\text{'}}}{\mathrm{B}}_{\mathrm{F}}=-7.084-{0.3089}^{\mathrm{*}}{\mathrm{kPCD}}_{15}+{0.3025}^{\mathrm{*}}{\mathrm{kPCD}}_{60} \end{gathered}$$ Equation 3

Table 3.

Performance Characteristics of Different GEBT End Points for Distinguishing Accelerated from Normal or Delayed Gastric Emptying

Test characteristic	Increased GEBT45 and/or GEBT90, and/or GEBT120 1	Increased kPCD45, and/or kPCD90, and/or kPCD120 a	kPCD45: kPCD240 b	kPCD90: kPCD240b	kPCD30 and kPCD45b	kPCD60 – kPCD15b	kPCD60 - kPCD30b
True positive, No.	2	1	8	9	9	9	9
True negative, No.	23	24	34	34	37	38	37
False positive, No.	2	2	11	11	9	8	9
False negative, No.	29	29	2	1	1	1	1
Total, No.	56	56	55	55	56	56	56
Percentage negative, %	93	95	65	64	68	70	68
Percentage positive, %	7	5	35	36	32	30	32
Performance statistics
Sensitivity, %	6	3	80	90	90	90	90
Specificity, %	92	92	76	76	80	83	80
Positive predictive value, %	50	33	42	45	50	53	50
Negative predictive value, %	44	45	94	97	97	97	97

From this table, kPCD values at 45, 90, 120, and 240 minutes are currently available in commercial, FDA-approved GEBT reports. But, samples at 15, 30, and 60 minutes are not in the FDA-approved GEBT kit.

^aDefined as any of these values is greater than the 90th percentile value in healthy persons.

^bBased on logistic regression models.

Based on the accelerated GE nomogram (Figure 2C), the probability of accelerated GE was ≤17%, 17% to 79%, and ≥80% in 39, 14, and 3 test results, respectively. These 3 lines represent 3 selected points on the ROC curve (Figure 3). The uppermost line represents high specificity (100%) and lower sensitivity (33%) and the lowermost represents high sensitivity (89%) and lower specificity (19%). These sensitivities and specificities are, respectively, the proportion of dark circles that fall above the line and the proportion of open circles that fall below the line. But the predictive probability associated with each line (17% in the case of the lowermost line), is the model-based estimate of the conditional probability of being a case (rapid GE) if combination of kPCD15 and kPCD60 values falls exactly on the line. Since the decision rule associated with this point on the ROC curve identifies the participant as a case for any set of values above and to the left of the line, this conditional probability of 17% is the extreme (lowest possible) value among all points that are identified as cases for that decision rule. Points above the line have a greater probability of being a case, and the higher they are above the line, the greater the predictive probability. Although 17% is seemingly low, it is comparable to the overall proportion (18%) of studies with rapid GE in this cohort.

In the ≤17% group, scintigraphy disclosed accelerated GE in 1 of 39 test results (3%); the GE T½ mean(SD) was 115(55) minutes. In the 17% to 79% group, GE was accelerated in 6 of 14 test results (43%); the GE T½ was 56(14) minutes. In the ≥80% group, GE was accelerated in all 3 test results (100%); the GE T½ was 41(3) minutes. The mean(SD) scintigraphy GE was different among these 3 groups (P< .001).

Comparison of FDA-Approved Interpretative Guideline Criteria and Nomograms

We sequentially applied Equation 3, which uses the kPCD60-kPCD15 metric to assess for accelerated vs nonaccelerated GE, then Equation 2, which uses the kPCD180–1.4*kPCD150 metric, to assess for delayed vs normal GE (Figure 3). The algorithm predicted that GE was accelerated in 17 studies. Among these, the scintigraphy disclosed accelerated GE (9[53%]), normal GE (7[41%]), or delayed GE (1[6%]). Among the remaining 39 test results, the algorithm predicted delayed GE in 31 test results, among which scintigraphy disclosed delayed GE in 26 results (84%). In 8 results, the algorithm suggested normal GE; scintigraphy also disclosed normal GE in 6 test results (75%).

Table 4 compares the utility of our algorithm, GEBT T½ values, and GEBT kPCD values that are part of the FDA-approved test vs scintigraphy. There are 2 key findings. All 3 GEBT-based criteria were highly sensitive but less specific for predicting delayed GE identified by scintigraphy. The algorithm was the best of 3 approaches (ie, T½, FDA-approved interpretative guideline criteria and kPCD-based criteria) for identifying GE determined by scintigraphy. It accurately predicted delayed GE in 26 of 29 (90%) test results with delayed GE and accelerated GE in 9 of 10 (90%) test results by scintigraphy. However, among 17 test results in which the scintigraphy GE was normal, the algorithm suggested that GE was accelerated in 7(41%) test results. The scintigraphy GE was also normal in 4 of 31(13%) test results in which the algorithm predicted delayed GE. The scintigraphy T½ was significantly different between test results in which the algorithm accurately predicted vs did not accurately predict GE by scintigraphy (Figure 3D). For example, among 31 test results in which the algorithm predicted delayed GE, the T½ was longer (P₌.002) when the algorithm was accurate vs inaccurate.

Table 4.

Comparison of GE Assessed With Scintigraphy and GEBT

	Scintigraphy (N=56)			Agreement between GEBT and scintigraphy, κ (95% CI)
GEBT diagnosis (criteria) a	Accelerated (n=10)	Normal (n=17)	Delayed (n=29)	Visit 1	Visit 2	Overall
Based on T½				0.47 (0.18 – 0.75)	0.40 (0.10 – 0.70)	0.43 (0.26 – 0.64)
Accelerated, n=0 b	0	0	0
Normal, n=19	8	9	2
Delayed, n=37	2	8	27
Based on kPCD values at 45, 90, 120, 150, 180 and 240 minutes c
Accelerated, n=3 d	1	2	0
Normal (FDA interpretative guideline criteria), n=18 e	7	8	3	0.19 (−0.13 – 0.52) f	0.21 (−0.11 – 0.52) f	0.20 (−0.02 – 0.43) f
Delayed (FDA interpretative guideline criteria) n=35 e	2c	7 d	26 d
Based on algorithm g				0.60 (0.35 – 0.86)	0.56 (0.31 – 0.82)	0.58 (0.40 – 0.76)
Accelerated, n=17	9	7	1
Normal, n=8	0	6	2
Delayed, n=31	1	4	26

^aNumber of test results in which the GEBT satisfied criteria for normal, delayed or accelerated GE as appropriate

^bThe lowest GEBT T½ was 52.4 minutes (ie, GEBT T½ 10th percentile value of 51.1 minutes). Hence, no patients had accelerated GE based on GEBT T½.

^cBased on results at these times.

^dkPCD45 and/or kPCD90 and/or kPCD120 are >90th percentile value.

^eThese criteria categorize test results as normal or delayed.

^fAgreement between GEBT and scintigraphy for diagnosis of normal or delayed GE.

^gBased on algorithm (Figure 3C).

Intraindividual and Interindividual CV

For GE T½, the intraindividual CV was 34% and 15% for scintigraphy and the GEBT, respectively, while the interindividual CV for scintigraphy and the GEBT was 46% and 34%, respectively (Table 1). By scintigraphy, the GE diagnosis was identical in both test results in 43% of patients. Based on the GEBT algorithm shown in Figure 3, the overall GE diagnosis category was identical in both visits in 18 of 28 (43%) patients.

Discussion

GE was simultaneously measured with scintigraphy and a GEBT on 2 occasions in 28 patients with poorly controlled, generally complicated, type 1 or type 2 diabetes. Among 56 scintigraphy test results, GE was normal, delayed, and accelerated in 30%, 52%, and 18% of patients, respectively. There was strong correlation between the GEBT and scintigraphy T½ values even though the equations that were used to estimate the GE from the GEBT were validated in an earlier study in which most participants did not have diabetes.²⁶ (The correlation between GE(%) measured with scintigraphy and the GEBT declines after 120 minutes because after 150 or 180 minutes, the ¹³CO₂ excretion rate decreases in persons with normal GE but increases in persons with delayed GE. This does not impact the clinical diagnostic utility of the GEBT as shown in Table 2.)

The FDA-approved interpretative guideline criteria were 90% sensitive and 67% specific for detecting delayed GE vs scintigraphy. These criteria suggested that GE was delayed in 9 of 27 patients with normal/accelerated GE by scintigraphy. By comparison, the combination of kPCD150 and kPCD180 values was more sensitive (90%) and specific (81%) for distinguishing between delayed vs scintigraphic normal/accelerated GE. These values are very similar to an earlier study.²⁶ However, these performance characteristics only apply to the optimum cut-off point on the ROC curves, ie, the point at which (sensitivity – [1-specificity]) has the greatest value. Hence, we created a delayed GE nomogram to estimate the probability of delayed GE for all combinations of kPCD150 and kPCD180 values. When the probability of delayed GE in this nomogram was ≤53%, 53% to 79%, and ≥80%, the actual probability of scintigraphic delayed GE was respectively low (15%), intermediate (60%), and high (95%), which suggests that the actual prevalence of delayed GE closely approximates to the probability across a range of values. Moreover, novel indices derived from kPCD15, kPCD30, and kPCD60 values, which are not collected in the commercially available GEBT, were more useful than the kPCD45, kPCD90, and kPCD120 values for diagnosing accelerated GE. Among these newer indices, the difference (kPCD60 – kPCD15) is the most useful for diagnosing accelerated GE. In the nomogram showing accelerated GE, when the probability of accelerated GE was ≤17%, 17% to 79%, and ≥80%, scintigraphy disclosed accelerated GE in 3%, 43%, and 100% of test results, respectively.

We devised an algorithm that applies kPCD15 and kPCD60 values to identify accelerated GE then kPCD150 and kPCD180 values to identify delayed GE. Test results that neither satisfy criteria for accelerated nor delayed GE are considered to represent normal GE. Compared to scintigraphy, the algorithm accurately predicted delayed GE in 26 of 29 (90%) test results and accelerated GE in 9 of 10 (90%) test results. However, scintigraphy disclosed normal GE in 7 of 17 (41%) test results in which the algorithm suggested that GE was accelerated and normal GE in 4 of 31 (13%) test results in which the algorithm predicted delayed GE. Among these test results in which the algorithm did not predict the GE diagnosis, the GE abnormality was less pronounced than in test results in which the outcomes of the algorithm and scintigraphy were concordant.

The kPCD values are expressed on a continuous scale. But the current ¹³C-spirulina GEBT test reports interpret kPCD values categorically as normal or delayed based on whether they are lower than the reference range specified by the FDA-approved label. These reports do not identify accelerated GE. By contrast, the nomogram and our proposed algorithm interpret the actual kPCD values in a continuous manner. They estimate the likelihood of not only delayed but also accelerated GE. Hence, this algorithm not only estimates the probability of delayed (or accelerated GE) but also implicitly estimates the severity of delayed (or accelerated) GE.

The Bland-Altman Plot for GEBT vs scintigraphy T½ shows heteroscedascity (i.e. estimated T½ for scintigraphy is relatively slower than GEBT for patients with fast gastric emptying and relatively faster for patients with delayed gastric emptying. Physiological factors may explain these differences. In persons with rapid GE, it is possible that GE is not the rate determining step in the process of the ¹³C label being transported from the meal to the breath. Perhaps GE occurs so quickly that the absorption and/or metabolism steps in the process become the determinant rates in the process. Another possible reason for the heteroscedascity is that the patient data set used to generate the multiple linear regression equations that are used to calculate T½ included only participants with delayed not rapid GE. The GEBT was validated and approved by FDA to determine “normal” vs “delayed” emptying. While the overall correlation between breath tests and scintigraphy, as well as the diagnostic agreement of scintigraphy with breath fraction emptied or t50 values, is generally consistent, some data points do not align perfectly, and we lack a clear explanation for these discrepancies.

The ¹³C-spirulina GEBT is the only FDA-approved and commercially available GEBT in the United States. It is approved for home-testing and for use in children. It is arguably the most widely used GE test in multicenter clinical trials. Thes findings also have several clinical implications, especially given concerns regarding delayed GE and the risk of aspiration in patients with DM who are being treated with GLP-1 agonists.¹⁰ First, they facilitate a more accurate assessment of the presence and severity of delayed GE. Second, the novel metrics suggest that kPCD150 and kPCD180 values are preferable to the current approach, which relies upon kPCD90, kPCD120, and kPCD150 values to diagnose delayed GE. Third, they show that additional breath samples at 15 and 60 minutes can be used to diagnose accelerated GE with the GEBT. This is a useful advance because accelerated GE cannot be distinguished from delayed GE by symptoms or the diabetic phenotype.^31,39 Fourth, germane to the ongoing debate about the intraindividual reproducibility of GE measurements and the distinction between dyspepsia and gastroparesis,^12,29 it is reassuring that the day-to-day reproducibility of GE measurements determined with the GEBT is excellent. Among patients with diabetes, the intraindividual CV for GE T½ measured with this 230 kcal scintigraphy meal was 34%, which is similar to the corresponding values of 33% and 37.9% for a 320 kcal meal (30% fat) in 21 patients with gastroparesis and 18 with functional dyspepsia,¹² but greater than in 21 patients with diabetes who were studied with a 320 kcal meal in an earlier study.²⁸ In the present study, the intraindividual reproducibility of GE T½ measured with the GEBT was less compared with scintigraphy.

Among several strengths, the patient population was representative of the poorly controlled, complicated patients with diabetes who undergo GE studies. GE was delayed in 25 of 56 (45%) test results. In 17 patients (30%), the GE T½ was >20 minutes above the upper limit of normal, indicative of markedly delayed GE. However, these patients had relatively mild upper gastrointestinal symptoms, which is not unusual among patients with diabetes mellitus.^1,32 The ability of this algorithm for distinguishing between normal and accelerated GE needs to be improved. Although patients who were taking GLP-1 or opioid agonists were excluded from this study, we anticipate that scintigraphy and GEBT should be equally accurate for detecting delayed GE in such patients.

In summary, GE can be reliably assessed with the ¹³C-spirulina GEBT in patients with diabetes. GE T½ values measured with scintigraphy and the GEBT correlated with each other. An algorithm based on 4 breath samples is useful for identifying not only delayed but also accelerated GE with the GEBT.

Key Points.

• The ¹³C-spirulina GEBT is FDA-approved for diagnosing delayed but not accelerated gastric emptying (GE). While this test has some advantages vs scintigraphy, it has not been thoroughly validated in patients with diabetes mellitus.

• Among patients with diabetes mellitus, we evaluated the sensitivity and specificity of this GEBT for diagnosing delayed and accelerated GE and assessed the day-to-day reproducibility of these measurements.

• There was strong positive correlation between GE assessed with scintigraphy and this GEBT. Using innovative metrics, this GEBT accurately and precisely assessed GE and effectively distinguished between normal, delayed, and accelerated GE.

Data availability:

Data are available from the corresponding author upon reasonable request.