Assessing different diagnostic tests for gastroesophageal reflux disease: a systematic review and network meta-analysis.

BACKGROUND: The aim of the current systematic review and network meta-analysis (NMA) was to assess the diagnostic characteristics of the gastroesophageal reflux disease questionnaire (GERDQ), proton-pump inhibitor (PPI) test, baseline impedance, mucosal impedance, dilated intercellular spaces (DIS), salivary pepsin, esophageal pH/pH impedance monitoring and endoscopy for gastroesophageal reflux disease (GERD).
METHODS: We searched PubMed and the Cochrane Controlled Trial Register database (from inception to 10 April 2018) for studies assessing the diagnostic characteristics of the GERDQ, PPI test, baseline impedance, mucosal impedance, DIS, or salivary pepsin and esophageal pH/pH impedance monitoring/endoscopy in patients with GERD. Direct pairwise comparison and a NMA using Bayesian methods under random effects were performed. We also assessed the ranking probability.
RESULTS: A total of 40 studies were identified. The NMA found no significant difference among the baseline impedance, mucosal impedance, and esophageal pH/pH impedance monitoring and endoscopy in terms of both sensitivity and specificity. It was also demonstrated that the salivary pepsin detected by the Peptest device had comparable specificity to esophageal pH/pH impedance monitoring and endoscopy. Results of ranking probability indicated that esophageal pH/pH impedance monitoring and endoscopy had highest sensitivity and specificity, followed by mucosal impedance and baseline impedance, whereas GERDQ had the lowest sensitivity and PPI test had the lowest specificity.
CONCLUSIONS: In a systematic review and NMA of studies of patients with GERD, we found that baseline impedance and mucosal impedance have relatively high diagnostic performance, similar to esophageal pH/pH impedance monitoring and endoscopy.

Introduction

Gastroesophageal reflux disease (GERD) is one of the most common healthcare issues, with an estimated worldwide prevalence of up to 33%[1,2] resulting in a heavy economic burden of approximately $13.0 billion/year to the healthcare system in the USA alone, due to the different diagnostic testing and overuse of proton-pump inhibitors (PPIs) to a great extent.[3] Albeit with the advances of diagnostic tests for GERD, the lack of a ‘gold standard’ has made identifying patients with GERD one of the biggest dilemmas in clinical practice.

Although GERD is generally empirically diagnosed based on typical reflux symptoms (heartburn and regurgitation),[4,5] the sensitivity and specificity of the symptom-based diagnosis of GERD is limited due to the complex symptom spectrum for GERD.[6] Patients with suspected GERD symptoms are often first tested for a response to PPI therapy, which definitely results in unnecessary overuse of PPIs because of its high placebo effect and low specificity.[7] So far, upper endoscopy and esophageal pH/pH impedance testing are usually performed to detect GERD complications, as well as documentation of the presence of reflux for an objective GERD diagnosis.[8] In order to develop a better understanding of the pathophysiology and improve appropriate GERD diagnosis, several new diagnostic tests, such as baseline impedance, esophageal mucosal impedance, salivary pepsin, and histopathology have been developed in recent years.[9]

To the best of our knowledge, there has been little published information regarding the comparison of diagnostic performance among individual tests. Therefore, we performed a systematic review and network meta-analysis (NMA) to assess the diagnostic characteristics of the GERDQ questionnaire, PPI test, baseline impedance, mucosal impedance, dilated intercellular spaces (DIS), salivary pepsin, esophageal pH/pH impedance monitoring and endoscopy for GERD.

Methods

Search strategy

An electronic and manual search of PubMed and the Cochrane Controlled Trial Register database for relevant articles from inception to April 2018 was performed by two authors independently. The combination of keywords and free text including: GERD Questionnaires; omeprazole, lansoprazole, pantoprazole, rabeprazole, esomeprazole, and PPI test; baseline impedance; mucosal impedance; dilated intercellular spaces; pepsin; esophageal pH/pH impedance monitoring; and endoscopy were used as search terms of different diagnostic tests for GERD. Additional search terms were: GERD or GORD, expanded to diagnosis, screening, reproducibility of results, sensitivity and specificity, false-negative reactions, false-positive reactions, predictive value, accuracy, and likelihood ratio.[10]

The searches were limited to English- or Chinese-language studies performed in adults (age > 18 years). Only data accessible in peer-reviewed journals were included to minimize potential sources of bias and inaccuracy.[11]

Inclusion criteria and exclusion criteria

Studies were screened for inclusion and final decisions on exclusion were made by two authors independently. Studies assessing the diagnostic characteristics of the GERDQ, PPI test, baseline impedance, mucosal impedance, DIS, salivary pepsin, esophageal pH/pH impedance monitoring, or endoscopy in adults with presumptive GERD were screened to be included. Studies were excluded if they focused only on children (age < 18 years) or patients who have specific diseases (such as cardiovascular disease) or who have had operations, if they focused exclusively on patients with extraesophageal GERD symptoms (such as asthma or laryngitis).

Data extraction

Data from the included studies were extracted by two authors independently. Information extracted included patient characteristics, study design, setting, gold standard and diagnostic modalities, and definitions of outcomes. Numbers were extracted directly from the tables or derived from percentages if only the total number of patients was available. Discrepancies were resolved by discussion until consensus was achieved for all data.

Quality assessment of studies

The quality of all included studies was assessed by researchers, according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool.[12] The QUADAS-2 tool included the following four key domains: patient selection, index test, reference standard, and flow of patients through the study, and timing of the index tests and reference standard (flow and timing). Review Manager 5 (RevMan 5.2.3, Cochrane Collaboration, Oxford, UK) statistical computing software was used to carry out quality assessment and investigation of publication bias.

The positive standard of diagnostic tests for GERD

The following standards for GERD were used in the current study, all of which were based on commonly accepted measures (Table 1).

Table 1.

The information of included studies.

Comparison (author)	Evaluable patients (n)	Men (%)	Average age	Prevalence (%) of esophagitis	Reference tests					Study test	Outcome measure extracted
					Upper endoscopy	pH/pH impedance monitoring
						AET	Number of reflux events	DeMeester score	Symptom reflux association
A versus B
Frazzoni et al.[13]	121	51.89	NA	42.33	Erosion	Total AET > 3.2%	>48	—	SAP ⩾ 95% or SI ⩾ 50%	MNBI	A cut-off value of 2292 Ω
Ravi et al.[14]	55	41.38	60	22.73	—	Total AET ⩾ 5.0%	—	—	—	Baseline impedance	A cut-off value of 2268 Ω
Frazzoni et al.[15]	289	45.67	50	23.53	Erosion	Total AET >3.2%	—	—	SAP ⩾ 95% or SI ⩾ 50%	MNBI	A cut-off value of 2292 Ω
Kandulski et al.[16]	36	30.77	55	30.77	Erosion	Total AET > 4.2%	—	—	SAP ⩾ 95% or SI ⩾ 50%	Baseline impedance	A cut-off value of 2100 Ω
A versus C
Hayat et al.[17]	198	44.14	49.7	NA	—	Total AET > 4.2%	—	—	SAP ⩾ 95%	Salivary pepsin	At least one sample +ve > 16 ng/ml
Saritas Yuksel et al.[18]	47	65.52	50	23.4	Erosion	Total AET > 4.2%	—	—	—	Salivary pepsin	A cut-off value of +1.0 or greater
A versus D
Cui et al.[19]	565	40.7	49.4	NA	Erosion	—	—	⩾14.7	SAP ⩾ 95%	DIS	A cut-off value of 0.9 μm
Zhou et al.[20]	576	41.67	49.3	27.04	Erosion	—	—	⩾14.7	SAP ⩾ 95%	DIS	A cut-off value of 0.9 μm
Cui et al.[21]	161	59.66	NA	48.74	Erosion	—	—	⩾14.7	—	DIS	A cut-off value of 0.85 μm
Mastracci et al.[22]	139	57.14	52	40.34	Erosion	Total AET > 5.5%	—	—	—	DIS	Mild and marked DIS
Vela et al.[23]	26	NA	NA	11.54	Erosion	Total AET > 5.3%	—	—	SAP ⩾ 95%	DIS	A cut-off value of 0.68 μm
A versus E
Zhou et al.[20]	636	41.67	49.3	27.04	Erosion	—	—	⩾14.7	SAP ⩾ 95%	GERDQ	A cut-off score of 8
Zavala-Gonzales et al.[24]	252	37	49.49	43.65	Erosion	Total AET > 4.2%	—	—	SI ⩾ 50%	GERDQ	A cut-off score of 8
Jonasson et al.[25]	169	53	47.4	81.07	Erosion	Total AET > 5.5%, or upright AET > 6.7%, or supine AET > 6.9%	—	—	SAP ⩾ 95%	GERDQ	A cut-off score of 8
Lacy et al.[26]	177	28.2	51.00	16.87	—	Total AET > 5.3%	—	—	—	GERDQ	A cut-off score of 8
Jones et al.[27]	308	46.43	47	37.66	Erosion	Total AET > 5.5%	—	—	SAP ⩾ 95%	GERDQ	A cut-off score of 8
Bai et al.[28]	8065	50.1	46.3	17.79	Erosion	—	—	—	—	GERDQ	A cut-off score of 8
A versus F
Zhou et al.[20]	636	41.67	49.3	27.04	Erosion	—	—	⩾14.7	SAP ⩾ 95%	Esomeprazole, 20 mg BID for 2 weeks	Complete symptom relief
Jonasson et al.[25]	124	53	47.4	81.07	Erosion	Total AET > 5.5%, or upright AET > 6.7%, or supine AET > 6.9%	—	—	SAP ⩾95%	PPI for 4 weeks	Symptoms decrease to 1 day/week
Lee et al.[29]	188	41.49	49.9	37.23	Erosion	—	—	—	—	Lansoprazole, 15 mg/30 mg/60 mg once daily for 2 weeks	Symptom score improved by more than 50%
Dent et al.[30]	296	46.43	47	37.66	Erosion	Total AET > 5.5%	—	—	SAP ⩾ 95%	Esomeprazole, 40 mg daily for 14 ± 3 days	Complete symptom relief
Cho et al.[31]	73	57.53	47	63.01	Erosion	Total AET > 4.4%	—	—	SAP ⩾ 95%	Lansoprazole, 30 mg BID for 2 weeks	Symptom score improved by more than 50%
Zheng et al.[32]	27	29.63	57	3.7	Erosion	Total AET > 4.0%	—	⩾ 14.0	SI ⩾ 50%	Esomeprazole, 20 mg BID for 2 weeks	NA
Lee et al.[33]	164	43.29	NA	42.07	Erosion	—	—	—	—	Rabeprazole, 20 mg/40 mg for 2 weeks	50% symptom reduction
Fan et al.[34]	32	37.5	47.3	0	—	—	—	⩾14.7	SI ⩾ 50%	Rabeprazole, 10 mg BID for 2 weeks	Symptom score reduced by more than a third
des Varannes et al.[35]	29	40.28	NA	31.94	Erosion	Total AET > 4.2%	—	—	SAP ⩾ 95% or SI ⩾ 50%	Rabeprazole, 20 mg BID for 1 week	The cut-off descriptor of at least a ‘clear improvement’
Aanen et al.[36]	67	62.16	51	NA	—	Total AET > 4.2%	—	—	SAP ⩾ 95% or SI ⩾ 50%	Esomeprazole, 40 mg daily for 2 weeks	Adequate symptom suppression
Dickman et al.[37]	35	65.71	55.6	34.29	Erosion	Total AET > 4.2%, or upright AET > 6.0%, or supine AET > 1.2%				Rabeprazole, 20 mg BID for 1 week	Symptom score improved by more than 50%
Juul-Hansen et al.[38]	52	33.85	47	0	—	Total AET > 4.0%	—	—	—	Lansoprazole, 60 mg for 1 week	Symptom relief: yes/no
Pandak et al.[39]	38	42.86	NA	26.19	Erosion	Total AET > 4.2%	—	—	—	Omeprazole, 40 mg/d orally BID for 2 weeks	Symptom score improved by more than two points
Juul-Hansen et al.[40]	56	31.25	54	0	—	Total AET > 4.2%	—	—	—	Lansoprazole 60 mg daily for 1 week	Antacid use decreased by more than 75%
Fass et al.[41]	35	94.29	55	100	Erosion	Total AET > 4.2%	—	—	—	Omeprazole (40 mg a.m. plus 20 mg p.m.) for 1 week	Heartburn score improved by more than 50%
Fass et al.[42]	42	76.19	55.2	50	Erosion	Total AET > 4.2%	—	—	—	Omeprazole (40 mg a.m. plus 20 mg p.m.) for 1 week	Heartburn score improved by more than 50%
Bate et al.[43]	58	55.07	47.4	49.28	—	Total AET > 4.0%	—	—	—	Omeprazole, 40 mg o.m. for 2 weeks	Symptoms improved by at least two grades (or from mild to none)
Johnsson et al.[44]	160	NA	NA	57.5	Erosion	Total AET > 4.0%	—	—	—	Omeprazole, 20 mg BID for 1 week	Symptoms decreased by at least one grade
Fass et al.[45]	37	97.44	60.1	43.24	Erosion	Total AET > 4.2%	—	—	—	Omeprazole (40 mg a.m. plus 20 mg p.m.) for 1 week	More than 50% reduction in symptoms
Carlsson et al.[46]	225	50.1	50	61	Erosion	—	—	—	—	Omeprazole, 20 mg for 4 weeks	Complete symptom relief
Galmiche, Barthelemy and Hamelin[47]	141	44.7	50	26	Erosion	—	—	—	—	Omeprazole, 20 mg for 4 weeks	Symptoms decrease to 1 day/week, less than mild
Hatlebakk et al.[48]	161	57	49	48	Erosion	—	—	—	—	Omeprazole, 20 mg for 4 weeks	Symptoms decrease to 1 day/week, less than mild
Venables et al.[49]	330	52	52	31	Erosion	—	—	—	—	Omeprazole, 20 mg for 4 weeks	Symptoms decrease to 1 day/week, less than mild
A versus G
Ates et al.[50]	268	NA	NA	22.76	Erosion	Total AET > 5.3%	—	—	—	Mucosal impedance	A cut-off value of 2019 Ω
Saritas Yuksel et al.[51]	69	NA	NA	27.54	Erosion	Total AET > 5.3%				Mucosal impedance	A cut-off value of 3200 Ω
F versus E
Xu et al.[52]	2014	126	46.83	45.6	NA	8-week PPI treatment				GERDQ	A cut-off score of 8

A: esophageal pH/pH impedance monitoring and/or endoscopy; B: baseline impedance; C: salivary pepsin; D: DIS; E: GERDQ; F: PPI test; G: mucosal impedance.

AET, acid exposure time; BID, twice daily; DIS, dilated intercellular space; GERD, gastroesophageal reflux disease; GERDQ, GERD questionnaire; MNBI, mean nocturnal baseline impedance; NA, not applicable; o.m., every morning; SAP, symptom association probability; SI, symptom index; PPI, proton-pump inhibitor.

Upper endoscopy or esophageal pH/pH impedance monitoring

Esophageal mucosal breaks on upper endoscopy suggest the presence of GERD. GERD was diagnosed in studies when patients had esophagitis of any grade in one of the commonly used classification systems, such as the Los Angeles or Hetzel–Dent grading systems.

Ambulatory esophageal pH/pH impedance monitoring is generally considered to provide the most objective evidence for pathologic reflux. The results were considered abnormal based on criteria defined in the individual studies. Whenever possible, we chose definitions that would reasonably be interpreted as abnormal in clinical practice, including: (a) acid exposure time (AET) ⩾ 3.2%–5.5% of the monitoring time; (b) DeMeester score ⩾ 14; (c) positive symptom reflux association, such as symptom-associated probability (SAP) ⩾ 95% or symptom index (SI) ⩾ 50%.

Baseline impedance

The baseline impedance was assessed on the pH/impedance system, and the cut-off value ranging from 2100 Ω to 2292 Ω was established and used in individual studies.

Salivary pepsin

Salivary pepsin was detected and quantitatively/semiquantitatively measured by non-invasive rapid salivary pepsin lateral flow device (LFD) (Peptest, RDBiomed, Hull, UK). The cut-off value was used based on criteria defined in individual studies.

Dilated intercellular space

The quantitative or semiquantitative measurement of DIS under light microscopy was performed, and the cut-off value was used based on criteria defined in individual studies.

GERDQ

GERDQ was considered positive if the score was >8.

Proton-pump inhibitor test

The definition of ‘a positive PPI test’ was based on criteria defined in individual studies. Whenever possible, we chose definitions that would reasonably be interpreted as representing success in clinical practice, and ‘complete relief of heartburn’ is the most commonly adopted criteria.

Mucosal impedance

The cut-off value of mucosal impedance was also established and used in the individual studies.

Statistical analysis

Firstly, traditional pairwise meta-analyses were performed for studies to compare different diagnostic modalities using the Stata version 12.0 software (StataCorp, College Station, TX, USA). The pooled estimates of odd ratios (ORs) and 95% confidence intervals (CIs) for sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR−), and diagnostic odds ratio of GERD were calculated if there were at least four studies included that had no threshold effect. Area under receiver operating characteristic (AUROC) was also calculated. When the AUROC is closer to 1, the clinical value is greater. When the AUROC is between 0.5 and 0.7, the clinical value is lower. An AUROC value > 0.7 indicates that the clinical value is good. Heterogeneity among studies was tested using the I[2] and Chi-square tests.[53] Secondly, the evidence network structure was drawn using the R version 3.5.1 statistical computing software and network package. Each node represents different diagnostic tests, with the node size reflecting the number of patients, and the thickness of lines between nodes indicating the number of included studies. Thirdly, Bayesian network meta-analyses were performed to combine the effective sizes of direct and indirect comparisons. Lack of autocorrelation and convergence were checked and confirmed by four chains and a 20,000-simulation burn-in phase; finally, direct probability statements were derived from an additional 50,000-simulation phase.[54] The consistency between direct and indirect evidence was assessed with the node-splitting method, and the consistency or inconsistency model was selected accordingly.[55] The ranking probability was then used to calculate the probability of each diagnostic test being the most effective diagnostic method based using a Bayesian approach, and the bar charts of the ranking probability were also produced; the larger the value is, the better the rank of the diagnostic test.[56,57] Comparison-adjusted funnel plots were performed to detect the small study effects on data.[56,58] R (version 3.5.1) package GeMTC was used for this network meta-analysis.

Results

Study search flow

A total of 6223 potentially relevant studies were initially retrieved and identified, of which 705 duplicate studies were excluded. Of the 5518 citations, 5450 citations were ruled out during the first screen by abstract review. A total of 68 studies were then evaluated for eligibility by full-text review. After full-text review, studies unrelated to diagnostic tests for GERD (n = 23), studies not in English/Chinese language (n = 2), and studies not eligible for enrollment (n = 3) were excluded. Altogether, a total of 40 published studies meeting the predetermined inclusion criteria were identified [13-52] (Figure 1).

Figure 1.

The PRISMA study search flow.

PRISMA, Preferred Reporting Items for Systematic Review and Meta-Analysis.

GERD, gastroesophageal reflux disease.

Study characteristics and qualities

Of 40 included studies, the evaluation of esophageal impedance for GERD diagnosis was performed in 4 studies, salivary pepsin by Peptest in 2 studies, DIS in 5 studies, GERDQ in 6 studies, and PPI test in 23 studies when compared with esophageal pH/pH impedance monitoring or endoscopy. One study compared the diagnostic accuracy of GERDQ and PPI test. The clinical information of the included studies is shown in Table 1. The diagnostic characteristic of diagnostic tests for GERD varied across studies (Table 2). The evaluation of the risk of bias and applicability concerns using the QUADAS-2 was shown in Figure 2 and Supplementary Figure 1.

Table 2.

Diagnostic evaluation of seven diagnostic tests for gastroesophageal reflux disease.

Comparison (author)	Patients					Prevalence of GERD according to reference test	SEN	SPE	PPV	NPV	LR+	LR−
	TP	FP	FN	TN	Total
A versus B
Frazzoni et al.[13]	60	3	20	38	121	0.66	0.75	0.93	0.95	0.66	10.25	0.27
Ravi et al.[14]	25	5	4	21	55	0.53	0.86	0.81	0.83	0.84	4.48	0.17
Frazzoni et al.[15]	62	31	6	190	289	0.24	0.91	0.86	0.67	0.97	6.50	0.10
Kandulski et al.[16]	15	5	4	12	36	0.53	0.79	0.71	0.75	0.75	2.68	0.30
A versus C
Hayat et al.[17]	66	40	18	74	198	0.42	0.79	0.65	0.62	0.80	2.24	0.33
Saritas Yuksel et al.[18]	11	2	11	23	47	0.47	0.50	0.92	0.85	0.68	6.25	0.54
A versus D
Cui et al.[19]	186	123	111	145	565	0.53	0.63	0.54	0.60	0.57	1.36	0.69
Zhou et al.[20]	188	118	119	151	576	0.53	0.61	0.56	0.61	0.56	1.40	0.69
Cui et al.[21]	111	0	8	42	161	0.74	0.93	1.00	1.00	0.84	NA	0.07
Mastracci et al.[22]	102	6	17	14	139	0.86	0.86	0.70	0.94	0.45	2.86	0.20
Vela et al.[23]	9	1	6	10	26	0.58	0.60	0.91	0.90	0.63	6.60	0.44
A versus E
Zhou et al.[20]	203	145	149	139	636	0.55	0.58	0.49	0.58	0.48	1.13	0.86
Zavala-Gonzales et al.[24]	129	20	51	52	252	0.71	0.72	0.72	0.87	0.50	2.58	0.39
Jonasson et al.[25]	115	11	32	11	169	0.87	0.78	0.50	0.91	0.26	1.56	0.44
Lacy et al.[26]	77	41	31	28	177	0.61	0.71	0.41	0.65	0.47	1.20	0.71
Jones et al.[27]	125	33	69	81	308	0.63	0.64	0.71	0.79	0.54	2.23	0.50
Bai et al.[28]	620	1405	815	5225	8065	0.18	0.43	0.79	0.31	0.87	2.04	0.72
A versus F
Zhou et al.[20]	248	158	104	126	636	0.55	0.70	0.44	0.61	0.55	1.27	0.67
Jonasson et al.[25]	90	4	27	3	124	0.94	0.77	0.43	0.96	0.10	1.35	0.54
Lee et al.[29]	55	82	15	36	188	0.37	0.79	0.31	0.40	0.71	1.13	0.70
Dent et al.[30]	106	35	91	64	296	0.67	0.54	0.65	0.75	0.41	1.52	0.71
Cho et al.[31]	49	4	15	5	73	0.88	0.77	0.56	0.92	0.25	1.72	0.42
Zheng et al.[32]	5	11	4	7	27	0.33	0.56	0.39	0.31	0.64	0.91	1.14
Lee et al.[33]	52	28	17	67	164	0.42	0.75	0.71	0.65	0.80	2.56	0.35
Fan et al.[44]	20	2	5	5	32	0.78	0.80	0.71	0.91	0.50	2.80	0.28
Des Varannes et al.[35]	15	6	3	5	29	0.62	0.83	0.45	0.71	0.63	1.53	0.37
Aanen et al.[36]	41	17	5	4	67	0.69	0.89	0.19	0.71	0.44	1.10	0.57
Dickman et al.[37]	12	2	4	17	35	0.46	0.75	0.89	0.86	0.81	7.13	0.28
Juul-Hansen et al.[38]	34	17	0	1	52	0.65	1.00	0.06	0.67	1.00	1.06	0.00
Pandak et al.[39]	19	7	1	11	38	0.53	0.95	0.61	0.73	0.92	2.44	0.08
Juul-Hansen et al.[40]	29	11	5	11	56	0.61	0.85	0.50	0.73	0.69	1.71	0.29
Fass et al.[41]	21	8	0	6	35	0.60	1.00	0.43	0.72	1.00	1.75	0.00
Fass et al.[42]	28	3	7	4	42	0.83	0.80	0.57	0.90	0.36	1.87	0.35
Bate et al.[43]	22	11	10	15	58	0.55	0.69	0.58	0.67	0.60	1.63	0.54
Johnsson et al.[44]	100	16	35	9	160	0.84	0.74	0.36	0.86	0.20	1.16	0.72
Fass et al.[45]	18	2	5	12	37	0.62	0.78	0.86	0.90	0.71	5.48	0.25
Carlsson et al.[46]	66	25	72	62	225	0.61	0.48	0.71	0.73	0.46	1.66	0.73
Galmiche et al.[47]	27	65	10	39	141	0.26	0.73	0.38	0.29	0.80	1.17	0.72
Hatlebakk et al.[48]	55	59	22	25	161	0.48	0.71	0.30	0.48	0.53	1.02	0.96
Venables et al.[49]	80	120	21	109	330	0.31	0.79	0.48	0.40	0.84	1.51	0.44
A versus G
Ates et al.[50]	108	6	34	120	268	0.53	0.76	0.95	0.95	0.78	15.97	0.25
Saritas Yuksel et al.[51]	37	9	5	18	69	0.61	0.88	0.67	0.80	0.78	2.64	0.18
F versus E
Xu et al.[52]	68	2	34	22	126	0.81	0.67	0.92	0.97	0.39	8.00	0.36

A: esophageal pH/pH impedance monitoring and/or endoscopy; B: baseline impedance; C: salivary pepsin; D: DIS; E: GERDQ; F: PPI test; G: mucosal impedance.

AET, acid exposure time; DIS, dilated intercellular space; FN, false negative; FP, false positive; GERD, gastroesophageal reflux disease; GERDQ, GERD questionnaire; LR, likelihood ratio; MNBI, mean nocturnal baseline impedance; NPV, negative-predictive value; PPI, proton-pump inhibitor; PPV, positive-predictive value; SAP, symptom association probability; SEN, sensitivity; SI, symptom index; SPE, specificity; TP, true positive; TN, true negative.

Figure 2.

The evaluation of risks of bias of included studies.

Pairwise meta-analysis for diagnostic tests for GERD

A direct pairwise meta-analysis of the diagnostic performance of six different tests for GERD diagnosis was conducted. The results revealed that the baseline impedance, GERDQ and PPI test exhibited lower sensitivity and specificity when compared with esophageal pH/pH impedance monitoring or endoscopy. We also calculated the AUROC for each diagnostic test and found that the esophageal impedance and PPI test were higher than 0.70, indicating that they had relatively high diagnostic value (Supplementary Table 1). The pairwise meta-analysis of DIS could not be performed successfully due to a threshold effect. The pairwise meta-analysis of salivary pepsin and mucosal impedance could not be performed either because there were only two studies included.

Evidence network of diagnostic tests for GERD

The evidence network structure included seven diagnostic tests. The highest number of evaluable patients performed the esophageal pH/pH impedance monitoring or endoscopy, and most studies compared PPI test with esophageal pH/pH impedance monitoring or endoscopy for GERD diagnosis (Figure 3).The effect of the direct comparison of different tests with esophageal pH/pH impedance monitoring or endoscopy had similar effect on the entire network meta-analysis (Supplementary Figure 2).

Figure 3.

The evidence network structure.

A: esophageal pH/pH impedance monitoring or endoscopy; B: baseline impedance; C: salivary pepsin; D: DIS; E: GERDQ; F: PPI test; G: mucosal impedance.

DIS, dilated intercellular spaces; GERD, gastroesophageal reflux disease; GERDQ, GERD questionnaire; PPI, proton-pump inhibitor.

Main results of network meta-analysis of diagnostic tests for GERD

The NMA found no significant difference among the baseline impedance, mucosal impedance, and esophageal pH/pH impedance monitoring or endoscopy in terms of both sensitivity and specificity. It was also demonstrated that the salivary pepsin detected by Peptest had comparable specificity with esophageal pH/pH impedance monitoring or endoscopy (Figure 4).

Figure 4.

The forest plots based on sensitivity, specificity, positive-predictive value and negative-predictive value of different diagnostic tests for GERD.

A: esophageal pH/pH impedance monitoring or endoscopy; B: baseline impedance; C: salivary pepsin; D: DIS; E: GERDQ; F: PPI test; G: mucosal impedance.

DIS, dilated intercellular spaces; GERD, gastroesophageal reflux disease; GERDQ, GERD questionnaire; PPI, proton-pump inhibitor.

Ranking probability of diagnostic tests for GERD

Ranking probability indicated that esophageal pH/pH impedance monitoring and/or endoscopy had the highest sensitivity, followed by baseline impedance or mucosal impedance, PPI test, salivary pepsin or DIS, and GERDQ [Figure 5(a)]. Moreover, esophageal pH/pH impedance monitoring and/or endoscopy also had the highest specificity, followed by the mucosal impedance, baseline impedance, DIS or salivary pepsin, GERDQ, and PPI test [Figure 5(b)]. The ranking probability of positive-predictive value and negative-predictive value was also provided in Figure 5(c) and 5(d).

Figure 5.

The ranking probability based on sensitivity, specificity, positive-predictive value and negative-predictive value of different diagnostic tests for GERD.

A: esophageal pH/pH impedance monitoring and/or endoscopy; B: baseline impedance; C: salivary pepsin; D: DIS; E: GERDQ; F: PPI test; G: mucosal impedance.

DIS, dilated intercellular spaces; GERD, gastroesophageal reflux disease; GERDQ, GERD questionnaire; PPI, proton-pump inhibitor.

Assessment of publication bias

The results of assessment of publication bias demonstrated symmetrical distribution, indicating no small sample effect or publication bias in this NMA (Supplementary Figure 3).

Discussion

We present the first systematic review and NMA comparing the diagnostic performance of GERDQ questionnaire, PPI test, baseline impedance, mucosal impedance, DIS, salivary pepsin and esophageal pH/pH impedance monitoring/endoscopy for GERD. The NMA and ranking probabilities reveal that the mucosal impedance and baseline impedance had comparable sensitivity and specificity with esophageal pH/pH impedance monitoring or endoscopy. GERDQ had the lowest sensitivity, and PPI test had the lowest specificity.

The results of direct pairwise comparison and NMA shows that esophageal reflux monitoring or endoscopy is superior to other diagnostic tests for GERD, which is in accordance with the recommendation in current guidelines.[59] In the current meta-analysis, the criteria for abnormal reflux varied across the included studies which may add clinical heterogeneity to our analysis to some extent. However, this heterogeneity couldn’t be avoided, since the understanding toward GERD pathophysiology and the diagnostic criteria for GERD had developed and changed during the past several decades. For instance, the latest Lyon Consensus lists conclusive evidence, including Los Angeles grade C and D erosive esophagitis from upper endoscopy or AET > 6% from pH/pH impedance monitoring for the definitive diagnosis of GERD.[60] However, most previous studies diagnosed GERD based on lower AET thresholds and esophageal mucosal breaks, regardless of grades. Further studies are needed to compare different grades of esophagitis and different criteria for reflux for the diagnosis and management of GERD.

We also found that the esophageal mucosal impedance as well as the baseline impedance had comparable diagnostic performance with esophageal reflux monitoring and endoscopy. Baseline impedance reflects the integrity of the esophageal mucosa,[61] with low values observed in patients having GERD, and associated with increased acid reflux, as well as DIS.[62,63] It is reported that mucosal impedance can help differentiate GERD from eosinophilic esophagitis (EoE), achalasia, and healthy controls with higher specificity (95%) when compared with reflux testing (64%).[50] Given its high diagnostic performance, the esophageal mucosal impedance, as well as the baseline impedance, may become promising diagnostic tools for GERD in the future. However, normative values for them still need to be determined. Also, mucosa impedance is currently not commercialized and widely used despite high diagnostic utility.

Additionally, the salivary pepsin detection was initially proposed as a non-invasive method for the diagnosis of GERD. Our results demonstrated that the measurement of salivary pepsin detected by the Peptest device had comparable specificity to esophageal pH/pH impedance monitoring or endoscopy. Nevertheless, Sifrim and coworkers failed to reproduce good specificity of salivary pepsin to diagnose GERD, and they found that salivary pepsin could not differentiate GERD from functional heartburn.[64] Therefore, salivary pepsin detection cannot be recommended for clinical application at present. Besides, our results found that the measurement of DIS only had only modest diagnostic characteristics. It has been reported that esophageal mucosal changes such as DIS may help differentiate GERD from other disorders.[65,66] Even so, the measurement of DIS on electron microscopy is not ready for clinical practice yet.

There are limitations to these findings. First, esophageal pH/pH impedance monitoring and endoscopy were examined together in the current NMA, which were found to have highest sensitivity and specificity. However, upper endoscopy alone has, actually, low sensitivity for GERD, so we should interpret results with caution. Second, we only analyzed GERDQ, while there are several other questionnaires for GERD diagnosis. However, all of them have different scoring systems and putting them together may lead to high heterogeneity in the NMA. And the GERDQ is a questionnaire derived from validated questionnaires including the Reflux Disease Questionnaire, Gastrointestinal Symptom Rating Scale, and the GERD Impact Scale. It is the most commonly used questionnaire for GERD. Thus, we had to choose only one questionnaire to be examined. Also, we did not include the postreflux swallow-induced peristaltic wave (PSPW) index in the current NMA because studies evaluating the diagnostic value of the PSPW index are limited at present; future NMA studies can include it, since it is also a promising metric for GERD diagnosis. Third, the rankings and probabilities can sometimes be misleading, ignoring the basic principles of certainty evaluation in evidence during NMA. We should analyze the results of NMA with caution and take the results of pairwise comparison into account. Furthermore, there exists high heterogeneity in the study populations, including various baseline characteristics, different procedure of the testing, and different outcome measure, etc., which may make the results incomparable, to an extent. Finally, the results of pairwise comparison and NMA were associated with wide confidence intervals and some included studies were of moderate-to-low quality.

In conclusion, the current systematic review and NMA shows that esophageal mucosal impedance and baseline impedance has a high diagnostic performance similar to esophageal reflux monitoring or endoscopy. The future direction of GERD management should be developing and improving techniques with high diagnostic performance; not only for a precision phenotype definition but also for a tailored treatment strategy.

Click here for additional data file.

Supplemental material, Supplementary_Fig._1._risk_of_bias_for_each_study for Assessing different diagnostic tests for gastroesophageal reflux disease: a systematic review and network meta-analysis by Mengyu Zhang, John E. Pandolfino, Xuyu Zhou, Niandi Tan, Yuwen Li, Minhu Chen and Yinglian Xiao in Therapeutic Advances in Gastroenterology

Click here for additional data file.

Supplemental material, Supplementary_Fig._2._the_effect_of_evidence for Assessing different diagnostic tests for gastroesophageal reflux disease: a systematic review and network meta-analysis by Mengyu Zhang, John E. Pandolfino, Xuyu Zhou, Niandi Tan, Yuwen Li, Minhu Chen and Yinglian Xiao in Therapeutic Advances in Gastroenterology

Click here for additional data file.

Supplemental material, Supplementary_Fig._3._Publication_Bias for Assessing different diagnostic tests for gastroesophageal reflux disease: a systematic review and network meta-analysis by Mengyu Zhang, John E. Pandolfino, Xuyu Zhou, Niandi Tan, Yuwen Li, Minhu Chen and Yinglian Xiao in Therapeutic Advances in Gastroenterology

Click here for additional data file.

Supplemental material, Supplementary_Table1._pairwise_meta_analysis-supplementary for Assessing different diagnostic tests for gastroesophageal reflux disease: a systematic review and network meta-analysis by Mengyu Zhang, John E. Pandolfino, Xuyu Zhou, Niandi Tan, Yuwen Li, Minhu Chen and Yinglian Xiao in Therapeutic Advances in Gastroenterology