Diagnostic accuracy of glycogen phosphorylase BB for myocardial infarction: A systematic review and meta‐analysis

Abstract Purpose We tried to investigate the diagnostic accuracy of glycogen phosphorylase BB as a cardiac marker for myocardial infarction. Methods We searched through different electronic databases (PubMed, Google‐scholar, Embase, and Cochrane Library) to locate relevant articles. Studies, with sufficient data to reconstruct a 2 × 2 contingency table, met our inclusion criteria were included. Three reviewers independently screened the articles. Discrepancies were resolved by other reviewers. Unpublished data were requested from the authors of the study via email. Subsequently, data extraction was done using a standardized form and quality assessment of studies using the QUADAS‐2 tool. Meta‐analysis was done using a bivariate model using R software. Results Fourteen studies were selected for the final evaluation, which yielded the summary points: pooled sensitivity 87.77% (77.52%–93.72%, I2 = 86%), pooled specificity 88.45% (75.59%–94.99%, I2 = 88%), pooled DOR 49.37(14.53–167.72, I2 = 89%), and AUC of SROC was 0.923. The lambda value of the HSROC curve was 3.670. The Fagan plot showed that GPBB increases the pretest probability of myocardial infarction from 46% to 81% when positive, and it lowers the same probability to 12% when negative. Conclusion With these results, we can conclude that GPBB has modest accuracy in screening myocardial infarction, but the limitations of the study warrant further high‐quality studies to confirm its usefulness in predicting myocardial infarction (MI).

diseases. 4 Most of the AMI-related mortality occurs within the first hour of onset of symptoms. 5 Hence, early detection of myocardial infarction (MI) is crucial to reduce coronary artery disease-related morbidity and mortality. 6 Since ECG alone has a low sensitivity and specificity for AMI diagnosis, the American College of Cardiology has included an elevation of cardiac troponin levels and other supplementary examinations in its guideline for AMI diagnosis. 7,8 Biomarkers are being used widely for the early detection of myocardial injury. To date, many cardiac markers have been identified for early identification of acute myocardial infarction, 9 all varying in sensitivity and specificity. In current worldwise practice, the most commonly used biomarkers are cardiac troponins and creatine kinase (CK-MB). Among the two isoforms of troponin, troponin T and troponin I, studies show that troponin I increases early, i.e., within 4-6 h, reaches peak concentration at 12 h, and finally returns to base level in 3 to 10 days. 9,10 Troponin T on the other hand remains elevated longer, i.e., 12-48 h, and falls to baseline level in about 10 days. Since cardiac troponins (Cantini) are found to have higher specificity and sensitivity over other cardiac enzymes, the elevation of cardiac troponin T is recommended as the standard biomarker criterion for establishing the diagnosis of acute myocardial infarction. 11,12 Similarly, another biomarker in practice is CK-MB, which although present in a significant amount in heart muscle, is not heartspecific as some of it is found in skeletal muscles and other tissues. 13 CK-MB rises early in the serum at almost 4-9 h after the onset of chest pain, reaches the peak concentration in blood nearly at ~24 h, and subsequently returns to baseline level at 48-72 h. Since CK-MB is cleared from the circulation early, it has higher chances of detection of reinfarction over troponins. 14 It has a specificity of 97 percent just 10-12 h after the onset of symptoms and good sensitivity if serial follow-up is done for 24-48 h, 15 but the diagnostic sensitivity is only 50% at 3 h. 16 The search for biomarkers that could aid in the early detection of MI with high specificity and sensitivity has led to the discovery of many novel molecules. Among the many emerging novel biomarkers, glycogen phosphorylase BB (GPBB) stands out as it increases in the earliest hours after AMI has set. This hints at a prospect in the use of GPBB as an early marker of AMI before significant damage ensues. 5 In most AMI patients, GPBB has been found to be increased as early as 1-4 h after the onset of chest pain and it usually peaks before CK-MB or troponin T with subsequent return to baseline level within 1-2 days after AMI onset. 17 These early results showing the potential use of GPBB for early diagnosis of AMI were our impetus in conducting this meta-analysis to analyze its diagnostic accuracy in acute myocardial infarction.

| Protocol registration
The review protocol, with a comprehensive methodology, inclusion criteria, exclusion criteria, search strategy, and review questions, was registered in Prospero. Registration Number: CRD42021252095.

| Information source and search strategy
We have analyzed the diagnostic accuracy of glycogen phosphorylase BB for myocardial infarction according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 18 The following databases served as sources for published studies prior to March 2021 to locate relevant articles that address the review question: PubMed, PMC, Embase, Google Scholar, and Cochrane Library. Our comprehensive search strategy included the terms: "acute coronary syndrome", "myocardial ischemia", "coronary artery disease", "glycogen phosphorylase, brain form", and "glycogen phosphorylase" under the medical subject heading (MeSH) terms for MEDLINE, and relevant Emtree terms were used for EMBASE search. Boolean logic was used for conducting a database search, and Boolean search operators "AND" and "OR" were used to link search terms. Various grey literature libraries, preprint servers, and thesis repositories were searched for unpublished studies. Furthermore, a secondary search was done by screening the references of retrieved studies and previous systematic reviews. The applied search strategy can be accessed in the Appendix S1.

| Inclusion and exclusion criteria
Eligible studies included in this meta-analysis had to fulfill the fol-

| Data extraction
The references retrieved from different databases were imported into Covidence (a primary screening and data extraction tool).
Covidence identified most of the duplicates and removed them automatically. The title and abstract of the remaining papers were then screened independently by NK, SR, and AT for potentially relevant studies that needed a full appraisal. AG, NK, and SR did a full-text review of selected studies based on inclusion and exclusion criteria.
Any discrepancies that arose were resolved by mutual discussion and consultation with other authors. Missing data were obtained via contacting the respective study author through email. Data were extracted from the shortlisted articles onto a standardized form designed in excel under the heading (author name, year, country, title, study design, exclusion and inclusion criteria, population characteristics, sensitivity, specificity, AUC value, cutoff value of GPBB used, and assay design used).

| Assessment of methodological quality
All authors individually used the Quality Assessment Tool for Diagnostic Accuracy Studies (QUADAS-2) 19 to assess the methodological quality of the included studies.

| Summary measures and meta-analysis
We prepared a 2 × 2 contingency table for each of the studies with TP, TN, FP, and FN, which was then imported to R (R version 4.0.3 [2020-10-10]) for statistical analysis. The author AG did the statistical analysis. We adopted a random-effect univariate statistical model using meta-package to calculate the pooled sensitivity and specificity with metaprop function and diagnostic odds ratio with metabin function, with their corresponding 95% confidence intervals. The Clopper-Pearson method 20 was used to calculate the confidence interval, and the Cochrane Q test 21  The SROC curve was constructed using a bivariate model (madapackage; reitsma function) (sensitivity as the vertical axis, falsepositive rate as the horizontal axis), and the area under the SROC curve (AUC) was calculated. 23 The hierarchical summary receiveroperating characteristic (HSROC) was used to examine the GPBB accuracy for the diagnosis of MI. We evaluated pretest probabilities (as prevalence) versus corresponding post-test probabilities following a positive or negative GPBB result based on the summary sensitivity and specificity using a Fagan plot. This showed the relationship between the prior probability, the likelihood ratio, and the post-test probability.

| Publication bias
The Deeks' funnel plot was used to assess the funnel plot asymmetry for publication bias. 24

| Study selection
We retrieved a total of 1586 articles through our search strategies:

| Study characteristics
The total number of participants from all the included studies  Table 1.

| Methodological quality
The quality of diagnostic studies was variable as assessed using the QUADAS-2 tool. The table assessing the risk of bias and con-

| Meta-analysis
The summary sensitivity obtained after meta-analysis was 87.77% Elevated GPBB was associated with increased MI, log DOR = 3.90 (2.68-5.12, I 2 = 89%, 95% CI) (Figures 3-5). There was considerable heterogeneity for all the above-mentioned statistical measures (I 2 > 50%), so moderator analysis was conducted for those statistical measures to account for the heterogeneity.
Spearman's rho was calculated to measure the strength of association between sensitivities and the false-positive rate, which showed a negative correlation between the two variables, rho = −0.432 (95% CI, −0.794-0.156), indicating that the heterogeneity among the included studies could be from other reasons but threshold effect.

| HSROC and SROC curve analysis
The sensitivity (y-axis) and false-positive rate (x-axis) were plotted in a graph to show the relationship between the true positive rate (TPR) and false-positive rate (FPR) of the test at various thresholds.
We used this to distinguish disease cases from noncases ( Figure 6).
The AUC for SROC was 0.923. The lambda, theta, beta, sigma2alpha, and sigma2theta values of HSROC were 3.670, 0.193, 0.165, 5.378, and 0.267, respectively, where 3.670 estimates the mean of the random effects for accuracy (i.e., lambda), 0.193 estimates the mean of the random effects for threshold (i.e., theta), 0.165 estimates the shape parameter (i.e., beta), 5.378 estimates the variance of the random effects for accuracy (i.e., sigma2alpha), and 0.193 estimates the variance of the random effects for threshold (i.e., sigma2theta). The resulting curve which depicts the expected trade-off between sensitivity and specificity across thresholds shows that there is certain accuracy of GPBB for diagnosing MI.

| Moderator analysis
Considering the heterogeneity obtained on analysis, moderator analysis (subgroup analysis and metaregression) was conducted to explore the effects on heterogeneity and summary estimates. The analysis was conducted by making subgroups under each tentatively homogenous group (sample size, study year, study type, and test kit used). We can see that study design is significantly associated with F I G U R E 1 Flow chart illustrating the electronic database searches and selection of studies in the meta-analysis TA B L E 1

| Sensitivity analysis
Sensitivity analysis was done by excluding one study at a time (leave-one-out method) that showed no significant differences in the pooled sensitivity, pooled specificity, and DOR (Appendix S3).

| Publication bias
A Deeks' funnel plot was used to assess the publication bias, and it is shown in Figure 8. The figure obtained is symmetrical. The Deeks' funnel plot showed no evidence of publication bias (p = 0.1288).

| DISCUSS ION
Timely diagnosis and prompt treatment of MI patients significantly decrease MI-associated morbidity and mortality. However, failing to diagnose the condition early stands as a major hindrance in achiev-  9 GPBB is abundant in normal tissues of the brain and myocardium. 39 GPBB, bound to the sarcoplasmic reticulum glycogenolytic complex of cardiomyocytes, is the key enzyme for glycogenolysis. 40,41 The myocardial metabolic state determines the level of association between GP with sarcoplasmic reticulum glycogenolytic complex and is found to be highly sensitive to ischemia-induced glycogenolysis. 41 Under ischemia during phosphorolysis, the bound form of GPBB becomes a soluble cytosolic form. As under the myocardial ischemia, glycogenolysis is increased and the cell membrane integrity is also lost resulting in GPBB release into extracellular space via the T-tubule system. 37,42,43 As the blood-brain barrier usually  Initial clinical trials suggested a high sensitivity and good specificity of GPBB assays for the early detection of myocardial infarction compared with the other biomarkers, including first-generation cardiac troponin T. 36 First generations of the troponin assays had unsatisfactory sensitivity, especially in the early phase of myocardial infarction. 49 Interestingly, recent data suggest that elevated GPBB may add prognostic information beyond hs-cTnI and brain natriuretic peptide (BNP). An association between the increased level of GPBB in a patient presenting with symptoms suggestive of ACS and a poorer midterm outcome has also been established. 44 Furthermore, studies found GPBB to be superior to cTnI and cTnT in a study for the diagnosis of anthracycline-induced cardiotoxicity and carbon monoxide-associated cardiotoxicity, both of which pertain to ischemic pathophysiology. 50

| Limitations of the study
The literature reviewing process was elaborate, but we could only extract studies from limited databases. Studies in a language other than English were not reviewed. There is significant heterogeneity in the study, which could have been a source of bias. We failed to analyze the time gap between sample blood collection to the onset of symptoms due to insufficient data. Our study has not analyzed the adjunctive role of GPBB in combination with other currently used biomarkers for the diagnosis of MI.

| CON CLUS ION
With the results of our analysis, we conclude that GPBB had moderate accuracy for diagnosis of MI and was superior to CK-MB and myoglobin. Whether GPBB assays could potentially replace hsCTn, especially during early hours of symptoms onset remains unclear.
For more conclusive evidence and for the GPBB to be made a standard cardiac biomarker for diagnosis in clinical practice, prospective studies employing rigorous laboratory and study design with robust criteria are needed to determine the clinical usefulness of these tests. Keeping the limitations of the study in mind, results may be subject to change in meta-analyses wherein more data have been extracted from a larger number of studies.

CO N FLI C T O F I NTE R E S T
The authors have declared that no competing interests exist.

AUTH O R CO NTR I B UTI O N S
AG and SG designed the study. AG, SG, and AB were engaged in the literature search. AT, NK, and SR carried out the title and abstract screening. AG, NK, and SR were involved in full-text screening.
Similarly, data extraction and quality assessment were done by each of the authors independently. AG carried out the statistical analysis, and all authors (AG, SG, NK, SR, AT, AB, and SD) drafted the manuscript. All authors were involved in revising the manuscript critically for important intellectual content. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary Materials.