• 
• 

Three Part Question

In [patients with left bundle branch block] do [the Sgarbossa electrocardiographic criteria] allow [accurate detection of acute myocardial infarction]?

Clinical Scenario

A seventy year-old man presents to the Emergency Department with a thirty-minute history of chest pain. ECG shows left bundle branch block (LBBB) and you are not sure whether it is pre-existent. You have heard of the scoring system derived by Sgarbossa for use in this situation but you wonder if it will enable you to accurately decide whether this man is having an acute myocardial infarction (MI), in order that thrombolytic therapy may be considered.

Search Strategy

OVID Medline 1966 - July 2005
EMBASE 1974 - July 2005 via the Dialog Datastar Interface
The Cochrane Library 2005 Issue 2
OVID:
[exp Myocardial Infarction/ OR exp Myocardial Ischemia/ OR exp Coronary Thrombosis/ OR (myocard$ ADJ infarct$).mp. OR heart attack.mp. OR AMI.mp. OR acute coronary syndrome.mp.] AND [(exp Bundle-Branch Block/ AND left.af.)OR left bundle branch block.mp. OR LBBB] AND [exp Electrocardiography/ OR electrocard$.mp. OR ECG.mp. OR EKG.mp.] limit to human and English language
EMBASE:
[ACUTE-HEART-INFARCTION#.DE. OR HEART-INFARCTION#.DE. OR ST-SEGMENT-ELEVATION#.DE. OR (myocard$ ADJ infarct$).mp.] AND [HEART-LEFT-BUNDLE-BRANCH-BLOCK#.DE. OR LBBB.mp.] AND [ELECTROCARDIOGRAM#.W..DE. OR ECG.mp. OR EKG.mp.] limit to human and English language
Cochrane:
left bundle branch block

Search Outcome

620 papers were identified with the reported search using OVID Medline, 239 using EMBASE and 90 papers using Cochrane. Eight papers were relevant to the three-part question, all of which were found in both OVID and EMBASE.

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Sgarbossa et al 1996 USA	131 patients with acute MI (documented by serum enzyme changes) and LBBB on their baseline ECG who were enrolled in the GUSTO-1 trial. The control group consisted of 131 patients randomly selected from the Duke Databank for Cardiovascular Disease, who had complete LBBB and stable, angiographically documented coronary artery disease. These patients did not have acute chest pain at the time of ECG. ECG's interpreted by one of four investigators. A random sample of 10% of ECG's was interpreted by all four investigators and kappa values reported. Only criteria for which the interobserver agreement was at least 0.7 were considered for inclusion in the clinical prediction rule. Once a model for the prediction rule had been derived, it was tested for validation in an independent population of 45 patients with LBBB, acute chest pain and a high likelihood of coronary artery disease (taken from the GUSTO-1 registry and GUSTO-2A study). 22 of these had a diagnosis myocardial infarction and 23 had unstable angina.	Diagnostic cohort	ST elevation >=1mm, concordant with QRS complex, for diagnosis of MI (multivariate analysis)	Odds ratio 25.2 (95% CI 11.6-54.7)	Control group did not have chest pain. The trial may have been more appropriately designed if it had included patients with undifferentiated chest pain and derived a clinical decision rule according to whether they tested positive or negative for MI. Significantly more patients in the control group had a previous MI (59% vs. 26%, p<0.001). Validation group was small (45 patients).
			ST depression >=1mm in lead V1, V2 or V3 (multivariate analysis)	Odds ratio 6.0 (1.9-19.3)
			ST elevation >=5mm and discordant with QRS complex	Odds ratio 4.3 (1.8-10.6)
			Predictive value of the clinical decision rule (Sgarbossa criteria, see Comments) on prospective validation (score >=3)	Sensitivity 30%, specificity 96%, LR+ 9.0, LR- 0.7, PPV 88%, NPV 61%. For every 100 patients with LBBB treated according to the Sgarbossa criteria, two would be thrombolysed despite not having an MI. 31 who were having MI's would not receive thrombolysis. 18 would be appropriately thrombolysed.
Gunnarsson et al 2001 Sweden	158 patients with left bundle branch block, chest pain <6 hours and clinical suspicion of MI. ECG's reviewed by the three authors. Where there were disagreements, the majority verdict was taken. If all three authors disagreed, the patient was excluded from analysis. Kappa 0.57-0.70. Diagnosis of MI was made according to CK-MB or CK-B in combination with total CK.	Prospective diagnostic cohort	Total score >=3 using Sgarbossa criteria for diagnosis of MI	Sensitivity 17.1% (95% CI 8.6-25.6%), specificity 94% (88.9%-99.1%), LR+ 2.85, LR- 0.88, post-test probability 72%. For every 100 patients with LBBB treated according to these criteria, three would be inappropriately thrombolysed, 40 with MI would not be thrombolysed and 23 with MI would be thrombolysed.	Outdated diagnosis of myocardial infarction. No mention of how many patients were excluded because of failure to reach an agreement about the ECG.
			Score >=3 on Sgarbossa criteria and survival at 30 days and 1 year	No significant difference in survival (p=0.97)
Gula et al 2003 Canada	414 ECG's with LBBB from a cohort of patients with AMI (who had been recruited another trial) compared with 85 controls with LBBB who had been attending hospital out-patient clinics. ECG's were interpreted by the three authors, with disagreements settled by consensus. Diagnosis of MI was made by the primary or "most responsible" diagnosis on the hospital record.	Retrospective diagnostic cohort	Interobserver agreement (kappa values)	LBBB 0.608; >5mm discordant ST elevation 0.368; ST depression V1, V2 or V3 0.523; Concordant ST elevation 0.393.	Lack of objective criteria for diagnosis of MI. Small control group (85 patients). Clinical utility of the scoring system proposed by Sgarbossa not assessed.
			Concordant ST elevation for diagnosis of MI	Sensitivity 6.3%, specificity 98.8%, LR+ 5.3, LR- 1.0
			ST depression V1-V3 for diagnosis of MI	Sensitivity 3.1%, specificity 100%
			Discordant ST elevation (measured 80ms after J point) for diagnosis of MI	Sensitivity 37.5%, specificity 62.4%, LR+ 1.0, LR- 1.0
			Any abnormality from the Sgarbossa criteria for diagnosis of MI	Sensitivity 44.5%, specificity 61.2%, LR+1.2, LR- 0.9
Kontos et al 2001 United States	182 patients presenting to the Emergency Department with suspected myocardial ischaemia and LBBB on initial ECG. All ECGs were reviewed by two blinded cardiologists, with disagreements resolved by a third cardiologist. MI diagnosed according to CK-MB levels, although 11 patients were discharged prior to CK-MB estimation, following myocardial perfusion imaging.	Prospective diagnostic cohort	Concordant ST elevation for diagnosis of MI	Sensitivity 8% (95% CI 2-26%), specificity 100% (98-100%), PPV 100%, NPV 88%. Note only two patients had this ECG feature (both had MI).	Not all patients had the 'gold standard' test for diagnosis of MI (CK-MB). The 'gold standard' test for diagnosis of MI is now outdated. Relatively small numbers had the outcome in question (24 had MI); very small numbers had each Sgarbossa criteria.
			ST depression V1-V3 for diagnosis of MI	Sensitivity 17% (7-36%), specificity 100% (98-100%), PPV 100%, NPV 89%. Note only 4 patients had this ECG feature (all had MI).
			Discordant ST elevation >=5mm for diagnosis of MI	Sensitivity 21% (9-41%), specificity 93% (88-96%), PPV 31%, NPV 89%. Note 16 patients had this ECG feature (5 had MI).
			Any Sgarbossa criteria for diagnosis of MI	Sensitivity 46% (28-65%), specificity 93% (88-96%), PPV 50%, NPV 92%. (22 had any criteria, 11 of whom had MI).
			Initial CK-MB or any ECG crtieria for diagnosis of MI	Sensitivity 63% (42-79%), specificity 99% (95-100%), PPV 88%, NPV 94%
Li et al 2000 United States	190 patients identified on the ECG database of an urban teaching hospital as having a diagnosis of LBBB and who were intially admitted with suspected cardiac ischaemia. Medical records were reviewed. All patients had cardiac enzyme estimation (CK-MB). Each ECG was reviewed by two physicians and evaluated for the Sgarbossa criteria.	Retrospective diagnostic cohort	Interobserver variability	91% or higher for each of the three criteria	Retrospective. Relatively small numbers - only 25 patients had the outcome in question. Now outdated criteria for diagnosis of MI (CK-MB), which may have underestimated the incidence. Clinical utility of the scoring system proposed by Sgarbossa not assessed.
			Concordant ST elevation for diagnosis of MI	Sensitivity 16%, specificity 100%, LR+ 16, LR- 0.8. Given a pre-test probability of 13% (study prevalence), a positive finding shifts the post-test probability of MI to 71%.
			Disocrdant ST elevation >=5mm for diagnosis of MI	Sensitivity 8%, specificity 95%, LR+ 1 (95% CI 0.3-4), LR- 1 (95% CI 0.8-1). Hence the post-test probability remains 13% following a positive or negative test.
			ST depression V1-V3 for diagnosis of MI	Sensitivity 0% (95% CI 0-14%), specificity 98%, LR+ 0 (0-8), LR- 1 (0.8-1)
			Any of the Sgarbossa criteria for diagnosis of MI	Sensitivity 20%, specificity 93%, LR+ 3, LR- 0.9. Post-test probability following a positive test 57%.
Edhouse et al 1999 United Kingdom	48 patients who presented to a teaching hospital with chest pain suspicious of myocardial ischaemia within 12 hours who had LBBB. ECG's were reviewed blindly, apparently by four independent observers	Retrospective diagnostic cohort	No Sgarbossa criteria on initial ECG and a diagnosis of MI	27% had MI	Retrospective. No report of how disagreements in ECG interpretation were resolved. Relatively small numbers.
			No Sgarbossa criteria on serial ECG's and a diagnosis of MI	17.2% had MI
			Any Sgarbossa criteria for diagnosis of MI	Sensitivity 79% (95% CI 63-95%), specificity 100%, PPV 100%, NPV 79%
			Interobserver agreement	Kappa scores 0.65, 0.69, 0.74, 0.77, 0.81 and 0.87 between observers, indicating good agreement.
Sokolove et al 2000 United States	ECG's from 224 patients with LBBB were independently interpreted by four cardiologists and four emergency physicians using the Sgarbossa clinical prediction rule. 100 ECG's were obtained from patients enrolled to the GUSTO-1 trial, who had MI confirmed by enzyme changes. 124 controls were selected from patients in the Duke Databank for Cardiovascular Disease who had stable, angiographically documented coronary artery disease and LBBB.	Retrospective cohort	Interobserver agreement	Excellent agreement both within groups (kappas 0.81, 0.84) and between cardiologists and emergency physicians (kappa 0.81)	Although the study does help to answer the three-part question, it was primarily designed to investigate interobserver agreement.
			Diagnosis of MI by cardiologists	Median sensitivity 73%, median specificity 98%
			Diagnosis of MI by emergency physicians	Median sensitivity 67%, median specificity 99%
Shlipak et al 1999 United States	103 presentations in 83 patients presenting with either chest pain, pulmonary oedema or post-cardiac arrest who had complete LBBB and cardiac enzyme estimation. ECG's evaluated as positive or negative according to the Sgarbossa clinical decision rule. All ECG's evaluated by the most senior electrocardiographer at the institution (intraobserver variability kappa 0.80). MI diagnosed according to characteristic presentation and either troponin or CK-MB elevation).	Prospective diagnostic cohort	Concordant ST elevation for diagnosis of MI	Sensitivity 7% (95% CI 1-21%), specificity 100% (95-100%), PPV 100%, NPV 71%	Not all patients had definitive troponin testing for the diagnosis of MI (some diagnosed using the now outdated CK-MB). Probabilities of survival were derived using data from other studies. Thus the statistics reported are inference not evidence and make considerable assumptions.
			ST depression V1-V3 for diagnosis of MI	Sensitivity 3% (0-17%), specificity 100% (95-100%), PPV 100%, NPV 71%
			Discordant ST elevation >=5mm for diagnosis of MI	Sensitivity 19% (7-37%), specificity 82% (71-90%), PPV 32%, NPV 70%, LR+ 1.1, LR- 0.99
			Overall Sgarbossa algorithm for diagnosis of MI	Sensitivity 10% (2-26%), specificity 82% (71-90%), PPV 100%, NPV 72%, LR+ undefined, LR- 0.9. Taking the mid-range of the 95% CI for specificity (98%), LR+ can be estimated at 5.0. Given pre-test probability of 30%, this gives a post-test probability of 68%. For every 100 patients treated using these criteria, no patients would be inappropriately thrombolysed, three would be appropriately thrombolysed and 27 would be ineligible for thrombolysis despite having an MI.
			Probability of stroke-free survival (derived by utilising data from other studies)	Thrombolysis for all: 92.6%; Thrombolysis for none 91.6%; Apply ECG algorithm to decide 91.6%

Comment(s)

The diagnosis of MI in the presence of LBBB is notoriously difficult. 50-70% of patients with LBBB and suspected MI are not actually having an infarct (Hands et al, 1988; Shlipak et al, 1999), but those who are have a high mortality rate and derive significant benefit from thrombolysis (Sgarbossa et al, 1998; Fibrinolytic Therapy Trialists, 1994). Edhouse et al found that 48% of patients with LBBB who were thrombolysed were not having MI, while 23% of those with MI were not thrombolysed. Sgarbossa et al proposed the use of a scoring system for the ECG diagnosis of acute MI with LBBB, as follows: * ST elevation >=1mm concordant with the QRS complex = 5 points * ST depression >=1mm in lead V1, V2 or V3 = 3 points * ST elevation >=5mm discordant with the QRS complex = 2 points. A score of >=3 points is consistent with a diagnosis of MI. The clinical utility of these criteria has been validated by several studies, all of which have demonstrated a high specificity but some have shown a low sensitivity. As such, although the algorithm cannot be used to rule out MI it can effectively help to rule it in, but should be utilised in conjunction with clinical features. In borderline cases, urgent measurement of cardiac enzymes such as CK-MB or serial ECG's may help to clarify the diagnosis. Where primary angioplasty is available, this patient group is likely to derive significant benefit, irrespective of ECG features.

Clinical Bottom Line

The Sgarbossa algorithm may be useful to help rule in myocardial infarction but cannot be used to rule it out. The algorithm should be used cautiously and in conjunction with other clinical information.

Level of Evidence

Level 2 - Studies considered were neither 1 or 3.

References

1. Sgarbossa EB; Pinski SL; Barbagelata A; Underwood DA; Gates KB; Topol EJ; Califf RM; Wagner GS, for the GUSTO-1 investigators Electrocardiographic diagnosis of evolving acute myocardial infarction in the presence of left bundle branch block New England Journal Of Medicine 1996; 334: 481-7
2. Gunnarsson G; Eriksson P; Dellborg M ECG criteria in diagnosis of acute myocardial infarction in the presence of left bundle branch block International Journal of Cardiology 2001; 78: 167-174
3. Gula LJ; Dick A; Massel D Diagnosing acute myocardial infarction in the setting of left bundle branch block: prevalence and observer variability from a large community setting Coronary Artery Disease 2003; 14: 387-393
4. Kontos MC; McQueen RH; Jesse RL; Tatum JL; Ornato JP Can myocardial infarction be rapidly identified in Emergency Department patients who have left bundle branch block? Annals of Emergency Medicine 2001; 37: 431-438
5. Li SF; Walden PL; Macrcilla O; Gallagher EJ Electrocardiographic diagnosis of myocardial infarction in patients with left bundle branch block Annals of Emergency Medicine 2000; 36(6): 561-565
6. Edhouse JA; Sakr M; Angus J; Morris FP Suspected myocardial infarction and left bundle branch block: electrocardiographic indicators of acute ischaemia Journal of Accident and Emergency Medicine 1999; 16: 331-335
7. Sokolove PE; Sgarbossa EB; Amsterdam EA; Gelber R; Lee TC; Maynard C; Richards JR; Valente R; Wagner GS Interobserver variability in the electrocardiographic diagnosis of acute myocardial infarction in patients with left bundle branch block Annals of Emergency Medicine 2000; 36(6): 566-571
8. Shlipak MG; Lyons WL; Go AS; Chou TM; Evans GT; Browner WS Should the electrocardiogram be used to guide therapy for patients with left bundle-branch block and suspected myocardial infarction? JAMA 1999; 281(8): 714-719
9. Sgarbossa EB; Pinski SL; Topol EJ; Califf RM; Barbagelata A; Goodman SG; Gates KB; Granger CB; Miller DP; Underwood DA; Wagner GS Acute myocardial infarction and complete bundle branch block at hospital admission: clinical characteristics and outcome in the thrombolytic era Journal of the American College of Cardiology 1998; 31(1): 105-110
10. Fibrinolytic Therapy Trialists' (FTT) Collaborative Group Indications for fibrinolytic therapy in suspected myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients Lancet 1994; 343: 311-322
11. Hands ME; Cook EF; Stone PH; Muller JE; Hartwell T; Sobel BE; Roberts R; Braunwald E; Rutherford J; and the MILIS Study Group Electrocardiographic diagnosis of myocardial infarction in the presence of complete left bundle branch block American Heart Journal 1988; 116: 23-32