A Perspective on Standardizing the Predictive Power of Noninvasive Cardiovascular Tests by Likelihood Ratio Computation: 2. Clinical Applications

doi:10.4065/74.11.1072

Mayo Clinic Proceedings

Volume 74, Issue 11, November 1999, Pages 1072-1087

https://doi.org/10.4065/74.11.1072 Get rights and content

Likelihood ratio measures may be used as a standard for expressing the predictive power of noninvasive cardiovascular tests, calculated from sensitivity and specificity measures or as ratios of the predictive value odds to pretest odds for positive and negative test results. The positive likelihood ratio, (+)LR, expresses the power of a positive test result to augment an estimate of disease probability independent of the pretest prevalence of disease in a given population; the negative likelihood ratio, (-)LR, expresses the power of a negative test result to augment an estimate of the probability of no disease independent of the pretest prevalence of no disease in the same population. The likelihood ratio principle is applicable to the evaluation of the predictive power of single or combined test results reported for either dichotomous or continuous end points. This part of the perspective exemplifies application of the likelihood ratio principle in a wide variety of testing con- ditions for coronary artery disease followed by a discussion of the limitations of likelihood ratio computation in test power evaluation. Likelihood ratios provide a more concise and unambiguous standard for calibrating the pre dictive power of single and combined noninvasive cardiovascular test results than are provided by measures of sensitivity, specificity, and predictive value.

Section snippets

Estimating The Predictive Power Of Noninvasive Test Results For Diagnosis Of Cad

A comparison of the relative predictive power of noninvasive tests is most ideally made when the tests to be compared are performed in parallel in the same index population. The index population must be representative of the population to which the tests are to be applied and should be of sufficient size to permit meaningful statistical analysis of the observed differences in test power. Such optimum circumstances for the comparison of noninvasive cardiovascular tests are rarely achieved. The

Estimating Predictive Independence For Combined Test Results

The calculation of the predicted and observed effective likelihood ratio for disease for multiple positive test results, (+)ELR_d, and the predicted and observed effective likelihood ratio for no disease for multiple negative test results, (-)ELR_n, provides a unique method for verifying the presence or absence of predictive independence among combined test results. This method is based on the premise that predictive independence among test results of like sign can be verified by the finding that

Nterpreting Multiple Threshold Tests Using The Roc Curveand The Likelihood Ratio

A common approach to assessing and comparing test results that can be interpreted at multiple thresholds (ie, tests reported in continuous quantitative terms or at multiple qualitative end points) is the ROC curve. The ROC curve is usually exhibited as a graphic plot of Se vs (l - Sp). In the context of the ROC curve, the Se is termed the true-positive rate plotted on the vertical axis, while the (1 - Sp) is termed the false-positive rate plotted on the horizontal axis. This ROC form is in fact

Estimating The Power Of A Multivariable Predictive Model

The potential error introduced by the use of pretest assumptions and test redundancy when applying Bayesian predictive formulations for 2 or more variables has led to increasing interest in the use of alternative multivariable probability models. In using such methods as multiple linear and logistic regression, discriminant function, and proportional hazards analysis, the predictive variables are adjusted or corrected by introduction of regression coefficients for each cofactor in the

Evaluating The Influence Of Verification Bias On The Diagnostic Power Of Noninvasive Cardiovascular Tests

Verification bias is a form of population selection bias that occurs when the sensitivity and specificity of a noninvasive test are derived from a population referred for definitive diagnostic evaluation on the basis of prior noninvasive testing. Verification bias includes 2 forms of selection bias: (1) knowledge ledge of previous test results that biases referral for verification in favor of those with positive test results and (2) pretest patient characteristics that influence a physician's

Evaluating The Predictive Power Of Cardiovascular Risk Factors

The likelihood ratio principle is readily adapted to the analysis of the predictive power of cardiovascular risk factors. Such risk factor analysis includes 2 forms: (1) epidemiological analysis, wherein the likelihood ratios are computed for population subsets with a given risk factor, irrespective of an association with other risk factors and (2) clinical analysis, wherein the likelihood ratios are computed for population subsets with specific risk factor profiles. Epidemiological analysis

Evaluating The Diagnostic Power Of Chest Pain Syndromes

Likelihood ratio analysis is uniquely applicable to an evaluation of the predictive power of clinical symptoms. Such an analysis is exemplified in a study of the prevalence of CAD in chest pain syndrome subsets reported by Chaitman et al⁵³ for patients enrolled in the Coronary Arteriography Surgery Study (CASS). In this study, 5334 men and 2806 women older than 30 years were referred for coronary arteriography; they were subdivided by 3 symptom complexes into populations with definite angina,

Limitations In Deriving And Applying Likelihood Ratio Estimates Of Test Power

Variables that limit the estimation and application of the likelihood ratio as a measure of the predictive power of a noninvasive test include the following: (1) index population size, (2) index population composition, (3) methodologic variation in testing, and (4) variance between testing and verification modalities. The limitation of index population size is especially pertinent to the likelihood ratio computation, while the remaining limitations apply to all measures of test power.

Acknowledgments

The author expresses his appreciation to Julienne M. Montgomery for her able assistance in preparing the manuscript.

References (54)

AM Weissler
A perspective on standardizing the predictive power of noninvasive cardiovascular tests by likelihood ratio computation, 1: mathematical principles
Mayo Clin Proc
(1999)
HJ Berger et al.
Global and regional left ventricular response to bicycle exercise in coronary artery disease: assessment by quantitative radionuclide angiocardiography
Am J Med
(1979)
JA Jengo et al.
Detection of coronary artery disease: comparison of exercise stress radionuclide angiocardiography and thallium stress perfusion scanning
Am J Cardiol
(1980)
DE Johnstone et al.
Comparison of exercise radionuclide angiocardiography and thallium-201 myocardial perfusion imaging in coronary artery disease
Am J Cardiol
(1980)
NL DePace et al.
Noninvasive assessment of coronary artery disease
Am J Cardiol
(1983)
WF Armstrong et al.
Effect of prior myocardial infarction and extent and location of coronary disease on accuracy of exercise echocardiography
J Am Coll Cardiol
(1987)
SG Sawada et al.
Exercise echocardiographic detection of coronary artery disease in women
J Am Coll Cardiol
(1989)
MM Pozzoli et al.
Exercise echocardiography and technetium-99m MIBI single-photon emission computed tomography in the detection of coronary artery disease
Am J Cardiol
(1991)
LJ Crouse et al.
Exercise echocardiography as a screening test for coronary artery disease and correlation with coronary arteriography
Am J Cardiol
(1991)
TH Marwick et al.
Accuracy and limitations of exercise echocardiography in a routine clinical setting
J Am Coll Cardiol
(1992)

AS Iskandrian et al.

Value of exercise thallium-201 imaging in patients with diagnostic and nondiagnostic exercise electrocardiograms

Am J Cardiol

(1981)

AP Morise et al.

Development and validation of a logistic regression-derived algorithm for estimating the incremental probability of coronary artery disease before and after exercise testing

J Am Coll Cardiol

(1992)

Cited by (21)

Optimized electrocardiographic criteria for prior inferior and anterior myocardial infarction
2012, Journal of Electrocardiology
Citation Excerpt :
The positive likelihood ratio expresses a diagnostic test's ability to detect the disease being tested for. A positive likelihood ratio greater than 10 exhibited by a diagnostic test is considered to indicate an excellent rule-in power for the test.10 Unlike the positive predictive value, positive likelihood ratios permit diagnostic inferences that are independent of the prevalence of the abnormality in the population being tested.10
The first purpose of the study was to optimize empirically the detection of prior inferior myocardial infarction (IMI) and prior anterior myocardial infarction (AMI) by electrocardiogram (ECG). The second purpose was to compare the diagnostic performances of the new criteria with those of 3 widely used commercial diagnostic ECG algorithms.
We analyzed the digital ECG data from 1138 subjects with suspected coronary artery disease in whom the presence or absence of prior IMI or AMI was documented by coronary angiography and left ventriculography. We used receiver operating characteristic curves to develop the new criteria for prior IMI and AMI using a training set of 562 subjects and then tested their diagnostic performances using a separate test set of 576 subjects. In both the training and test sets, we used χ² test to compare the performances of the new criteria with those of 3 commercial computerized diagnostic algorithms.
The best criterion for prior IMI was the algebraic sum of the Q and T amplitudes in leads III and aVF. Its sensitivities/specificities were 71%/98% and 74%/98% in the training and test sets, respectively. The best criterion for prior AMI was the algebraic sum of the Q, R, and T amplitudes minus the Q duration in leads V₂, V₃, and V₄. Its sensitivities/specificities were 68%/98% and 65%/98% in the training and test sets, respectively. In both the training and test sets, these diagnostic performances were generally superior to those of the 3 commercial algorithms.
Using digital ECG data, we developed and tested new criteria for prior IMI and AMI whose diagnostic performances are generally superior to each of 3 widely used commercial ECG diagnostic algorithms.
Noninvasive Detection of Left Ventricular Systolic Dysfunction by Acoustic Cardiography in Cardiac Failure Patients
2008, Journal of Cardiac Failure
Citation Excerpt :
We also generated receiver operator characteristic curves to determine the diagnostic sensitivities and specificities for LVSD and used them to calculate positive and negative likelihood ratios. Unlike positive and negative predictive values, positive and negative likelihood ratios are independent of the prevalence of the abnormality in the population being tested.21 To avoid dividing by zero, we set the positive likelihood ratio equal to sensitivity in the cases in which specificity was 100%.
Despite its shortcomings, ejection fraction (EF) is widely used to detect left ventricular systolic dysfunction (LVSD) as has prolonged QRS duration as indirect evidence of LVSD. However, acoustic cardiography provides other parameters for detecting LVSD without these limitations. One parameter, the electromechanical activation time (EMAT), is prolonged in LVSD. We compared the abilities of acoustic cardiography, EF, and QRS duration to detect LVSD.
We studied a sample of 108 patients who underwent elective diagnostic cardiac catheterization. The diagnostic findings included left ventricular filling pressures, angiographic EF, and maximum left ventricular dP/dt. We defined LVSD as a maximum left ventricular dP/dt of <1600 mm Hg/s. At thresholds of 35% and 50% to detect LVSD, the sensitivities of angiographic EF were 39% and 74%; specificities were 89% and 70%, respectively. At QRS duration thresholds of 100 and 120 ms, sensitivities were 63% and 35%; specificities, 63% and 89%, respectively. At thresholds of 100 and 110 ms, EMAT had sensitivities of 53% and 42%, and specificities of 90% and 100%, respectively. EMAT performed better than LV EF or QRS duration in hemodynamic subgroups.
Acoustic cardiography is a convenient, automated diagnostic method whose performance for detecting LVSD exceeds both angiographic EF and QRS duration alone.
Usefulness of Acoustic Cardiography to Resolve Ambiguous Values of B-Type Natriuretic Peptide Levels in Patients With Suspected Heart Failure
2007, American Journal of Cardiology
Citation Excerpt :
A positive or a negative likelihood ratio >10 shown by a diagnostic test was considered to have excellent rule-in or rule-out power for the condition being tested, respectively. Unlike positive and negative predictive values, diagnostic inferences based on positive and negative likelihood ratios were independent of the prevalence of the abnormality in the population being tested.12 Statistical analyses were performed using SPSS, version 13.0 (SPSS, Inc., Chicago, Illinois).
B-type natriuretic peptide (BNP) levels are helpful to diagnose left ventricular (LV) systolic and/or diastolic dysfunction. BNP levels that are only moderately increased have limited diagnostic ability, and an additional test to resolve this problem would be desirable. The hypothesis that acquiring combined electrocardiographic and electronic cardiac acoustical data can improve the detection of LV dysfunction in patients with nondiagnostic values of BNP was tested. Both BNP and combined 12-lead electrocardiograms with electronic heart sound (acoustic cardiographic) recordings were obtained from 164 outpatients referred for echocardiographic evaluation for suspected heart failure. Acoustic cardiographic parameters included the third heart sound (S₃) and percentage of electromechanical activation time, measured as the interval from onset of the Q wave of the electrocardiogram to the first heart sound (S₁) and expressed as a proportion of the cardiac cycle. Sixty-nine of 164 patients (42%) had BNP values in the “gray zone” of 100 to 500 pg/ml. Sensitivity and specificity for LV dysfunction of BNP in the gray zone were 55% and 75%, with a positive likelihood ratio of 2.3. The use of acoustic cardiographic parameters in these 69 patients increased sensitivity and specificity to 69% and 100%, with a corresponding positive likelihood ratio of 69. In conclusion, easily obtainable acoustic cardiographic data substantially improved the diagnostic evaluation of patients with nondiagnostic BNP values and therefore can increase the confidence with which physicians diagnose and treat LV dysfunction.
Assessment of patients with low-risk chest pain in the emergency department: Head-to-head comparison of exercise stress echocardiography and exercise myocardial SPECT
2005, American Heart Journal
Citation Excerpt :
To estimate the predictive power of single tests for predicting the presence or absence of disease, sensitivity, specificity, positive (+) and negative (−) predictive values, accuracy (defined as percentage of true test results, ie, (true positives + true negatives)/total number tests performed), and likelihood ratios [LR] ((+)LR = sensitivity/(1−specificity); (−)LR = specificity/(1−sensitivity)) were calculated considering follow-up data. Standard deviations, 95% confidence limits,30 and odds LR (posttest odds of disease for a positive test = (+)PTOD; posttest odds of no disease for a negative test = (−)PTON)31,32 were also calculated. The same LRs were adopted for estimating the power of combinations of tests, applied in sequence, for predicting the presence or absence of disease (see Appendix A).
The aim of the study was to compare head-to-head the performance of exercise tolerance test–stress echocardiography (ex-Echo) and exercise stress-perfusion nuclear imaging (exercise–single-photon emission computed tomography [ex-SPECT]) for the diagnosis of coronary artery disease (CAD) in patients evaluated at the chest pain unit with delay from chest pain (CP) onset.
As an early triage strategy for CAD in emergency medicine, ex-Echo could have the advantage of widespread availability and low costs.
In the years 2000-2002, 503 consecutive patients (mean age 60 years) with recent (<24 hours) CP and nonischemic electrocardiogram (ECG), in whom CAD remained undiagnosed after first line 6-hour work-up including serum markers of myocardial injury and resting echocardiogram, underwent ex-Echo and ex-SPECT within 24 hours. Patients with (+)ex-Echo or (+)ex-SPECT or (+)ex-ECG or abnormal troponin I were referred to coronary angiography; otherwise, they were discharged and followed up. End points were coronary stenosis ≥50% and cardiovascular events at 6-month follow-up.
Ninety-nine patients (20%) had (+)ex-Echo and 121 (24%) (+)ex-SPECT; CAD was diagnosed in 81% and 67%, respectively; positive tests were concordant in 69%. In negative ex-Echo and ex-SPECT, final evidence of CAD emerged in 14 and 13, respectively. Ex-Echo demonstrated higher accuracy than ex-SPECT (93% ± 1% vs 89% ± 1%), optimal specificity (95% ± 5% vs 90% ± 5%), and positive predictive value (81% ± 4% vs 67% ± 4%); moreover, in the case of (−)ex-ECG, observed effective likelihood ratio indicates a (+)synergy between ex-ECG and ex-Echo.
Ex-Echo can be an effective diagnostic strategy in the early triage of CP patients, improving diagnosis in case of (−)ex-ECG and reducing unnecessary angiography number. Its drawback is represented by the 5% of missed diagnosis.
Use of Likelihood Ratio Computation to Standardize the Predictive Power of Noninvasive Cardiovascular Tests
2000, Mayo Clinic Proceedings
A perspective on standardizing the predictive power of noninvasive cardiovascular tests by likelihood ratio computation: 2. Clinical applications
1999, Mayo Clinic Proceedings
Likelihood ratio measures may be used as a standard for expressing the predictive power of noninvasive cardiovascular tests, calculated from sensitivity and specificity measures or as ratios of the predictive value odds to pretest odds for positive and negative test results. The positive likelihood ratio, (+)LR, expresses the power of a positive test result to augment an estimate of disease probability independent of the pretest prevalence of disease in a given population; the negative likelihood ratio, (-)LR, expresses the power of a negative test result to augment an estimate of the probability of no disease independent of the pretest prevalence of no disease in the same population. The likelihood ratio principle is applicable to the evaluation of the predictive power of single or combined test results reported for either dichotomous or continuous end points. This part of the perspective exemplifies application of the likelihood ratio principle in a wide variety of testing con- ditions for coronary artery disease followed by a discussion of the limitations of likelihood ratio computation in test power evaluation. Likelihood ratios provide a more concise and unambiguous standard for calibrating the pre dictive power of single and combined noninvasive cardiovascular test results than are provided by measures of sensitivity, specificity, and predictive value.

View all citing articles on Scopus

View full text

Original ArticleA Perspective on Standardizing the Predictive Power of Noninvasive Cardiovascular Tests by Likelihood Ratio Computation: 2. Clinical Applications

Section snippets

Estimating The Predictive Power Of Noninvasive Test Results For Diagnosis Of Cad

Estimating Predictive Independence For Combined Test Results

Nterpreting Multiple Threshold Tests Using The Roc Curveand The Likelihood Ratio

Estimating The Power Of A Multivariable Predictive Model

Evaluating The Influence Of Verification Bias On The Diagnostic Power Of Noninvasive Cardiovascular Tests

Evaluating The Predictive Power Of Cardiovascular Risk Factors

Evaluating The Diagnostic Power Of Chest Pain Syndromes

Limitations In Deriving And Applying Likelihood Ratio Estimates Of Test Power

Acknowledgments

Mayo Clin Proc

Am J Med

Am J Cardiol

Am J Cardiol

Am J Cardiol

J Am Coll Cardiol

J Am Coll Cardiol

Am J Cardiol

Am J Cardiol

J Am Coll Cardiol

J Am Coll Cardiol

J Am Soc Echocardiogr

Mayo Clin Proc

Am J Cardiol

Am J Cardiol

J Am Coll Cardiol

J Am Coll Cardiol

J Am Coll Cardiol

Mayo Clin Proc

Am J Cardiol

J Am Coll Cardiol

J Am Coll Cardiol

J Am Coll Coll Cardiol

J Am Coll Cardiol

Am J Cardiol

Am J Cardiol

J Am Coll Cardiol

Original Article
A Perspective on Standardizing the Predictive Power of Noninvasive Cardiovascular Tests by Likelihood Ratio Computation: 2. Clinical Applications