Analysis group, study weight (optional), study quality score (optional)
Data sorting options
dr-ROC handles up to 25 studies per analysis. For larger analyses, use dr-ROC 100, which is available free of charge to all registered dr-ROC users.
Output
Manuscript-ready table with study-by-study sensitivity and specificity and meta-analysis results
Manuscript-ready graphs:
Summary ROC
Data points and confidence intervals
Logit plot (D vs. S)
DOR forest plot (study by study effect sizes)
Three-part sensitivity/specificity forest plot (as called for by Cochrane Collaboration guidelines)
Comparison of summary ROC, fixed-effects, and random-effects results
Post-test probability
Two-page results summary including:
Total of patients and studies, total included in meta-analysis
Summary of data analysis options
Sensitivity and specificity at mean threshold, with confidence interval
Effect size (D)
Area under the ROC
Area under the upper left quadrant of the ROC
Fixed-effects meta-analysis (Mantel-Haenszel)
Random-effects meta-analysis (DerSimonian-Laird)
Logit regression results with standard error and p-value
Sensitivity, specificity, PPV, NPV with upper and lower confidence limits, at mean, minimum, and maximum thresholds
Pooled sensitivity, specificity, PPV, NPV, prevalence
Results at point where sensitivity = specificity (S = 0)
Corresponding specificity at user-selected sensitivity, sensitivity at user-selected specificity
Specificity at mean sensitivity
Mean, standard deviation, minimum, maximum, and median values for study-by-study sensitivity, specificity, PPV, NPV, prevalence, S (threshold), and D (effect size)
Spearman rank correlation for sensitivity vs. specificity, with p-value
Pearson product-moment correlation for sensitivity vs. specificity, with p-value
r2 correlation coefficient for sensitivity vs. specificity
Heterogeneity tests: Cochran's Q, p(Q), i2
Mean, standard deviation, etc. for study quality scores of all studies and included studies
Analysis options
Study inclusion and exclusion: include or exclude each individual study from meta-analysis --no cutting, pasting, or retyping!
Weighting: no weighting, N, inverse variance, quality score, user-specified weight
Continuity correction: all studies or just studies with zero TP, FN, FP, or TN
Constrain ROC to constant D (symmetric curve) or unconstrained regression
Calculation of overall prevalence: pooled across studies or average of studies
Confidence interval: user specified
Show or hide data check alerts
Display options
Sort table and forest plot: Study author, study ID, N, effect size (D), quality score, prevalence, or as entered
Label studies: Study author, study ID, or both
Include or exclude excluded studies from per-study summaries
Label graphs with analysis parameters: on or off
Include or exclude excluded studies from output table
ROC and data graph options
Display summary ROC
Display fixed-effects or random-effects results instead of summary ROC
Display summary ROC confidence interval
Display extended range of summary ROC and confidence interval
Display individual data points
Display excluded data points using contrasting marker
Display confidence interval of individual data points
Display mean threshold point
Display confidence interval of mean threshold point
Display minimum and maximum threshold points
Display point at user-selected sensitivity and specificity
Show gridlines
Show diagonal line
Label graph with percentages or decimals
Label graph as sensitivity/specificity or TPR/1-FPR
Additional options with other graphs
Manual
32-page manual includes documentation of all analysis and display options
Hard copy and PDF editions provided with each order
Quick start page also included
dr-ROC Requirements
Microsoft Excel dr-ROC runs under Microsoft Excel. It has been developed and tested with the following versions of Excel.
Windows: Excel 2000*, Excel 2002, Excel 2003
Macintosh: Excel v.X, Excel 2004
*--Excel 2000 cannot display the extended range results for the summary ROC and confidence interval lines. All other graph functions, including standard range summary ROC and confidence intervals are now compatible with Excel 2000.
Manuscript-ready tables require any current version of Microsoft Word
DOR meta-analysis: dr-ROC now performs random-effects (DerSimonian-Laird) and fixed-effects (Mantel-Haenszel) meta-analysis of the logs of study diagnostic odds ratios (D). Note that these are one-dimensional analyses which do not take into account the dependence of DOR on effective threshold (S).
Q*: The point on the ROC where sensitivity equals specificity is labeled Q*, which should not be confused with the Q-test for heterogeneity.
Sensitivity at mean specificity: A few published articles report objectively choosing a point on the ROC by taking the mean specificity across studies and finding the corresponding sensitivity. dr-ROC now calculates this point.
Pooled sensitivity and specificity: Sensitivity and specificity are meta-analyzed separately. This method is simply to pool all the data, calculate the particular ratio, and calculate its confidence interval in the same manner that individual study confidence intervals are calculated. Heterogeneity of these figures will be introduced in Version 2.1.
Correlations: The Spearman rank correlation and Pearson product-moment correlations for sensitivity versus specificity are calculated, along with their statistical significance. The coefficient of determination, r2, measures the proportion of variation in specificity that would be accounted for by differences in sensitivity.
Graphs
In the summary ROC plot, you may choose between summary ROC meta-analysis results, random-effects, or fixed-effects. Confidence intervals, mean theshold, and minimum/ maximum thresholds correspond to the chosen meta-analysis results. When necessary, an alert will appear in the graphs to prompt you to run graphing macros after changing analysis method.
In the Forest plot, you may also choose between summary ROC, random effects, and fixed-effects for the summary log DOR plotted at the bottom of the graph.
New "Different Methods" plots summary ROC, random-effects, and fixed effects results on a single graph. When both random-effects and fixed-effects lines are plotted, the random-effects line will be red and the fixed-effects line will be blue.
New "Sensitivity/Specificity" graph provides side-by-side Forest plots of sensitivity and specificity, including your choice of the pooled results (see above) or the summary ROC results.
Behind the scenes
Some reorganization of the internal worksheets to facilitate this and future upgrades, and to make sheets a little less complicated.
dr-ROC is copyright (C) 2005 by dr2 Consulting; all rights reserved.
Specifications current as of April 2006. Subject to change without notice.
Windows and Excel are trademarks of Microsoft Corporation.
Macintosh is a trademark of Apple Computer
