Generate summary ROC curves for diagnostic test accuracy meta-analysis. Enter sensitivity and specificity pairs or 2x2 tables from multiple studies. The tool fits the Moses-Littenberg model, plots individual study points as weighted bubbles, computes the AUC and Q* index, and generates R code for the mada package.
Load sample data to see how the tool works, or clear all fields to start fresh.
Drag & drop a file or
CSV, TSV, Excel (.xlsx/.xls) - max 500 rows
| Study | Sensitivity | Specificity | N | |
|---|---|---|---|---|
Select between sensitivity/specificity pairs (most common) or full 2x2 tables with true positives, false positives, false negatives, and true negatives. Both formats are accepted, and the tool computes all derived measures automatically.
Input the diagnostic accuracy data for each study. Include sample size (N) for bubble weighting. Sensitivity and specificity can be entered as proportions (0 to 1) or percentages (0 to 100). Import from CSV/Excel or type directly into the table.
The D3.js ROC space plots each study as a bubble (sized by sample size) at its sensitivity and 1-specificity coordinates. The fitted SROC curve shows the summary relationship, and the Q* point marks where sensitivity equals specificity.
Review the AUC (area under the SROC curve), Q* index, regression intercept, and slope. The intercept represents the log diagnostic odds ratio when there is no threshold effect, and the slope indicates the magnitude of the threshold effect.
Copy the auto-generated methods paragraph for your manuscript. Export reproducible R code using the mada package, which fits the bivariate model (Reitsma et al., 2005) for a more statistically rigorous analysis.
Download the SROC plot as a high-resolution PNG or PDF suitable for journal submission. Export study-level data including sensitivity, specificity, sample size, and log diagnostic odds ratios as CSV or Excel.
Need this done professionally? Get a complete systematic review or meta-analysis handled end-to-end.
Get a Free QuoteThe area under the SROC curve (AUC) provides a single number summarizing the overall accuracy of the diagnostic test across all possible thresholds. An AUC of 0.9 or above indicates excellent accuracy, 0.8 to 0.9 indicates good accuracy, and 0.7 to 0.8 indicates moderate accuracy. Unlike the AUC from a single-study ROC curve, the SROC AUC accounts for between-study variation in thresholds.
The Q* index is the point on the SROC curve where sensitivity equals specificity. It provides an easily interpretable summary measure that is not influenced by threshold differences across studies. The Q* index ranges from 0.5 (no discrimination) to 1.0 (perfect discrimination) and corresponds to the diagnostic odds ratio at zero threshold effect.
Each study is plotted as a bubble in ROC space, with the bubble area proportional to the sample size. Larger bubbles represent studies with more participants and therefore greater precision. Visually inspecting the relationship between bubble size and position can reveal whether larger studies cluster differently from smaller studies, which may indicate publication bias or study quality effects.
In the Moses-Littenberg model, the slope (b) of the regression of log diagnostic odds ratio on the threshold variable indicates the extent of threshold effect. A slope near zero suggests that the diagnostic odds ratio is relatively constant across thresholds, while a larger absolute slope indicates that diagnostic accuracy varies substantially with the threshold used.
The Moses-Littenberg approach implemented here is the traditional method. For peer-reviewed publication, the bivariate model (Reitsma et al., 2005) or the hierarchical SROC model (Rutter and Gatsonis, 2001) are generally preferred because they account for the correlation between sensitivity and specificity within studies and provide proper confidence and prediction regions.
Best practice in diagnostic test accuracy meta-analysis is to present both the SROC curve and paired forest plots showing study-level sensitivity and specificity. The SROC curve provides the summary visual, while forest plots reveal heterogeneity patterns. The Cochrane Handbook for DTA Reviews recommends this combined presentation.
The summary receiver operating characteristic curve is the primary graphical tool for presenting the results of diagnostic test accuracy meta-analysis. First proposed by Moses, Shapiro, and Littenberg (1993), the SROC curve addresses a fundamental challenge in pooling diagnostic accuracy data: different studies often use different thresholds (cut-off values) for defining a positive test result, creating an inherent trade-off between sensitivity and specificity that cannot be captured by simply averaging these measures separately.
The Moses-Littenberg model works by transforming each study's sensitivity and specificity into two derived variables: D (the log diagnostic odds ratio) and S (a proxy for the threshold). The regression of D on S produces the SROC curve parameters. The intercept represents the log diagnostic odds ratio when there is no threshold effect, and the slope captures how the diagnostic odds ratio changes with threshold variation. When the slope is zero, all variation in sensitivity and specificity across studies is attributable to threshold differences, and the diagnostic odds ratio is constant.
The AUC (area under the SROC curve) provides a global summary of diagnostic accuracy. A perfect test would have an AUC of 1.0, with the SROC curve passing through the upper-left corner of ROC space. A test with no discriminative ability would have an AUC of 0.5, with the curve lying along the diagonal. The Q* index, the point where sensitivity equals specificity on the SROC curve, provides a complementary summary measure that is easy to interpret clinically.
Modern approaches to diagnostic test accuracy meta-analysis increasingly use the bivariate model (Reitsma et al., 2005) or the hierarchical SROC model (Rutter and Gatsonis, 2001). These methods jointly model logit-transformed sensitivity and specificity using a bivariate normal distribution, properly accounting for their negative correlation within studies and between-study heterogeneity. The R code generated by this tool includes the mada package's reitsma() function, which implements the bivariate model.
When reporting SROC curves in publications, present the curve alongside individual study estimates plotted as bubbles in ROC space, with bubble size proportional to study sample size. This allows readers to visually assess the heterogeneity pattern, identify potential outliers, and judge whether the summary curve adequately represents the data. The Cochrane Handbook for Diagnostic Test Accuracy Reviews (Macaskill et al., 2023) recommends presenting both the SROC curve and paired forest plots of sensitivity and specificity.
For a complete diagnostic test accuracy meta-analysis workflow, start by computing study-level accuracy measures with our diagnostic test accuracy calculator. Assess risk of bias with our QUADAS-2 risk of bias tool. Translate the pooled accuracy into clinical probabilities using our Fagan nomogram calculator. Present individual study results with our forest plot generator.
A summary receiver operating characteristic (SROC) curve is a graphical method for summarizing the results of a diagnostic test accuracy meta-analysis. Unlike a standard ROC curve (which plots sensitivity against 1 minus specificity for a single study across different thresholds), an SROC curve plots the pairs of sensitivity and specificity from multiple studies, each using their own threshold, and fits a summary curve through these points. The SROC curve accounts for the trade-off between sensitivity and specificity that occurs as diagnostic thresholds vary across studies, providing a single visual summary of overall test accuracy.
A standard ROC curve is generated from a single study by varying the diagnostic threshold and plotting sensitivity against (1 minus specificity) at each threshold. An SROC curve is generated from multiple studies, each contributing a single sensitivity-specificity pair at their chosen threshold. The SROC curve fits a regression through these study-level points to summarize the overall diagnostic accuracy, accounting for the fact that different studies may use different thresholds. The SROC curve is essential in diagnostic test accuracy meta-analysis because it separates genuine accuracy variation from threshold variation across studies.
The Q* index is the point on the SROC curve where sensitivity equals specificity. It represents the intersection of the SROC curve with the diagonal line from the lower-left to the upper-right corner of the ROC space (the line where sensitivity equals specificity). The Q* index provides a single summary measure of diagnostic accuracy that is independent of threshold effects. A Q* of 1.0 represents a perfect test, while a Q* of 0.5 represents a test with no discriminative ability (equivalent to chance). The Q* index is derived directly from the intercept of the Moses-Littenberg regression model.
The Moses-Littenberg SROC curve is the traditional and simpler approach, suitable when you want a visual summary and straightforward AUC and Q* index estimates. The bivariate model (Reitsma et al., 2005) and the hierarchical SROC model (Rutter and Gatsonis, 2001) are more statistically rigorous because they jointly model sensitivity and specificity while accounting for their correlation and between-study heterogeneity. For publication in major journals, the bivariate or HSROC models are generally preferred. This tool implements the Moses-Littenberg approach, and the generated R code includes mada package code that fits the bivariate model for a more complete analysis.
The area under the SROC curve (AUC) summarizes the overall discriminative ability of the diagnostic test across all possible thresholds. An AUC of 1.0 indicates perfect discrimination, while an AUC of 0.5 indicates no discrimination (chance level). General interpretation guidelines are: 0.9 to 1.0 is excellent, 0.8 to 0.9 is good, 0.7 to 0.8 is moderate, and below 0.7 is limited. However, these thresholds are guidelines rather than strict cut-offs, and the clinical context determines what level of accuracy is acceptable for a given application.
A minimum of 3 studies is needed to fit the Moses-Littenberg regression, but at least 5 to 10 studies are recommended for a reliable SROC curve. With fewer than 5 studies, the regression estimates will be unstable, and the AUC and Q* index should be interpreted with considerable caution. The bivariate model implemented in the mada R package also requires a minimum number of studies for stable estimation, typically at least 4 to 5 studies. With very few studies, reporting the individual sensitivity-specificity pairs without attempting to fit a summary curve may be more appropriate.
Compute sensitivity, specificity, and likelihood ratios from 2x2 tables with our diagnostic test accuracy calculator. Assess risk of bias in diagnostic accuracy studies with our QUADAS-2 risk of bias tool. Visualize individual study estimates with our forest plot generator for meta-analysis. Translate summary accuracy into clinical probabilities with our Fagan nomogram calculator.
Reviewed by
Dr. Sarah Mitchell holds a PhD in Biostatistics from Johns Hopkins Bloomberg School of Public Health and has over 15 years of experience in systematic review methodology and meta-analysis. She has authored or co-authored 40+ peer-reviewed publications in journals including the Journal of Clinical Epidemiology, BMC Medical Research Methodology, and Research Synthesis Methods. A former Cochrane Review Group statistician and current editorial board member of Systematic Reviews, Dr. Mitchell has supervised 200+ evidence synthesis projects across clinical medicine, public health, and social sciences. She reviews all Research Gold tools to ensure statistical accuracy and compliance with Cochrane Handbook and PRISMA 2020 standards.
Whether you have data that needs writing up, a thesis deadline approaching, or a full study to run from scratch, we handle it. Average turnaround: 2-4 weeks.