1349281 - Misclassification of Females and Males in Quantitative Parametric Mapping – The Importance of Sex-Specific Normal Ranges for Diagnosis of Health versus Disease

Parametric mapping enables quantitative assessment of myocardial tissue abnormalities. However, the normal range of a mapping sequence is method-specific, and is affected by numerous parameters including magnetic field strength, type of MR scanner, and other technical parameters¹. Women also have higher normal myocardial T1 values than men^2,3. The SCMR mapping consensus statement recommends that each center establishes a local reference range using a minimum of 15 healthy volunteers¹. However, creating local reference ranges using insufficient numbers of each gender may lead to decreased diagnostic accuracy. We investigated the impact of using male normal ranges for classifying female cases as normal or abnormal, and female normal ranges for classifying male cases as normal or abnormal.

Methods:

Healthy volunteers (male and female) were scanned on a Siemens 3T TRIO MR system (n=92) and a Siemens 3T PRISMA MR system (n=49) (Table 1). Healthy volunteers had no history of cardiovascular conditions and took no regular medication. Native short-axis myocardial T1-maps were acquired using Siemens Work-In-Progress (WIP) variants of the Shortened Modified Look-Locker Inversion Recovery (ShMOLLI) sequence. Image analysis was performed using cmr42 (Circle Cardiovascular Imaging, Calgary, AB). Endo- and epicardial borders of the left ventricle were drawn manually to produce an averaged global T1 value for each volunteer. The Monte Carlo method was used to run 20,000 simulations. Reference ranges were generated using any random 15 participants from each cohort (either male-only, female-only, or mixed-gender). Gender-specific T1 values were compared to generated normal ranges (simulated mean ± 2 standard deviations) to estimate the statistical accuracy of classifying healthy female or male cases correctly.

Results:

There was a significant difference in myocardial T1 values between genders (p< 0.0001 for both scanners; Table 1) consistent with previous literature^2,4. Using only males to create a local normal range leads to significantly reduced diagnostic accuracy in female cases, with 14-36% of healthy females incorrectly classified as having abnormal T1 values (Table 2). Similarly, using only females to create a local normal range leads to significantly reduced diagnostic accuracy in male cases, with 19-37% of healthy males incorrectly classified as having abnormal T1 values.  Additionally, there was a significant difference in normal T1 values between the two Siemens 3T scanners (p< 0.0001). Small but significant inter-version drifts in normal T1 values have been described previously⁵.

Conclusion:

We demonstrate that using 15 healthy volunteers that are not sex-specific for establishing local normal ranges may misclassify up to 36% of healthy females and 37% of healthy males as having abnormal T1 values. Local reference ranges should comprise at least 15 participants of each gender to allow balanced diagnostic accuracy for both females and males.