# Utility of the Predictive Summary Index for Comparison of Diagnostic Protocols in Myasthenia Gravis With Repetitive Nerve Stimulation and Concentric Needle Jitter

## TL;DR

This study compares two diagnostic tests for myasthenia gravis using a new measure called the predictive summary index, showing that one test is more reliable than the other.

## Contribution

The paper introduces the predictive summary index as a novel quality measure for comparing diagnostic tests in myasthenia gravis.

## Key findings

- Concentric needle jitter (CNJ) had a higher predictive summary index (0.82) than repetitive nerve stimulation (RNS) (0.53).
- The number needed to predict for CNJ was 1.22, compared to 1.89 for RNS.
- Amplitude decrement in RNS was found to be more reliable than area decrement.

## Abstract

As sensitivity of a test can be increased at the expense of specificity, a quality measure taking into account both parameters is desirable. The “predictive summary index” (PSI) is calculated from the prevalence of the disease, sensitivity and specificity. We characterize diagnostic yield of repetitive nerve stimulation (RNS) and stimulated concentric needle jitter (CNJ) in the diagnosis of myasthenia gravis (MG) with PSI. From PSI, the number of tests that is required for the prediction of one correct diagnosis (“number needed to predict” [NNP] in analogy to the number needed to treat) can be calculated as 1/PSI.

Among patients consecutively referred to the electrophysiology laboratory for a workup of a disorder of the neuromuscular junction (NMJ), we identified 35 patients with MG (clinical diagnosis supported by antibody findings or treatment response) and 85 patients with diagnosis unrelated to the NMJ in whom both RNS and stimulated CNJ had been performed.

The PSI was 0.82 and 0.53 for CNJ and RNS, respectively. The NNPs were 1.22 for CNJ and 1.89 for RNS. This means that 1.49 patients have to be tested with CNJ instead of RNS to arrive at one correct diagnosis. In data from the literature, PSI illustrates that 7% as a cut‐off in RNS can be helpful and that amplitude decrement is more reliable than area decrement.

CNJ was more reliable in the diagnosis of MG than RNS. PSI provides a rational quantitative framework to discuss both the validity of different techniques and their cost in terms of time spent to arrive at a correct diagnosis.

We use retrospective data from our electrophysiology laboratory to show that to diagnose or rule out myasthenia gravis one in every 1.22 concentric needle jitter tests and one in 1.89 repetitive nerve stimulations tests leads to the correct prediction.