To develop new scales on the severity of myasthenia gravis (MG) which are suitable for clinical practice in China.

A final version of new comprehensive scale was developed by item optimization from a preliminary scale established by combination of Delphi method and previous evaluation of items from source scales. With the original quantitative data from 60 MG patients, items were selected on the basis of interobserver and test-retest reliability of items, the contributions of items on internal consistency and construct validity of the preliminary scale. Evaluation of this final scale includes internal consistency, interobserver and test-retest reliability, construct validity and correlation with currently used scales. A self-rating version was simplified from the final scale by including items that can be measured by patients, which was then evaluated in 57 generalized MG patients.

One item (palpebral fissure) was deleted from the preliminary scale due to poor interobserver and test-retest reliability. The remaining nine items constituted the final comprehensive scale which covered six commonly involved muscle groups in MG. All items in the comprehensive scale had fair to excellent interobserver reliability (K=0.43-0.80, P<0.01) and fair to excellent test-retest reliability (K=0.41-1.00, P<0.01). The comprehensive scale had acceptable internal consistency (Cronbach ɑ=0.63), excellent interobserver reliability (intraclass correlation coefficient (ICC)=0.88, P<0.01) and test-retest reliability (ICC=0.86, P<0.01). The comprehensive scale had four common factors and good construct validity (cumulative dedication rate 68.46%). There was high correlation between the comprehensive scale and Myasthenia Gravis Composite (MGC), Absolute and Relative Score of MG (ARS-MG) and Quantitative Myasthenia Gravis score (QMGs) (r=0.87-0.90, P<0.01), and moderate correlation between the comprehensive scale and Myasthenia Gravis Activities of Daily Living (MG-ADL) (r=0.68, P<0.01). A self-rating scale was developed by deleting ocular and facial items and keeping the remaining six items, which covers the bulbar, limb, neck and respiratory muscles. Internal consistency (Cronbach ɑ=0.69), interobserver reliability and test-retest reliability (ICC=0.90 for both, P<0.01) of the self-rating scale were better than those of the comprehensive scale. The self-rating scale had two common factors and good construct validity (cumulative dedication rate 64.24%). There was high correlation between the self-rating scale and the comprehensive scale, MGC and ARS-MG (r=0.74-0.86, P<0.01), and moderate correlation between the self-rating scale and QMGs and MG-ADL (r=0.55-0.68, P<0.01).

Item optimization was performed from a preliminary scale established by Delphi method. A new comprehensive scale and a self-rating scale were developed. The comprehensive scale and self-rating scale had good reliability and validity, which can be used to assess the severity of MG reliably and effectively.

Myasthenia gravis; Severity; Scale; Reliability; Validity