In total, 1185 suspected ALS patients were enrolled, making this study the largest single-center ALS diagnostic study to date. Results demonstrated a higher sensitivity of the Gold Coast criteria in comparison to the rEEC and Awaji criteria, when considering definite, probable, and possible ALS as a positive diagnosis, and the sensitivity remained robust across subgroups defined by the site of onset, or patients with or without EMG. The sensitivity enhancement of the Gold Coast criteria was more prominent in the limb-onset ALS patients and among those who completed EMG tests. In contrast, the specificity of the Gold Coast criteria was much lower when compared to that of the rEEC and Awaji criteria.
The main limitation of the rEEC and Awaji criteria is the complexity with multiple diagnostic categories. A sizeable proportion of patients do not progress along the diagnostic categories. A fraction of pure LMN-phenotype patients die without fulfilling the criteria [3, 19] while signs of corticospinal tract pathology can be detected in post-mortem studies [20, 21]. A study has found that the diagnostic category of rEEC has no relation with the prognosis [3] and another study proved that only “definite ALS” is a significant factor for poor prognosis [22]. Therefore, the Gold Coast criteria is expected to have higher sensitivity due to the considerable simplification of diagnosis criteria, which particularly classifies LMN dysfunction in at least 2 body regions as a positive sign.
In addition, the present study reaffirmed the higher sensitivity of Awaji in both bulbar- and limb-onset ALS when compared to the rEEC. The inclusion of possible ALS as a positive diagnosis significantly enhanced the sensitivity in both the bulbar- and limb-onset subgroups, with more gains in the bulbar-onset ALS, and the results became comparable between rEEC and Awaji. Bulbar-onset patients are particularly sensitive to Awaji, as bulbar muscles do not show spontaneous activities in the early phase of ALS, but fasciculation potentials and motor unit instability can be easily detected in these muscles [6]. Bulbar involvement is very strong evidence for the diagnosis of ALS in clinical practice, which is supported by the current finding that none of the non-ALS diseases has a bulbar onset. It is obvious that when a patient with bulbar symptoms and progressing limb weakness and wasting that all point exclusively to ALS still cannot meet the criteria of rEEC, a diagnosis of ALS can be made by the Gold Coast criteria. In the present study, the Gold Coast criteria were more sensitive than rEEC and Awaji in both bulbar- and limb-onset patients, particularly in the limb-onset ALS patients.
As far as we know, this is the first diagnostic study of ALS that analyzed patients without EMG, which was an exclusion criterion in previous studies. Of the 309 patients without EMG, 306 were diagnosed as ALS (19.3% definite, 40.5% probable) and 214 reached the end point. The main reasons that ALS patients did not complete EMG in our center included having already received EMG in other hospitals, fulfilling definite or probable categories by clinical data, and intolerance to the tests. According to the new proposal, LMN involvement can be identified by clinical examination and/or EMG, which implies that EMG is not mandatorily required. However, this is certainly not advocated in clinical practice. On the one hand, with clinical examination alone it may be difficult to confirm weakness and wasting in early disease stages, and EMG can detect LMN involvement in non-weak muscles; on the other hand, EMG could provide evidence of progressive neurogenic damage by detecting ongoing denervation. Besides, our subgroup analyses also indicated that EMG can significantly improve the diagnostic sensitivity.
Upadhyay et al. [23] included 156 patients and published the first article on the diagnostic utility of Gold Coast Criteria in MND, in which the sensitivity of revised criteria was 88.2% (135/153) and the diagnostic accuracy was 89%. In a later study of Hannaford et al. including 506 patients, the sensitivity of Gold Coast criteria (92%) was comparable to that of the possible rEEC (88.6%) and Awaji criteria (90.3%) [24]. Specifically, in the latter study, the sensitivity of the Gold Coast criteria was shown to be similar in ALS patients with shorter (< 18 months) and longer (> 18 months) disease durations, as well as in patients with lower (functional scores > 38) and higher severity (functional scores < 38), implying a robust diagnostic utility of the novel criteria throughout the disease process, even in early stages of ALS when functional disability might be still mild [24].
Our result was inconsistent with Hannaford’ study, which showed that the specificity was comparable across the 3 criteria (88.5% for Gold Coast, 96.2% and 95.5% for liberalized rEEC and Awaji). Notably, the specificities of the ALS diagnostic criteria in our study were much lower than those reported previously (all above 80%) [5, 14, 25]. This difference may arise from the different inclusion criteria used. Previous studies assessing the specificity of ALS diagnostic criteria included (1) suspected diagnosis of ALS by the referring physician and neuromuscular disorder, defined as muscle weakness and wasting for at least 6 months in at least 1 body region [14]; or (2) patients who had been referred with MND included in the differential diagnosis [5]. In contrast, our study only included patients highly suspected as ALS, and this is one of the strengths of this study. The ALS diagnostic criteria were designed to be highly specific. Application of full clinical assessments, including laboratory investigations, neuroimaging, muscle biopsy and experimental immunotherapy, to exclude “mimic disorders” is an essential precondition during diagnostic evaluation. We examined the diagnostic criteria in the very population in which the test will be used (those with suspected ALS), and thus provided the most robust estimates of the test’s sensitivity and specificity in clinical settings. In such cases, however, a higher risk of false-positive ALS diagnosis of the Gold Coast criteria would be expected as a result of less comprehensive requirements. A patient with cervical spondylotic myelopathy may fulfil the criteria of disease progression with UMN and LMN signs in one region and an atypical MMN can sometimes fulfil the criteria of progressive LMN damage in two regions. Only longitudinal follow-ups will tell the truth.
Another factor that may contribute to the discrepancies of specificity is the different races of participants. There is evidence that Chinese ALS patients have younger age of disease onset but a longer median survival [26], and more importantly, higher proportions of progressive muscular atrophy (13.1% here compared to 5%–8.9% in previous studies [24, 27]) and other atypical MND in phenotypes [28]. Thus, there are still challenges for differential diagnosis of ALS, particularly the LMN form. The Gold Coast criteria showed a high sensitivity in those who did not achieve a rEEC diagnostic category, highlighting its diagnostic utility in atypical ALS.
It should be noted that 14 patients did not get an established diagnosis, and 2 of them even did not have EMG (No. 11 and 17 in supplement data), but since none of them manifested disease progression, it seems reasonable to determine them as at least not classic ALS. Genetic tests may be beneficial. For example, the SOD1 p.H46R mutation is consistently associated with a relatively benign form of ALS with slow progression [29]. Considering the atypical manifestations and incomplete investigations of these non-ALS diseases in the present study, the reduced specificity of the Gold Coast criteria is less likely to impact patient recruitment in clinical trials. Another limitation of the study lies in the reliability of the clinical and EMG findings because of the retrospective design. The low interobserver reliability of UMN signs may also be a potential source of bias in such studies [15]. Further unbiased, prospective studies are warranted.