2.1 Study population

Incident ALS cases, diagnosed between 2011 and 2015 according to the revised El Escorial criteria,¹⁴ were retrospectively selected from a prospective population-based registry in the Apulia region (SLAP). The SLAP registry was established in 1997, and the surveillance began on 1 January 1998. Cases were matched by sex and age (±3 years) with a cohort of healthy volunteers. Additional methodological information on the SLAP registry was published elsewhere.¹⁵ ALS patients fulfilling one of the following criteria were excluded from this study: (1) genetic ALS; (2) a concomitant diagnosis of any type of dementia including frontotemporal dementia; (3) history of major psychiatric disorders; and (4) therapy with antidepressant at the time of enrolment and/or over the previous three months (i.e. selective serotonin reuptake inhibitors, SSRIs). For the control group, the exclusion criteria were as follows: (1) presence of a first and/or a second-degree relative with a diagnosis of ALS, frontotemporal dementia or Alzheimer's disease; (2) history of major psychiatric disorders; (3) therapy with antidepressant at the time of enrolment and/or over the previous three months (i.e. SSRIs); and (4) presence of major cardiac, renal, liver or other systemic diseases.

2.2 Procedures of assessment

Cases and controls underwent a detailed interview about familiar and personal history. All cases underwent a neurological examination by ALS-expert neurologists, blinded to any biochemical result, who focused on identifying signs of upper motor neuron (UMN) and lower motor neuron (LMN) involvement and their distribution over several body regions. Diagnosis was made according to the revised El Escorial Criteria.¹⁴

Based on the site of symptom onset, patients were classified as (1) ‘bulbar onset’, when the onset of symptoms was in the bulbar region, or (2) ‘spinal onset’ when the onset of symptoms was in cervical, thoracic or lumbar regions. The spreading pattern of the disease was described using two variables: time to diffusion (TTD), defined as the time of symptom spreading from the onset region to a second one, and time to generalization (TTG), defined as the time of symptom spreading from the spinal or bulbar localization to both.¹⁶ As already described, these two clinical variables were based mostly on the personal history of the patient or on the neurological examination at baseline in a minority of cases (mostly when the neurologist detected signs, as fasciculation or spasticity in one region referred as not affected by the patient); in this case, TTD and/or TTG were considered to be present at the time of enrolment.¹⁷

Six clinical phenotypes were considered at the time of assessment: 1—ALS with prevalence of upper motor neuron signs, 2—ALS with prevalence of lower motor neuron signs, 3—‘flail arm’, 4—‘flail leg’, 5—bulbar ALS (patients who had not developed any spinal involvement in the first 6 months from the onset of symptoms and who had developed pyramidal signs before or after 6 months from symptoms’ onset) and 6—classical ALS.¹⁸

Functional status was evaluated based on the total score of the revised ALS functional rating scale (ALSFRS-R),¹⁹ whereas muscular impairment was assessed using the manual muscle testing (MMT).²⁰ Respiratory involvement was assessed based on forced vital capacity (FVC; percentage of the predicted value) and sniff nasal inspiratory pressure (SNIP).

The disease duration was calculated as the difference (expressed in months) between the date of assessment and the date of symptom onset. The onset-diagnosis interval (ODI) was defined as the time-difference (in days) between date of symptom onset and date of diagnosis. To standardize the disease progression, the progression rate was calculated as 48 minus the ALSFRS-R score at the time of assessment and divided by disease duration from symptom onset.¹¹

Other clinical information such as the presence/date of percutaneous endoscopic gastrostomy (PEG) tube placement and/or tracheostomy was collected after 24 months from the enrolment, using a telephone interview. Staging of the disease at the time of enrolment was based on the King's staging system.²¹

Mortality data were checked using a medical administrative database of Apulia region (Edotto) in which dates of death are registered. Censored date was 03 May 2019. We assumed that no ALS patient migrated outside the region during the study period.

The study was approved by the Institutional Review Board of the ‘Azienda Sanitaria Locale, Lecce’. Written informed consent was obtained from all participants or their legal next of kin if they were unable.

2.3 Sample collection and storage

Venous blood was drawn by venipuncture from all cases and controls; blood samples were collected in EDTA vacutainers, which were immediately centrifuged for 15 min at ~2000 g at room temperature within 1h. After centrifugation, plasma was removed, aliquoted (0.5 ml/aliquot) into screw-cap polypropylene tubes and stored at −80°C until biochemical analyses. Samples were thawed at room temperature only once before analysis.

2.4 pNfH analysis

An enzyme-linked immunosorbent assay (ELISA) was used to quantify the pNfH plasma level [Neurofilament (pNf-H) high sensitive ELISA, REF EQ6562-9601, For Research Use Only, Euroimmun AG], according to the manufacturer's protocol. The lower detection limit was 6 pg/ml. Briefly, in the first analysis step, the calibrators, controls and samples were diluted with monoclonal biotin-labelled anti-pNfH antibodies and added to microplate wells coated with polyclonal anti-pNfH antibodies. In this process, pNfH is bound in a complex. In a second incubation, streptavidin peroxidase conjugate binds the biotin. A following incubation with substrate and chromogen promotes a colour reaction. The colour intensity is proportional to the pNfH concentration in the sample. The absorbance is then measured at 450 nm. Plasma pNfH concentrations were presented as pg/ml. Calibrators, controls and samples were measured in duplicate for each test run. The analytical performance of the assay was verified with the within-run, between-day and within-laboratory precision, using two quality controls, high positive (C1) and low positive (C2), tested in triplicate in five consecutive days (CLSI EP15-A).²² The within-run CV, between-day CV and within-laboratory CV were 2.2%, 2.9% and 3.4%, respectively, for the C1 control (mean concentration 91.05 pg/ml), and 1.5%, 2.5% and 2.8%, respectively, for the C2 control (mean concentration 42.21 pg/ml).

2.5 Statistical analysis

Data are expressed as mean ±standard deviation or median ±interquartile range as appropriate, while categorical variables with absolute numbers and relative frequencies. All tests are two tailed, and a p < 0.05 was deemed as statistically significant.

The differences of numerical covariate between groups were evaluated with linear ANOVA and Kruskal–Wallis tests, depending on the residuals’ distributions (normal distribution assessed with the Shapiro–Wilk test) in a linear regression model and on variance homogeneity. All p-values were corrected for multiple tests (Bonferroni method).

Correlations between pNfH and numerical covariates were calculated using the Spearman rank correlation coefficient.

The diagnostic performance characteristics of plasma pNfH (as a continuous variable) to discriminate between cases and controls were calculated and compared in a logistic regression analysis setting.

Age and/or gender were also tested as covariates in such context, when significant (p < 0.05). The area under the curve (AUC) constructed with the predicted probabilities of the logistic regression analysis and corresponding performance characteristics were reported if the AUC differed significantly from the AUC constructed with the pNfH values. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and AUC with corresponding 95% CI for plasma pNfH were calculated with receiver operating characteristic (ROC) curves. The highest Youden index was used to calculate the optimal cut-off on a ROC analysis and consequent best serum pNfH concentration (pg/ml) discerning between ALS patients and controls.

Kaplan–Meier (KM) univariate analysis was carried out to determine the effect of plasma pNfH on survival (intended as tracheostomy and/or death events) from date of diagnosis. Log rank tests were used to test for differences among groups. Subsequently, univariable and multivariable analysis with Cox proportional hazards model was performed to estimate the hazard ratios (HR) of pNfH on survival adjusting for clinically relevant covariates. Particularly, different multivariable models were considered with age, plasma pNfH and sex as base model, consequently adding remaining clinically relevant covariates with a hierarchical strategy in order to assess the resulting adjusted hazard ratios (aHR).

All analyses were carried out using R (R development core team, Vienna, Austria) version 3.4.0.

3.1 Population characteristics

A total of 256 subjects were enrolled in the study, 128 ALS patients and 128 healthy controls matched for age (± 3 years) and sex. Demographic and clinical characteristics of the study population are listed in Table 1.

3.2 Plasma pNfH levels in ALS and controls

The median plasma pNfH level was 100.7 pg/ml (IQR: 39.6 to 190.2) for ALS patients and 16.0 pg/ml (IQR: 7.3 to 39.6) for controls. Median values were statistically different between the two groups (p<0.001; Figure 1).

With ROC analysis, consequent to a logistic regression model with case/control status as dependent variable and biomarker as unique predictor, an optimal plasma pNfH cut-off value of 39.74 pg/ml was calculated to discriminate between ALS cases and controls, which resulted in a sensitivity and a specificity of 75% with an area under the curve (AUC) of 0.82 (95% CI = 0.77 – 0.87). The accuracy was 73% whereas the positive predictive and the negative predictive values were 79% and 62% respectively (Figure 2).

3.3 Correlation with demographic characteristics and clinical parameters and group differences in cases.

In cases and controls, the correlation between plasma pNfH and age was very weak and not statistically significant (cases: r = −0.1, p = 0.26; controls: r = 0.057, p = 0.52). Plasma pNfH levels did not differ between males and females both in patients and in controls.

Restricting the analysis to cases, plasma pNfH were found significantly higher in cases with bulbar onset of the disease, compared to cases with spinal onset (p = 0.0033), whereas no differences were found between diagnostic categories, phenotypes at diagnosis or King's staging levels.

Plasma pNfH positively correlated with disease progression rate. The strength of the correlation was mild–moderate, and it was statistically significant (r = 0.19, p = 0.031; Figure 3).

The strength of correlation between plasma pNfH levels and ALSFRS-R total score was lower and not statistically significant (r = −0.13, p = 0.15). However, dividing patients into two subgroups based on the median ALSFRS-R total score (34), mean plasma pNfH were significantly different between the two groups (p = 0.015). In particular, in the group with an ALSFRS-R total score below the median the mean plasma pNfH was 244.617 (326.360 SD) pg/ml vs 142.208 (190.486 SD) pg/ml in the group with an ALSFRS-R total score above the median (Figure S1).

The correlations of plasma pNfH levels and disease duration (r = −0.09, p = 0.33), ODI (r = −0.17, p = 0.052), TTD (r = −0.25, p = 0.358) and TTG (r = −0.37, p = 0.181) were all not statistically significant.

Eleven cases presented very high levels of plasma pNfH (>500 pg/ml). Those cases had a fast progression rate (median 0.58, range 0.26–5.5) and a short disease duration (median 7, range 2–31), although comparable with other cases with much lower pNfH plasma levels (Figure S2).

Restricting the analysis on this cluster of 11 case, the strength of the correlation between plasma pNfH levels and progression rate sharply increased, but the level of significance remained the same, probably due to the very small sample size (r = 0.64, p = 0.033; Figure S3).

3.4 Outcomes and survival analysis

The Kaplan–Meier product limit analysis showed no survival differences after stratifying patients according to the plasma pNfH median value (101 pg/ml; log-rank test ${\chi }^2$ = 0.5; p = 0.49; Figure S4).

In univariate Cox proportional hazard (PH) regression models, higher age (HR 1.04, 95% CI = 1.01 – 1.06, p = 0.001), King's Stages 4a (HR 4.32, 95% CI = 1.57 – 11.886, p = 0.005) and 4b (HR 2.86, 95% CI 1.03 – 7.94, p = 0.044), faster disease progression rate (HR 1.57, 95% CI = 1.26 – 1.96, p < 0.001), presence of NIV (HR 2.62, 95% CI = 1.38 – 4.96, p = 0.003) and presence of PEG (HR 2.74, 95% CI = 1.77 – 4.26, p < 0.001) were independently and significantly associated with a reduced survival (defined as time from diagnosis to death/tracheostomy) in cases. On the contrary, higher ALSFRS-R total scores at clinical evaluation were significantly associated with an increased survival (HR 0.93, 95% CI = 0.91– 0.96, p < 0.001).

A multivariable PH Cox regression model confirmed that age was associated with an increased adjusted HR (aHR) of 1.03 (95% CI = 1–1.06, p = 0.024) and that higher ALSFRS-R scores were associated with an increased survival (aHR 0.94, 95% CI = 0.91 – 0.98, p = 0.001). None of the tested PH Cox regression models showed that higher plasma pNfH concentrations were independently associated with a reduced survival in cases (aHR 1, 95% CI = 1–1, p = 0.060).