Here’s my blog from the fascinating session 4B at the Orlando Symposium.
Ammar Al-Chalabi takes the stage to a disco effect of flashing lights that might be more appropriate for the next speaker’s lurid pink suit, but he promises controversy regardless. His starting point is a definition of ALS that is about as uncontroversial as can be, but then he puts this into some historical context and points out how troubling ‘incomplete’ clinical forms can be for classification purposes, particularly if they then evolve into ALS. He wonders how many more syndromes will be removed from the MND diagnostic class given that MMN and KD have both been already withdrawn, keep going and we will all put ourselves out of a job eventually. He also points out how redundant ICD-10 classifications are, perhaps not that controversial, unless you wrote them (I’m guessing some of the audience members did). He sees ALS as a solar system of related conditions. He then presented the results of a brief survey he did amongst 72 global ALS experts to appraise what descriptive terms were actually in current use, the answer being ‘loads’. Confusingly most respondees felt their own approaches to be somewhat illogical (but seemingly irresistible) and what’s more we learnt that Australia is part of Europe. The diagnostic criteria don’t seem to improve matters either, despite serial revisions, and although they should assess the probability of ALS being the phenotype, most neurologists don’t seem the find them useful. In essence there are a set of continuous variables that the clinical community feels the need to binarise (think hypertension) by setting a cut-off. To complicate matters further some of these variables (UMN burden for example) are hard to measure and perhaps non-stationary. Ammar’s proposed solutions include classification by underlying cause (where possible) and perhaps using symmetry as an additional phenotypic marker of slow progression. Mamede de Carvalho was clearly listening though and pointed out that monomelic disease can contradict this. Also we were urged to consider the intention behind the El-E criteria – for inclusion/exclusion to clinical trials, although as Ammar pointed out the label on the tin says ‘diagnostic criteria’.
Next up was Rick Bedlack of ALSuntangled fame. This presentation was of particular interest to me as it described a novel analysis of the PROACT database. I’ve also taken a look at this collection of outcome measures from numerous clinical trials.., and we chose to concentrate on the reliability of initial rates of disease progression and how best to measure this rate over time, especially given that a good proportion of participants will drop out or die during the trial period. Our paper has just been accepted in the ALSFTD journal. I also noticed that a small proportion (under 1%) of participants seemed not to get any worse (or even got better by ALSFRS measures), my immediate assumption was that they probably didn’t have ALS, so I excluded them from further analysis. Rick Bedlack has taken a diametrically opposite approach in analysing data from ONLY these small number of participants, looking to find evidence in support of the apparently 21 sufficiently documented reversals of ALS progression that are reported in the literature. In summary it seemed more likely to happen in men with long disease durations, but he didn’t check the FVC data yet. What’s happening here? He doesn’t know, but likens it to HIV ‘elite controllers’, and thinks it’s a topic worthy of significant further research. Of course there are sceptics in the audience (you can have stable disease and improve ‘function’ with cunning life adaptions), but an on-the-spot show of hands suggested that about 1% of audience members had witnessed similar phenomena in their patients. So. It’s just been published in Neurology.
Team Utrecht was up next with Henk-Jan Westeneng predicting survival in ALS patients by integrating multiple data sources. This looks like proper data-driven personalised medicine and it also seems that Dutch doctors wear fetching short-sleeved white coats(!). The variables were narrowed down using a Cox model but sadly imaging parameters (so far…) don’t seem to improve the prediciton accuracy. Data from 2000 patients showed the ALSFRS slope to be most predictive, but 8 variables were independently contributing and survival in various countries seems quite similar. Questions concerned the effect of imputing missing data (you can tolerate ‘known unknowns’ here apparently) and seemingly 70% of patients like to get an (emphasis on uncertain) estimate of prognosis. A patient-friendly version with readily understandable data is planned for easy web access.
Patricia Andres then described data generated by quantitative strength testing using the ATLIS system with a view to improving the efficiency of future clinical trials. This will be essential to detect small effects within a heterogenous population. She also reminded the audience that QoL measures are REQUIRED in phase 3 clinical trials. The concept of functional cliffs was introduced, for example steps between ASLFRS grades are not even, so the scale is best reserved for large trials that can average out this fluctuation. The pros and cons of quantitative muscle testing were introduced and her dataset of 63 patients with 3 or more visits described. The impact would decrease the sample size required by about a third over using ALSFRS. Basically it just reminds us how useless the ALSFRS is on an individual basis. Conclusions: Invest in outcome measures… and consider use of a lead in phase.
Last up was Susana Pinto from Lisbon, reminding us that respiratory failure limits life-expectancy for people with ALS. They have 15y of clinic data and she agrees there can be a symptomatic improvement, e.g. if you no longer walk much so you don’t feel breathless. Conventionally though we directly measure the strength of ventilatory muscles, they typically did this at 4 month interval visits. FVC and SVC are highly correlated at both time points and generally there’s lots of correlation with the other respiratory measures also. For example both FVC and SVC are dependent on MEP. Correlations are a bit weaker in patients with bulbar onset, SVC might be a bit more accurate for them (less air leak, but then they don’t use whole face masks in Lisbon apparently), but would SNIP be a better measure all round? We don’t know. It also remains uncertain the impact of NIV on these measures.
That’s all, thanks for reading.