Multiple analytes were found to be statistically significantly associated with ALS severity after multiple test correction. With biggest effects, creatinine metabolism markers, creatinine and creatine kinase were reversely associated with the limb FRS component for limb onset patients; inflammation markers were directly associated with disease severity
Associations between clinical chemistry analytes and ALS severity in pooled clinical trials
1. Analysis of biochemical analytes
associated with ALS progression
Vladimir Morozov
Boston, 2016
2. Amyotrophic lateral sclerosis (ALS),Motor Neuron Disease (MND),Lou
Gehrig's disease
• Fatal neurodegenerative disorder manifested via
muscle weakness.
• The cause and mechanism is not known. Muscle
weakness is thought to be result of motor neuron
degeneration.
• Median ALS survival is 3 years from the onset of
symptoms. However, about 10 percent survive for
10 or more years
• Depending on appearance of first signs of disease
patients can be broadly classified into limb and
bulbar onset groups
3. Measures of ALS progression
• ALS Functional Rating Scale (FRS) was
proposed as proposed as measure of ALS
progression and severity. The scale is
composed from 10 (12 for the revised FRS)
questions covering limb, speech, respiratory
function
• Most people with ALS die from respiratory
failure when the diaphragm and chest wall
muscles fail. So Forced Vital Capacity (VFC)
can be used as measure of ALS severity and
predictor of survival
4. Why analyze biochemical analytes?
• Can provide easily and objectively measured
endpoints for clinical trials. Currently self
functional rating (FRS) and forced vital
capacity (FVC) are used
• Can provide hints about a pathogenic
mechanism of ALS
5. Pooled Resource Open-Access ALS Clinical Trials Database, PRO-ACT
• Created by Prize4Life in partnership with the
Northeast ALS Consortium and the Neurological
Clinical Research Institute at Massachusetts
General Hospital
• Placebo and treatment-arm data from 18 Phase
II/III clinical trials. Only placebo/treatment group
information. Trial ID/name is not provided
• Demographic, laboratory test analytes, medical
and family history, and other data elements.
More than 10 million longitudinally collected
data points.
6. Aggregated FRS and FVC by years from ALS onset
rvival :3 years. Afterward “Drop off” upward effect
Limb functions are more affected than bulbar
7. Association between analyte and ALS progression.
• Hypothesis to test: concentration of an analyte is correlated with
ALS progression
• Each lab test analyte (log scaled concentration) is fit into two
mixed effect linear models with/out weight effect:
log(Analyte)~Sex*Age*severity+Sex*Age*Weight
log(Analyte)~Sex*Age*severity+Sex*Age*Weight
“severity”=1-FRS/40 .
• Body weight might be useful explanatory variable for biochemical
analytes because many analytes are known to be correlated with
weight. Weight decreases during ALS progression. It can be
attributed to muscle loss and/or decreased food intake/digestion
due to salivation difficulties.
• Study participants is the model random effect. Hence the models
can be thought as multivariate linear regressions fitted separately
for each participants so that intercepts can be different, but
slopes are constrained to be same between study participants
8. Some associations disappeared after
adjusting for weight
FRS effect sizes and p-values
CREATINE
KINASE
CREATININ
E
ALKALINEPHOSHPOTASE
GAMMAGLUTAMYLTRANSFERASE
9. Coefficients from model:
Some associations have significant gender:FRS
interaction. However the associations stay
significant for both genders
10. Creatinine by FRS, stratified by age and gender
Creatinine plasma
concentration is
associated with
muscle mass. So it is
less in women and
older people. There is
no obvious non-
linearity in association
with ALS progression
11. Creatine kinase(CK) by FRS, stratified by age, gender
There might be no-linear
relation with FRS when
creatine kinase don’t
change in the beginning of
disease.
12. Creatine kinase(CK) by FRS, stratified by age, site of onset
Serum total creatine kinase
activity composes from brain
and muscle isoenzymes. The
lab test is NOT specific for the
type of CK. Though it make to
sense to look separately for
limb and bulbar groups of ALS
patients.
Can it distinguish bulbar vs limb sub-groups of ALS patients
13. G93A SOD1 gene expression for the protein analytes associated with
FRS
•Brain, spinal cord,
sciatic nerve,
muscle Affymetrix
time profile
•Statistically
significant
monotonic
patterns are
selected
•Sign of effects
agree with the
blood association
signs
14. Conclusions
• There are a few analytes that are statistically significantly
associated with ALS functional score
• Creatinine and creatine kinase have the largest association
effect sizes
• Decrease of creatinine is known to be associated with
muscle loss, its elevation is used as marker of acute muscle
damage. Independent studies show decrease in ALS
patients comparing to normal.
• Creatine kinase, alkaline phosphatase, gamma glutamyl
transferase mRNA concentration are associated with ALS
progression in the SOD1 mouse model when the association
directions agree with the lab test analyte associations
15. Modifiers of ALS progression
Test if a concentration of an analyte in beginning of clinical trial is associated with speed
of ALS progression (slope of the functional score) :
• FRS is fit into mixed effect linear model:
FRS~delta+Analyte180+delta:Analyte180, “Analyte180”: averaged log scaled analyte
concentration during first 180 days of trial.
• Study participants is the model random effect. Hence the models can be thought as
multivariate linear regressions fitted separately for each participants so that
intercepts can be different, but slopes are constrained to be same between study
participants
17. Dependence of FRS slope from phosphorus concentration
Linear relationship. It speaks against an that the slope
modifying effect is result of an acute condition(e.g. kidney
failure)
18. FRS slope effects from a model including known modifiers (Dec 2016
PRO-ACT version)
Value Std.Error DF t-value p-value
(Intercept) 2.966092 0.037255 19968 79.6151 0.00E+00
AgeQ(49.6,77.4] 0.032739 0.029518 2390 1.109112 2.67E-01
delta -0.54969 0.012893 19968 -42.6347 0.00E+00
SexMale 0.135697 0.028341 2390 4.787997 1.79E-06
SiteOfOnsetBulbar 0.161174 0.03297 2390 4.888559 1.08E-06
SiteOfOnsetBulbar,Limb -0.15815 0.114053 2390 -1.38663 1.66E-01
cov1Q(1.2,1.62] -0.21717 0.027096 2390 -8.01478 1.71E-15
AgeQ(49.6,77.4]:delta -0.02941 0.010157 19968 -2.8952 3.79E-03
delta:SexMale -0.06432 0.010217 19968 -6.29494 3.14E-10
delta:SiteOfOnsetBulbar -0.2554 0.011842 19968 -21.567 5.34E-102
delta:SiteOfOnsetBulbar,Limb -0.36256 0.052657 19968 -6.88532 5.94E-12
delta:cov1Q(1.2,1.62] -0.17604 0.009654 19968 -18.2353 1.07E-73
FRS~delta*AgeQuntile+delta*SiteOfOnset + delta*Sex+delta*AnalyteQuantile
Age and site of onset are known modifiers of ALS progression. Age, analyte
concentration were converted into 2-class variables based on median values for
better interpretation and comparison.
There are small but statically significant associations between
phosphorus with age(-0.02 older than 50) and gender (-0.04 in
male). However these association are in opposite directions with the
slope modifier effects
19. Forced Vital Capacity(FVC) slope effects
Value Std.Error DF t-value p-value
(Intercept) 90.8 1.7 6390 53.4 0
AgeQ(52,84.1] 1.24 1.32 963 0.942 0.346
delta -0.0708 0.00294 6390 -24.1 1.16E-122
SexMale -2.16 1.36 963 -1.59 0.111
SiteOfOnsetBulbar -3.5 1.65 963 -2.12 0.0341
SiteOfOnsetBulbar,Limb -11 4.82 963 -2.29 0.0224
cov1Q(1.21,1.66] -3.18 1.29 963 -2.47 0.0138
AgeQ(52,84.1]:delta -0.00039 0.00231 6390 -0.17 0.865
delta:SexMale 0.0212 0.00237 6390 8.95 4.56E-19
delta:SiteOfOnsetBulbar -0.0274 0.00284 6390 -9.65 7.22E-22
delta:SiteOfOnsetBulbar,Limb 0.00579 0.0176 6390 0.33 0.741
delta:cov1Q(1.21,1.66] -0.0112 0.00224 6390 -4.99 6.27E-07
The same explanatory variable as in the previous FRS model.
However only 969 participants have both FVC and phosphorus
reported comparing 2396 for FRS. This association has less
statistical power
Bulbar onset and phosphorus are still statistically
significant modifiers of the ALS progression slope
20. Cox proportional hazards regression model with phosphorus and known
modifiers of ALS progression
The survival data are unbalanced, 1040 death and 80
censored. Censoring time for non-death were not
provided, so were replaced by last observation time
during trial
Hazard
Ratio CI (lower) CI (upper) Pr(>|z|)
Sex: Male 1.183 1.040 1.346 0.0107
AgeQ: (55.4,84.1] 1.127 0.983 1.293 0.0873
SiteOfOnset: Bulbar 1.662 1.437 1.921 6.93E-12
SiteOfOnset: Bulbar,Limb 1.826 1.107 3.012 0.0184
SiteOfOnset: Other 2.450 1.920 3.128 6.29E-13
cov1Q: (1.22,1.6] 1.142 1.009 1.293 0.036
21. Phosphorus by years from onset. Analysis of long-term survivors
long-term survival cutoff
22. Statistics for association between “long survival” group and phosphorus
concentration
•Define “long survival” group as participants with more than 5 years
since disease onset
•Run logistic regression for odds to be in the “long survival group”
explained by phosphorus quintile
or Welch t-test for difference in phosphorus concentration
between “long survival” and the rest:
cov1 by I(YearsFromOnset > 5)
t = 4.1714, df = 79.094, p-value = 7.702e-05
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
0.03243997 0.09165205
Odds RatioCI (lower) CI (upper) Pr(>|z|)
(Intercept) 0.028 0.021 0.036<0.001
cov1Q: (1.21,1.66] 0.498 0.301 0.800 0.005
23. Possible pitfalls of interpreting measures made in beginning of trial
• Confounding pathology. For example high phosphorus is result
of kidney failure. Then participants with kidney failure will die
faster from ALS and have steeper functional decline .
Phosphorus spikes caused by kidney failure are much higher than
observed in the data . If such cases happened, they probably were
removed from the database. Also the plots of relationships
between phosphorus concentration and slope of progression and
survival are linear, there are no spikes for high concentrations
• Accumulated changes. Phosphorus concentration is slightly
positively associated with FRS. Fast progressing participants
would have higher phosphorus in beginning of clinical trial
because of this association, then these participant would
continue to progress faster.
However analysis of participants who enters into studies shortly
after disease diagnosis came with same sign and still statistically
significant effects as for the whole data set
24. Conclusion
• Higher level of serum phosphorus at
beginning of trial is associated with faster FRS
decline and shorter survival (time since
disease onset)
• High phosphate levels can be avoided
with phosphate binders and dietary restriction
of phosphate.