#Outline, new data from 4/16 [generally in baseline ftld cdr >+0.5 - 2, M+]

###Q: wanna see whether this thing is any better at tracking change in the genetic FTD ####LME w/ stnd betas. Compare 2 models. 1 w/ this vari, another with Sbs. Maybe something else. could also calculate change scores. Could do first and last difference, just to see over length of clincal trial which has greater effect size of change. then move to sample size.

###Q: any difference in the longitudinal slope between carriers and non-carrier ####LME. Factor vari with carrier vs non carrier. timeXgroup interaction. W/ plot

###Q: so interested in how this sample size maps onto the sample sizes you’ve seen in the DPM paper ####sample size estimates as noted above. Add to what did for DPM.

save below

Comp score appears to be better able to catch change in GRN as compared to FTLD CDR Sbs.

A little messy though, because I think the FTLD CDR Sbs was used in the creation of this score. Though that might be ok.

LME (all beta standardized)

Summary: LME predicting comp score or sum of boxes. Variable of interest is the interaction of time (months) and group (control vs case). Controls are FTLD CDR of 0, mutation negative. Case are mutation positive. If run with full data (up to 30 months), models do not converge with random slope + intercept. When restricted to random intercept, Sbs performs a little better, but results are still nonsig. If limited to 24 months, same convergence issues (or singular boundaries). With 24 month limit, looks like the Comp score might actually be performing (marginally) better. It’s got a wider SE though.

Details: Ran two LMEs, both with standardized variables. Both look at interaction between time and group (carry vs control). First predicts comp score, second predicts Sbs. Model did not converge with random slope + intercept. Reran as just random intercept. When limited to 24 month follow up, beta for comp score interaction is slightly higher than beta for Sbs.

Comp Score LME

Variable

Estimate

Std. Error

df

t value

Pr(>|t|)

(Intercept)

-0.1184

0.0380

1,731.424

-3.1118

0.0019

months_from_prior_visit

-0.0051

0.0027

4,893.486

-1.8837

0.0597

CarryvsControlCase

-0.0240

0.0441

1,818.191

-0.5440

0.5865

BASE_FTLDCDR_SB

-0.0156

0.0209

1,547.181

-0.7457

0.4560

age_at_visit_rng

0.0834

0.0209

1,418.814

3.9981

0.0001

as.factor(sex)2

0.0569

0.0384

1,392.391

1.4800

0.1391

months_from_prior_visit:CarryvsControlCase

0.0438

0.0035

4,896.432

12.4589

0.0000

Sum of Boxes LME

Variable

Estimate

Std. Error

df

t value

Pr(>|t|)

(Intercept)

-0.0380

0.0102

1,719.059

-3.7170

0.0002

months_from_prior_visit

0.0000

0.0007

4,893.427

-0.0123

0.9902

CarryvsControlCase

-0.0139

0.0119

1,806.290

-1.1723

0.2412

BASE_FTLDCDR_SB

0.9422

0.0056

1,533.573

167.7755

0.0000

age_at_visit_rng

0.0289

0.0056

1,404.347

5.1497

0.0000

as.factor(sex)2

0.0104

0.0103

1,378.601

1.0088

0.3133

months_from_prior_visit:CarryvsControlCase

0.0136

0.0009

4,896.597

14.3665

0.0000

LME 2

Same as above, but with GRN vs any other mutation. Limited to 24 months. All participants have a baseline ftld cdr global > 0. No controls included.

Both interactions significant. Comp score is marginally better. This interaction is saying that the GRN group has a greater increase in comp or Sbs than the other mutation groups. This is consistent with what we see in the first figure at the top of the markdown.

Comp Score LME (GRN vs other Mutation)

Variable

Estimate

Std. Error

df

t value

Pr(>|t|)

(Intercept)

-0.2735

0.0706

516.1099

-3.8753

0.0001

months_from_prior_visit

0.0563

0.0046

1,489.0904

12.3201

0.0000

GRNvsOtherGRN

-0.0578

0.0965

552.9475

-0.5992

0.5493

BASE_FTLDCDR_SB

0.0376

0.0412

464.4973

0.9138

0.3613

age_at_visit_rng

0.1460

0.0493

461.2404

2.9638

0.0032

as.factor(sex)2

0.0500

0.0807

449.1687

0.6199

0.5357

months_from_prior_visit:GRNvsOtherGRN

0.0375

0.0101

1,525.9116

3.6997

0.0002

Sum of Boxes LME (GRN vs other Mutation)

Variable

Estimate

Std. Error

df

t value

Pr(>|t|)

(Intercept)

-0.0978

0.0263

515.6416

-3.7154

0.0002

months_from_prior_visit

0.0234

0.0017

1,492.2340

13.6030

0.0000

GRNvsOtherGRN

-0.0106

0.0360

553.9075

-0.2956

0.7676

BASE_FTLDCDR_SB

0.9566

0.0153

462.4687

62.3468

0.0000

age_at_visit_rng

0.0616

0.0184

458.3832

3.3577

0.0009

as.factor(sex)2

0.0073

0.0301

446.8056

0.2441

0.8073

months_from_prior_visit:GRNvsOtherGRN

0.0155

0.0038

1,529.4754

4.0489

0.0001