explore_gmcsf
GMCSF Status as a Binary Predictor
Surival Analysis
Conclusion: In an unadjusted log-rank test, survival does NOT significantly differ by GMC-CSF status.
In a Cox proportional hazard model stratified by race and gender, and adjusted for age_at_bleed, survival does NOT significantly differ by GM-CSF status (p = 0.5522027).
GMCSF Status as a Predictor of Antibody Diagnosis
Likelihood ratio tests of Multinomial Models
Response: antibody
Model Resid. df Resid. Dev Test Df
1 age_at_bleed + gender + race 480 238.1581
2 gmcsf_status + age_at_bleed + gender + race 475 209.7803 1 vs 2 5
LR stat. Pr(Chi)
1
2 28.37779 3.070336e-05Conclusion: In a multinomial logistic regression for antibody diagnosis which was adjusted for gender, race, and age_at_bleed, the resultant likelihood ratio test for the entire antibody variable showed that GMCSF status DOES significantly affects the distribution of antibody status (p = 3.0703361^{-5}.
GMCSF Status as a Predictor of RO52 Positivity
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.44727066 2.49426505 0.1793196 0.85768674
gmcsf_statuspositive 2.12993215 0.84756756 2.5129939 0.01197114
age_at_bleed -0.01357418 0.04500315 -0.3016272 0.76293626
gendermale -0.98177974 0.84540244 -1.1613164 0.24551324
racenot_white -0.24363326 1.00804797 -0.2416882 0.80902180 OR 2.5 % 97.5 %
(Intercept) 1.5640376 0.01382541 345.531445
gmcsf_statuspositive 8.4142959 1.74433476 52.175075
age_at_bleed 0.9865175 0.89479355 1.075347
gendermale 0.3746437 0.06238788 1.885606
racenot_white 0.7837750 0.10567270 6.135206Conclusion: In a binary logistic regression model adjusted for gender, race, and age_at_bleed, GMCSF status DOES significantly predict RO52 status.
When converting OR to RR, we get RR 2.3563327, or a positive GM-CSF results in RR 2.3563327 of having a positive RO52.
GMCSF Status as a Predictor of being on a B Cell Medication
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.85666226 1.35455549 -1.3706801 0.17047470
gmcsf_statuspositive 1.23719677 0.49473575 2.5007224 0.01239403
age_at_bleed 0.02526249 0.02262461 1.1165936 0.26416815
gendermale -0.15917424 0.46586935 -0.3416714 0.73259818
racenot_white -0.41100990 0.47118988 -0.8722808 0.38305519 OR 2.5 % 97.5 %
(Intercept) 0.1561931 0.01021761 2.210914
gmcsf_statuspositive 3.4459401 1.33705315 9.419273
age_at_bleed 1.0255843 0.98107929 1.073316
gendermale 0.8528477 0.33826714 2.123149
racenot_white 0.6629804 0.26046468 1.670121Conclusion: In a binary logistic regression model adjusted for gender, race, and age_at_bleed, GMCSF status DOES significantly predict RO52 status (p = 0.012394).
When converting OR to RR, we get RR 1.8103448, or a positive GM-CSF results in RR 1.8103448 of having been on a B cell medication at the time of bleed draw.
GMCSF Status as a Predictor of FVC Change Over Time
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.284933305 0.089015322 3.2009467 0.002381572
gmcsf_statuspositive 0.010273357 0.028381133 0.3619784 0.718893561
age_at_bleed -0.004564523 0.001438933 -3.1721578 0.002586482
gendermale 0.055368333 0.031283397 1.7698951 0.082842167
racenot_white -0.054987488 0.031103292 -1.7678993 0.083179201
Conclusion: In an ANCOVA adjusted for initial FVC values, (as well as gender, race, and age_at_bleed), GMCSF status does NOT significantly predict change in FVC over time.
In a Wilcoxon Rank-Sum test not adjusted for initial FVC values, GMCSF status does NOT significantly predict change in FVC over time (p = 0.305433915504474).
GMCSF Status as a Predictor of Radiologic Severity of ILD
Value Std. Error t value p value
gmcsf_statuspositive -4.956905e-01 0.47795552 -1.0371059746 0.2996864711
age_at_bleed 2.283776e-05 0.02464979 0.0009264892 0.9992607687
gendermale 3.749472e-01 0.46073598 0.8138006217 0.4157591789
racenot_white 1.651103e+00 0.49739566 3.3194971477 0.0009017973
mild|moderate 5.043850e-01 1.52035960 0.3317537539 0.7400752073
moderate|severe 3.036275e+00 1.55559552 1.9518408816 0.0509570989Conclusion: In an ordinal logistic regression for radiologic ILD severity adjusted for gender, race, and age_at_bleed, GMCSF status does NOT significantly predict radiologic severity of ILD on CT chest (p = 0.2996865).