Glioma analysis
Ismael Calandri
18/4/2020
##
## Attaching package: 'table1'## The following objects are masked from 'package:base':
##
## units, units<-## Loading required package: ggplot2##
## Attaching package: 'ggplot2'## The following object is masked from 'package:expss':
##
## vars## Loading required package: ggpubr##
## Attaching package: 'ggpubr'## The following object is masked from 'package:expss':
##
## compare_means##
## Attaching package: 'dplyr'## The following objects are masked from 'package:expss':
##
## between, compute, contains, first, last, na_if, recode, vars## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionMetodos:
Estudio de cohorte, se registraron pacientes que ingresaron a cirugia con una imagen preopertaroria y diagnóstico anatomopatologico de Glioma de bajo grado de malignidad (lower grade Glioma) considerando a estos como los Gliomas con idh 1 mutado, a la unidad de neurocirugia de Fleni entre los años 2009 al 2019. Los pacientes fueron clasificados bajo los diagnosticos de Oligodendroglioma u astrocitoma de acuerdo a los criterios de (insertar los que sean), teniendo en cuenta la presencia de …….., …… y de Codelecion-1p19q en el estudio citogenetico del tejido.
Para el analisis estadistico se utilizo el software r y r studio y los paquetes table1, caret y tidiverse. Las variables continuas son reportadas en terminos de media y desvio estandard, las variables categoricas como frecuencias. Se evaluo la frecuencia del signo de missmatch, la localizacion ,lobar y la edad como predictores del diagnostico postoperatorio de astrocytoma. Se determino la sensibilidad, especificidad y valores predictivos positivos y negativos en funcion de la frecuencia de los hallazgos en un modelo de regresion logistica binaria con el diagnostico como variable dependendiente. La magnitud de efecto de cada una de estas variables en el diagnostico de astrocitoma es reportado como un odds ratio con su intervalo de confianza del 95%. Para evaluar conjuntamente todas las variables se construyó un modelo de regresión logistica multivariado incluyendo, el missmatch, el sexo, la edad y la localizacion de la lesión.
Resultados
Dentro de la cohorte estudiada se observo una prevalencia del 0.57% para el diagnóstico de astrocytoma, no hubo diferencias significativas en la edad de ambos grupos (p=0.11) ni en el sexo (p=0.99) (table 1). La presencia de missmatch se asocio con el diagnostico de astrocitoma (OR=42.84[8.04,796.62], p=0.0003). La presencia de este signo presento una sensibilidad=0.66, una especificidad=0.95 y un VPP y VPN de 0.95 y 0.68 respectivamente. El sexo masculino aumento las chances del diagnostico de un astrocitoma (OR=0.23[0.05,0.84] p=0.036). En cuanto a la localización, el diagnostico de astrocitoma se relaciono negativamente con la localizacion frontal(OR=0.44[0.20,0.92], p=0.033) y positivamente con la localizacion Temporal ( OR=5.04[1.77,18.22] p=0.005). No se halló relacion significativa con el resto de las localizacion o la edad de los sujetos. El gráfico 1 resume OR para cada una de las variables. El modelo multivariable de mayor verosimilitud presento una sensibilidad=0.75, una especificidad = 0.93 y un VPP y VPN de 0.93 y 0.73 respectivamente. La tabla 2 resume estos hallazgos
##
## Welch Two Sample t-test
##
## data: base$Age by base$dx
## t = 1.5984, df = 113.34, p-value = 0.1127
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.7299626 6.8273751
## sample estimates:
## mean in group 0 mean in group 1
## 38.55556 35.50685##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: base$Masculino and base$dx
## X-squared = 6.9371e-31, df = 1, p-value = 1Tabla 1, Demographics features
| Astrocytoma (N=73) |
Oligodendroglioma (N=54) |
Overall (N=127) |
|
|---|---|---|---|
| Sex(males) | |||
| 0 | 39 (53.4%) | 28 (51.9%) | 67 (52.8%) |
| 1 | 34 (46.6%) | 26 (48.1%) | 60 (47.2%) |
| Age at diagnosis | |||
| Mean (SD) | 35.5 (10.5) | 38.6 (10.7) | 36.8 (10.7) |
| Median [Min, Max] | 33.0 [18.0, 66.0] | 36.0 [23.0, 67.0] | 35.0 [18.0, 67.0] |
| Frontal | |||
| 0 | 34 (46.6%) | 15 (27.8%) | 49 (38.6%) |
| 1 | 39 (53.4%) | 39 (72.2%) | 78 (61.4%) |
| Parietal | |||
| 0 | 67 (91.8%) | 45 (83.3%) | 112 (88.2%) |
| 1 | 6 (8.2%) | 9 (16.7%) | 15 (11.8%) |
| Temporal | |||
| 0 | 52 (71.2%) | 50 (92.6%) | 102 (80.3%) |
| 1 | 21 (28.8%) | 4 (7.4%) | 25 (19.7%) |
| Insular | |||
| 0 | 66 (90.4%) | 51 (94.4%) | 117 (92.1%) |
| 1 | 7 (9.6%) | 3 (5.6%) | 10 (7.9%) |
| Occipital | |||
| 0 | 72 (98.6%) | 54 (100%) | 126 (99.2%) |
| Missing | 1 (1.4%) | 0 (0%) | 1 (0.8%) |
| Missmatch sign | |||
| 0 | 19 (26.0%) | 37 (68.5%) | 56 (44.1%) |
| 1 | 22 (30.1%) | 1 (1.9%) | 23 (18.1%) |
| Missing | 32 (43.8%) | 16 (29.6%) | 48 (37.8%) |
Modelo 1: Missmatch como predictor de astrocytoma
##
## Call:
## glm(formula = dx ~ Missmatch, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.5042 -0.9104 0.2982 0.2982 1.4703
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.6665 0.2822 -2.361 0.018206 *
## Missmatch1 3.7575 1.0606 3.543 0.000396 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.40 on 78 degrees of freedom
## Residual deviance: 79.97 on 77 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 83.97
##
## Number of Fisher Scoring iterations: 5OR del missmatch
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 0.5135135 0.2894282 0.8814635
## Missmatch1 42.8421032 8.0427260 796.6211022sensibilidad y especificiad del missmatch
## Loading required package: lattice##
## Attaching package: 'caret'## The following object is masked from 'package:survival':
##
## cluster## missmatch
## astrocytoma 0 1
## 0 37 1
## 1 19 22## [1] "sensitivity"## [1] 0.6607143## [1] "specificity"## [1] 0.9565217
## [1] "PPV"## [1] 0.9527055## [1] "NPV"## [1] 0.6801823Localizacion frontal aislada
##
## Call:
## glm(formula = dx ~ Frontal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.5387 -1.1774 0.8549 1.1774 1.1774
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.8183 0.3100 2.640 0.00829 **
## Frontal1 -0.8183 0.3839 -2.132 0.03303 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 173.21 on 126 degrees of freedom
## Residual deviance: 168.50 on 125 degrees of freedom
## AIC: 172.5
##
## Number of Fisher Scoring iterations: 4OR del frontal
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 2.2666667 1.2585386 4.2837042
## Frontal1 0.4411765 0.2039644 0.9246137sensibilidad y especificiad del frontal
## missmatch
## astrocytoma 0 1
## 0 39 15
## 1 39 34## function (data, ...)
## {
## UseMethod("sensitivity")
## }
## <bytecode: 0x0000000027557cf0>
## <environment: namespace:caret>## [1] 0.5## function (data, ...)
## {
## UseMethod("specificity")
## }
## <bytecode: 0x000000002756a758>
## <environment: namespace:caret>## [1] 0.6938776
## [1] "PPV"## [1] 0.6840558## [1] "NPV"## [1] 0.5114571Localizacion temporal aislada
##
## Call:
## glm(formula = dx ~ Temporal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.9145 -1.1941 0.5905 1.1608 1.1608
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.03922 0.19807 0.198 0.84303
## Temporal1 1.61901 0.58039 2.790 0.00528 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 173.21 on 126 degrees of freedom
## Residual deviance: 163.35 on 125 degrees of freedom
## AIC: 167.35
##
## Number of Fisher Scoring iterations: 3OR del temporal
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 1.040000 0.7048887 1.535947
## Temporal1 5.048077 1.7721821 18.226774sensibilidad y especificiad del temporal
## temporal
## astrocytoma 0 1
## 0 50 4
## 1 52 21## function (data, ...)
## {
## UseMethod("sensitivity")
## }
## <bytecode: 0x0000000027557cf0>
## <environment: namespace:caret>## [1] 0.4901961## function (data, ...)
## {
## UseMethod("specificity")
## }
## <bytecode: 0x000000002756a758>
## <environment: namespace:caret>## [1] 0.84## [1] "PPV"## [1] 0.8024191## [1] "NPV"## [1] 0.5541677Localizacion parietal aislada
##
## Call:
## glm(formula = dx ~ Parietal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.350 -1.350 1.014 1.014 1.354
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.3980 0.1927 2.065 0.0389 *
## Parietal1 -0.8035 0.5612 -1.432 0.1522
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 173.21 on 126 degrees of freedom
## Residual deviance: 171.11 on 125 degrees of freedom
## AIC: 175.11
##
## Number of Fisher Scoring iterations: 4OR del parietal
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 1.4888889 1.0241104 2.185024
## Parietal1 0.4477612 0.1413922 1.327398sensibilidad y especificiad del parietal
## missmatch
## astrocytoma 0 1
## 0 9 45
## 1 6 67## function (data, ...)
## {
## UseMethod("sensitivity")
## }
## <bytecode: 0x0000000027557cf0>
## <environment: namespace:caret>## [1] 0.6## function (data, ...)
## {
## UseMethod("specificity")
## }
## <bytecode: 0x000000002756a758>
## <environment: namespace:caret>## [1] 0.5982143
## [1] "PPV"## [1] 0.6643771## [1] "NPV"## [1] 0.5301218Localizacion insular aislada
##
## Call:
## glm(formula = dx ~ Insular, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.552 -1.289 1.070 1.070 1.070
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.2578 0.1864 1.383 0.167
## Insular1 0.5895 0.7148 0.825 0.410
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 173.21 on 126 degrees of freedom
## Residual deviance: 172.49 on 125 degrees of freedom
## AIC: 176.49
##
## Number of Fisher Scoring iterations: 4OR de la insula
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 1.294118 0.8996545 1.872242
## Insular1 1.803030 0.4756464 8.682533sensibilidad y especificiad de la insula
## missmatch
## astrocytoma 0 1
## 0 37 1
## 1 19 22## function (data, ...)
## {
## UseMethod("sensitivity")
## }
## <bytecode: 0x0000000027557cf0>
## <environment: namespace:caret>## [1] 0.6607143## function (data, ...)
## {
## UseMethod("specificity")
## }
## <bytecode: 0x000000002756a758>
## <environment: namespace:caret>## [1] 0.9565217
## [1] "PPV"## [1] 0.9527055## [1] "NPV"## [1] 0.6801823Graphic 1
Modelo 2 (incluye edad y localizacion)
## [1] "Model significance"## [1] 1.412675e-07##
## Call:
## glm(formula = dx ~ Missmatch + Age + base$Masculino + Frontal +
## Temporal + Parietal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.62773 -0.63902 0.05913 0.50032 1.96062
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.27329 1.73297 -0.158 0.874694
## Missmatch1 4.46376 1.27480 3.502 0.000463 ***
## Age -0.02656 0.02900 -0.916 0.359785
## base$Masculino1 -1.45232 0.69289 -2.096 0.036081 *
## Frontal1 0.89129 1.55258 0.574 0.565919
## Temporal1 3.11389 1.86469 1.670 0.094935 .
## Parietal1 0.84152 1.68335 0.500 0.617140
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 66.824 on 72 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 80.824
##
## Number of Fisher Scoring iterations: 6OR de las variables
## Waiting for profiling to be done...## OR 2.5 % 97.5 %
## (Intercept) 0.7608729 0.01775466 22.6816462
## Missmatch1 86.8130178 11.19941891 2350.2980131
## Age 0.9737917 0.91692443 1.0288821
## base$Masculino1 0.2340272 0.05306655 0.8442238
## Frontal1 2.4382809 0.13432671 85.5437295
## Temporal1 22.5084430 0.74056871 1520.9268442
## Parietal1 2.3198853 0.09271104 94.2432825sensibilidad y especificidad del modelo
## modelo
## astrocytoma 0 1
## 0 36 2
## 1 12 29##
## Attaching package: 'gtsummary'## The following objects are masked from 'package:expss':
##
## contains, vars## Estimate Lower 90% limit Lower 95% limit Upper 95% limit
## Sensitivity 0.7073171 0.5976290 0.5693484 0.8216165
## Specificity 0.9473684 0.8659499 0.8434427 0.9905675## Estimate Lower 90% limit Lower 95% limit Upper 95% limit
## NPV 0.7094567 0.6405551 0.6197492 0.7853313
## PPV 0.9468496 0.8795995 0.8501721 0.9824341modelo 3 (mismatch y edad)
## [1] "Model significance"## [1] 2.74564e-07##
## Call:
## glm(formula = dx ~ Missmatch + Age, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.5829 -0.9188 0.2492 0.3868 1.6782
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.15443 0.97295 0.159 0.873885
## Missmatch1 3.79737 1.06739 3.558 0.000374 ***
## Age -0.02249 0.02583 -0.871 0.383865
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 79.187 on 76 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 85.187
##
## Number of Fisher Scoring iterations: 5sensibilidad y especificidad del modelo
## modelo
## astrocytoma 0 1
## 0 37 1
## 1 19 22## [1] "sensitivity"## [1] 0.6607143## [1] "specificity"## [1] 0.9565217
## [1] "PPV"## [1] 0.9736842## [1] "NPV"## [1] 0.5365854Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.
Modelos que faltaban
## [1] "Model significance"## [1] 8.568389e-09##
## Call:
## glm(formula = dx ~ Missmatch + Temporal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.36918 -0.80942 0.09772 0.35304 1.59709
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.9478 0.3147 -3.011 0.002600 **
## Missmatch1 3.6920 1.0752 3.434 0.000596 ***
## Temporal1 2.5978 1.1305 2.298 0.021566 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 72.253 on 76 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 78.253
##
## Number of Fisher Scoring iterations: 6## astrocytoma
## modelo 0 1
## 0 36 14
## 1 2 27## [1] "specificity"## [1] 0.9473684## [1] "sensitivity"## [1] 0.6585366## [1] "Model significance"## [1] 4.121028e-08##
## Call:
## glm(formula = dx ~ Missmatch + Temporal + Frontal, family = binomial,
## data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.32391 -0.82329 0.09526 0.37296 1.67227
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.1145 0.6920 -1.611 0.107281
## Missmatch1 3.7452 1.0975 3.413 0.000643 ***
## Temporal1 2.7625 1.2841 2.151 0.031447 *
## Frontal1 0.2067 0.7569 0.273 0.784774
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 72.177 on 75 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 80.177
##
## Number of Fisher Scoring iterations: 6## astrocytoma
## modelo 0 1
## 0 36 14
## 1 2 27## [1] "specificity"## [1] 0.9473684## [1] "sensitivity"## [1] 0.6585366## [1] "Model significance"## [1] 1.769532e-07##
## Call:
## glm(formula = dx ~ Missmatch + Frontal, family = binomial, data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.6390 -0.8263 0.2499 0.3616 1.5751
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.1427 0.4875 -0.293 0.769819
## Missmatch1 3.5936 1.0666 3.369 0.000754 ***
## Frontal1 -0.7564 0.5856 -1.292 0.196465
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 78.309 on 76 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 84.309
##
## Number of Fisher Scoring iterations: 5## astrocytoma
## modelo 0 1
## 0 37 19
## 1 1 22## [1] "specificity"## [1] 0.9736842## [1] "sensitivity"## [1] 0.5365854modelo oligo
## [1] "Model significance"## [1] 1.769532e-07##
## Call:
## glm(formula = oligo ~ Missmatch + Frontal, family = binomial,
## data = base)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.5751 -0.3616 -0.2499 0.8263 2.6390
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.1427 0.4875 0.293 0.769819
## Missmatch1 -3.5936 1.0666 -3.369 0.000754 ***
## Frontal1 0.7564 0.5856 1.292 0.196465
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 109.403 on 78 degrees of freedom
## Residual deviance: 78.309 on 76 degrees of freedom
## (48 observations deleted due to missingness)
## AIC: 84.309
##
## Number of Fisher Scoring iterations: 5## astrocytoma
## modelo 0 1
## 0 22 1
## 1 19 37## [1] "specificity"## [1] 0.5365854## [1] "sensitivity"## [1] 0.9736842