Table 3 Comparison of patient characteristics between TCGA training and testing cohorts

From: A deep learning model to predict glioma recurrence using integrated genomic and clinical data

 

Training (133)

Testing (29)

p-value

Recurrence outcome [n]

Late

75

17

1.0

Early

58

12

Tumor grade [n]

Grade 2

51

11

0.955

Grade 3

62

13

Grade 4

20

5

Race [n]a

White

121

28

0.872

African American or Black

8

1

American Indian or Alaska Native

2

0

Asian

1

0

Hispanic or Latino [n]a

No

119

27

0.73

Yes

4

1

Histology [n]b

Astrocytoma

59

14

0.903

Glioblastoma

46

8

Oligodendroglioma

19

5

Astrocytoma wildtype

9

2

IDH & 1p19q codeletion status [n]

IDH-mutant, non-codeleted

59

14

0.778

IDH-wildtype

55

10

IDH-mutant, codeleted

19

5

Vital status [n]

Deceased

80

17

1.0

Alive

53

12

Days to recurrence

Median

427.0

468.0

0.585

IQR

622.0

545.0

Age [yr]

Median

47

39

0.191

IQR

22.0

28.0

  1. Chi-square tests of independence were used for all categorical variables, unless expected cell counts were <5 in a 2 × 2 contingency table, in which case we applied Fisher’s exact test. The Mann–Whitney U test was used for age and days to recurrence. All statistical tests were two-sided (significance α = 0.05).
  2. Bracketed numbers next to Training/Testing indicate the number of included patients.
  3. aRace and ethnicity were unavailable for 1 and 11 patients, respectively.
  4. bHistology represents the reassigned labels, as described in Methods and Fig. 2.