Table 1 Accuracy of prognoses rules.

From: Prediction of short-term antidepressant response using probabilistic graphical models with replication across multiple drugs and treatment settings

Training (PGRN-AMPS and ISPC): SSRI (Citalopram/Escitalopram) N = 947 outpatients

Baseline strata

4-week strata

Number of patients making transition

Most-likely outcome

Prognoses Rule and Coverage

Probability of most-likely outcome (accuracy = 100*probabilty)

Odds ratio (OR)

95% confidence interval

p value of accuracy with NIR = 0.53

Change in symptom severity (Baseline—4 week)

Number of symptoms needing the change

Coverage (Fraction of patients covered by prognoses rule)

A3

B3

96

Nonresponse

≤1

≥3

0.75

0.60

6.90

(2.03, 23.74)

0.05

B2

104

Response

≥2

≥2

0.65

0.85

3.27

(1.26, 8.5)

4.83E-13

B1

59

Remission

≥2

≥2

0.86

0.75

3.63

(0.9, 16)

5.64E-07

A2

B3

88

Nonresponse

≤1

≥3

0.87

0.80

7.70

(1.46, 40.41)

1.12E-09

B2

148

Response

≥1

≥2

0.95

0.85

5.40

(2.3, 12.87)

4.83E-13

B1

102

Remission

≥2

≥2

0.88

0.75

2.20

(0.6, 7.66)

5.64E-07

A1

B2 or B3

100

Nonresponse

≤1

≥3

0.94

0.70

4.71

(0.81, 27.23)

7.85E-05

B1

160

Remission

≥2

≥1

0.93

0.82

4.35

(1.18, 16)

6.15E-11

Testing (ISPC+Eli Lilly): SSRIs (Escitalopram, fluoxetine, sertraline, paroxetine) N = 464 outpatients

A3

B3

82

Nonresponse

≤1

≥3

0.89

0.72

5.30

(1.2, 23.2)

1.26E-05

B2

63

Response

≥2

≥2

0.86

0.86

6.50

(1.3, 33)

8.28E-14

B1

41

Remission

≥2

≥2

0.83

0.70

6.00

(1, 36.3)

7.85E-05

A2

B3

39

Nonresponse

≤1

≥3

0.87

0.85

12.60

(1.14, 65.9)

4.83E-13

B2

98

Response

≥1

≥2

0.80

0.62

4.10

(1.13, 12.7)

0.02

B1

60

Remission

≥2

≥2

0.67

0.63

2.50

(0.84, 7.6)

0.01

A1

B2 or B3

33

Nonresponse

≤1

≥3

0.85

0.70

7.20

(0.64, 82)

7.85E-05

B1

48

Remission

≥2

≥1

0.67

0.84

4.20

(1.1, 16.5)

2.61E-12

Testing (Eli Lilly): SNRI (Duloxetine) N = 1067 outpatients

A3

B3

201

Nonresponse

≤1

≥3

0.86

0.73

1.80

(1, 4.2)

4.69E-06

B2

125

Response

≥2

≥2

0.65

0.84

2.80

(1.2, 6.8)

2.61E-12

B1

82

Remission

≥2

≥2

0.56

0.83

2.30

(0.9, 6.5)

1.31E-11

A2

B3

156

Nonresponse

≤1

≥3

0.95

0.78

9.10

(1.6, 9.14)

1.59E-08

B2

237

Response

≥1

≥2

0.95

0.62

2.20

(1.2, 4.04)

0.02

B1

129

Remission

≥2

≥2

0.67

0.77

2.14

(0.93, 4.9)

5.51E-08

A1

B2 or B3

89

Nonresponse

≤1

≥3

0.77

0.63

5.05

(1.4, 17.9)

0.01

B1

48

Remission

≥2

≥1

0.88

0.83

25.00

(1.9, 35)

1.31E-11

Testing (MARS): COMEDICATIONS N = 465 inpatients

A3

B3

57

Nonresponse

≤1

≥3

0.88

0.66

4.80

(0.9, 27)

0.002

B2

71

Response

≥2

≥2

0.76

0.73

2.80

(0.84, 8.3)

4.69E-06

B1

29

Remission

≥2

≥2

0.84

0.69

9.00

(0.9, 91)

0.0002

A2

B3

51

Nonresponse

≤1

≥3

0.86

0.76

4.40

(1, 23)

1.81E-07

B2

101

Response

≥1

≥2

0.80

0.60

3.25

(1, 10.6)

0.06

B1

50

Remission

≥2

≥2

0.56

0.57

2.90

(1.1, 8.9)

0.1933479

A1

B2 or B3

49

Nonresponse

≤1

≥3

0.88

0.76

12.80

(0.74, 37)

1.81E-07

B1

57

Remission

≥2

≥1

0.75

0.57

4.60

(1.1, 19)

0.2

Prognoses performance in placebo-treated patients (Eli Lilly N = 575)

A3

B3

96

Nonresponse

≤1

≥3

0.95

0.80

0.40

(0.03, 4.9)

0.60

B2

104

Response

≥2

≥2

0.40

0.69

2.00

(0.45, 8.9)

0.85

B1

59

Remission

≥2

≥2

0.95

0.73

1.40

(0.1, 19)

0.75

A2

B3

88

Nonresponse

≤1

≥3

0.96

0.87

0.20

(0.02, 3.4)

0.80

B2

148

Response

≥1

≥2

0.87

0.47

0.48

(0.13, 1.8)

0.85

B1

102

Remission

≥2

≥2

0.86

0.67

1.20

(0.26, 5.7)

0.75

A1

B2 or B3

100

Nonresponse

≤1

≥3

0.86

0.67

0.67

(0.23, 1.92)

0.70

B1

160

Remission

≥2

≥1

0.91

0.71

2.50

(0.7, 13.85)

0.82

  1. Prognoses performance of prognostic symptoms in patients making specific transitions between baseline and 4-week strata. The ranges of depression severity scores in each strata are as follows: A1 [14–18], A2 [19–24], A3 [25–39]; B1 [0–8], B2 [9–15], B3 [16–31]. The OR represents the odds that the expected treatment outcome at 8 weeks will occur if patients are covered by the prognoses rule, compared to the odds of the same outcome occurring in patients not covered by the prognoses rule. The statistical significance (p-value) of the prognoses’ accuracy was established using the null information rate (NIR).
Back to article page