Skip to main content

Table 3 Operating Characteristics for Various Predictive Models and Definitions of Substantial BMI Increase

From: Early evaluation of patient risk for substantial weight gain during olanzapine treatment for schizophrenia, schizophreniform, or schizoaffective disorder

  

Evaluated from Study 1

Based on validation dataset (Study 2)

SWGa (kg/m2)

Predictive Modelb

BMIC cut-off

N

Sensitivity,c%

Specificity,d %

PPVe %

NPVf %

+TEST n (%)g

+SWG N (%)h

Sensitivity,i %

Specificity,j %

PPVk %

NPVl %

1

BMIC at 1 wk

.13

661

56.8

63.1

73.9

44.3

329 (49.8)

428 (64.8)

53.1

69.4

75.6

45.5

1

BMIC at 2 wks

.30

660

63.8

68.1

78.7

50.5

347 (52.6)

428 (64.9)

57.8

62.2

72.6

46.0

1

BMIC at 3 wks

.40

661

64.5

68.7

79.1

51.3

349 (52.8)

428 (64.8)

60.9

62.2

73.6

47.9

1

BMIC at 4 wks

.53

660

71.0

74.3

83.5

58.3

363 (55.0)

427 (64.7)

68.8

66.7

78.6

54.6

1

BC alone

N/A

667

59.4

53.8

70.1

42.1

365 (54.7)

431 (64.6)

67.2

54.1

71.7

48.8

1

BMIC at 1 wk+BC

N/A

661

61.0

58.4

72.9

44.9

358 (54.2)

428 (64.8)

70.3

69.4

80.4

56.8

1

BMIC at 2wks+BC

N/A

660

63.6

62.5

75.8

48.2

359 (54.4)

428 (64.9)

65.6

64.9

76.4

52.2

1

BMIC at 3wks+BC

N/A

661

67.8

67.8

79.5

53.4

365 (55.2)

428 (64.8)

64.1

73.0

80.4

54.0

1

BMIC at 4wks+BC

N/A

660

69.8

71.2

81.6

56.3

365 (55.3)

427 (64.7)

67.2

66.7

78.2

53.3

2

BMIC at 1 wk

.18

661

53.4

66.1

56.4

63.3

282 (42.7)

298 (45.1)

50.0

76.0

67.6

60.3

2

BMIC at 2 wks

.39

660

59.6

68.0

60.4

67.3

293 (44.4)

297 (45.0)

60.0

70.6

66.7

64.3

2

BMIC at 3 wks

.52

661

67.2

68.0

63.4

71.5

317 (48.0)

299 (45.2)

66.0

70.6

68.8

67.9

2

BMIC at 4 wks

.69

660

69.9

72.8

67.7

74.9

306 (46.4)

296 (44.9)

66.0

74.0

71.7

68.5

2

BC alone

N/A

667

58.2

52.2

49.7

60.6

350 (52.5)

299 (44.8)

72.0

39.2

53.7

58.5

2

BMIC at 1wk+BC

N/A

661

59.1

60.9

55.4

64.4

318 (48.1)

298 (45.1)

62.0

62.0

62.0

62.0

2

BMIC at 2wks+BC

N/A

660

62.3

64.5

58.9

67.6

314 (47.6)

297 (45.0)

66.0

70.6

68.8

67.9

2

BMIC at 3wks+BC

N/A

661

66.6

67.7

63.0

71.0

316 (47.8)

299 (45.2)

64.0

70.6

68.1

66.7

2

BMIC at 4wks+BC

N/A

660

69.9

71.4

66.6

74.5

311 (47.1)

296 (44.9)

72.0

76.0

75.0

73.1

3

BMIC at 1 wk

.24

661

54.6

66.3

38.3

79.3

261 (39.5)

183 (27.7)

44.8

69.0

37.1

75.4

3

BMIC at 2 wks

.48

660

58.2

68.6

41.4

81.2

256 (38.8)

182 (27.6)

62.1

69.4

45.0

82.0

3

BMIC at 3 wks

.64

661

66.9

69.2

45.6

84.4

270 (40.9)

184 (27.8)

72.4

73.6

52.5

86.9

3

BMIC at 4 wks

.83

660

66.5

70.5

46.2

84.7

262 (39.7)

182 (27.6)

69.0

78.9

57.1

86.2

3

BC alone

N/A

667

58.7

54.9

33.1

77.7

326 (48.9)

184 (27.6)

69.0

43.1

32.8

77.5

3

BMIC at 1wk+BC

N/A

661

64.5

65.5

41.7

82.8

283 (42.8)

183 (27.7)

58.6

56.3

35.4

76.9

3

BMIC at 2wks+BC

N/A

660

62.6

67.2

42.1

82.5

271 (41.1)

182 (27.6)

65.5

68.1

45.2

83.1

3

BMIC at 3wks+BC

N/A

661

66.3

69.0

45.2

84.1

270 (40.9)

184 (27.8)

69.0

69.4

47.6

84.8

3

BMIC at 4wks+BC

N/A

660

66.5

68.6

44.7

84.3

271 (41.1)

182 (27.6)

72.4

71.8

51.2

86.4

  1. aSWG (substantial weight gain) defined as an increase of at least 1, 2, or 3 kg/m2 in BMI from baseline at Week 30 (Study 1) or Week 28 (Study 2).
  2. bFor each combination of SWG definition (≥ 1, 2 or 3 kg/m2) and set of predictors (baseline characteristics [BC] and/or BMI Change [BMIC] at 1, 2, 3, or 4 weeks), the prediction rules (classification scores) were constructed with linear discriminant analysis models. When early BMI Change is used as a single predictor, the estimated cut-off for early BMI change is shown at the time at which early BMI change was assessed. Baseline characteristics, when included in the prediction rules for substantial BMI increase (alone or in combination with early BMI change) were age, sex, BMI, and Caucasian ethnicity.
  3. cSensitivity, proportion of patients with SWG that were correctly predicted (estimated using leave-one-out cross validation).
  4. dSpecificity, proportion of patients with no SWG that were correctly predicted (estimated using leave-one-out cross validation).
  5. ePositive predictive value, proportion of patients predicted to have SWG who had substantial weight gain (estimated using leave-one-out cross validation).
  6. fNegative predictive value, proportion of patients predicted not to have SWG who did not have substantial weight gain (estimated using leave-one-out cross validation).
  7. gNumber of patients (%) predicted to have substantial weight change.
  8. hNumber of patients (%) who met the criteria for substantial BMI change as defined in the first column of this table. Number and percent of patients with SWG for a given defined SWG vary slightly because values are based on subjects included in the model who had non-missing values,
  9. iSensitivity of prediction rule derived from Study 1 data as estimated from validation data (Study2).
  10. jSpecificity of prediction rule derived from Study 1 data as estimated from validation data (Study2).
  11. kPositive predictive value of prediction rule derived from Study 1 as estimated from validation data (Study2).
  12. lNegative predictive value of prediction rule derived from Study 1 as estimated from validation data (Study2).
  13. Abbreviations: SWG, substantial weight gain; PPV, positive predictive value; NPV, negative predictive value; N/A, not applicable.
Back to article page

BMC Psychiatry

ISSN: 1471-244X

Contact us