Skip to main content

Table 3 Diagnostic accuracy when basing endoscopy referral on varying SCD probability thresholds for the basic and the five faecal biomarker extended models, as observed in 810 Dutch patients with lower abdominal complaints referred for endoscopy in the CEDAR studya

From: Is there an added value of faecal calprotectin and haemoglobin in the diagnostic work-up for primary care patients suspected of significant colorectal disease? A cross-sectional diagnostic study

Ā 

Referred, % (95Ā % CI)

SCD detected, n

Missed SCDb, n

Accuracy for SCD

Diagnostic model SCD probability threshold

Total

CRC

IBD

D

AA

Sensitivity, % (95Ā % CI)

PPV, % (95Ā % CI)

Specificity, % (95Ā % CI)

NPV, % (95Ā % CI)

Basic

ā€ƒā‰„2.5Ā %

98.0 (96.4ā€“99.2)

141

0

ā€“

ā€“

ā€“

ā€“

99.7 (96.7ā€“100.0)

17.7 (15.2ā€“20.5)

2.3 (1.0ā€“4.1)

98.0 (76.5ā€“100.0)

ā€ƒā‰„5.0Ā %

86.9 (83.9ā€“89.6)

136

5

ā€“

ā€“

2

3

97.0 (92.6ā€“98.8)

19.4 (16.6ā€“22.5)

15.2 (12.1ā€“18.8)

95.9 (90.2ā€“98.5)

ā€ƒā‰„7.5Ā %

72.5 (69.1ā€“75.6)

126

15

1

ā€“

4

10

89.5 (83.2ā€“93.8)

21.5 (18.3ā€“25.0)

31.1 (27.5ā€“35.0)

93.4 (89.2ā€“96.1)

Calprotectin POC extended

ā€ƒā‰„2.5Ā %

97.9 (96.4ā€“98.9)

141

0

ā€“

ā€“

ā€“

ā€“

100.0 (97.3ā€“100.0)

17.8 (15.3ā€“20.6)

2.6 (1.4ā€“4.3)

100.0 (81.1ā€“100.0)

ā€ƒā‰„5.0Ā %

83.6 (80.3ā€“86.6)

138

3

ā€“

ā€“

1

2

97.4 (93.1ā€“99.2)

20.3 (17.4ā€“23.5)

19.3 (15.8ā€“23.3)

97.2 (92.6ā€“99.2)

ā€ƒā‰„7.5Ā %

68.5 (64.9ā€“71.9)

125

16

1

1

5

9

88.1 (81.5ā€“92.7)

22.4 (19.1ā€“26.1)

35.6 (31.7ā€“39.7)

93.4 (89.6ā€“96.0)

Calprotectin ELISA extended

ā€ƒā‰„2.5Ā %

97.6 (96.0ā€“98.8)

141

0

ā€“

ā€“

ā€“

ā€“

100.0 (97.3ā€“100.0)

17.8 (15.3ā€“20.7)

2.9 (1.5ā€“4.8)

100.0 (82.7ā€“100.0)

ā€ƒā‰„5.0Ā %

84.0 (81.1ā€“86.6)

135

6

ā€“

1

2

3

96.3 (91.5ā€“98.7)

20.0 (17.1ā€“23.2)

18.6 (15.5ā€“22.0)

96.0 (90.7ā€“98.6)

ā€ƒā‰„7.5Ā %

68.1 (64.7ā€“71.4)

126

15

ā€“

1

5

9

90.0 (83.7ā€“94.2)

23.0 (19.7ā€“26.7)

36.5 (32.8ā€“40.4)

94.5 (91.0ā€“96.8)

POC FIT extended

ā€ƒā‰„2.5Ā %

92.8 (89.0ā€“96.2)

140

1

ā€“

ā€“

ā€“

1

99.4 (96.2ā€“100.0)

18.6 (16.0ā€“21.6)

8.6 (4.5ā€“13.1)

98.7 (91.2ā€“99.9)

ā€ƒā‰„5.0Ā %

69.9 (64.4ā€“75.3)

131

10

1

ā€“

4

5

93.0 (87.4ā€“96.4)

23.2 (19.6ā€“27.1)

34.9 (28.7ā€“41.4)

96.0 (92.6ā€“97.9)

ā€ƒā‰„7.5Ā %

54.3 (49.6ā€“59.1)

124

17

1

4

6

6

88.1 (81.4ā€“92.8)

28.2 (24.0ā€“32.9)

52.8 (47.4ā€“58.1)

95.5 (92.7ā€“97.2)

Calprotectin POC and POC FIT extended model

ā€ƒā‰„2.5Ā %

92.8 (89.4ā€“95.7)

141

0

ā€“

ā€“

ā€“

ā€“

100.0 (97.3ā€“100.0)

18.8 (16.1ā€“21.8)

8.7 (5.2ā€“12.8)

100.0 (93.4ā€“100.0)

ā€ƒā‰„5.0Ā %

69.6 (64.7ā€“74.3)

132

9

1

ā€“

4

4

93.7 (88.2ā€“96.8)

23.5 (19.9ā€“27.3)

35.5 (29.9ā€“41.2)

96.4 (93.1ā€“98.2)

ā€ƒā‰„7.5Ā %

53.3 (48.7ā€“57.8)

124

17

1

3

7

6

88.4 (81.7ā€“93.1)

28.9 (24.7ā€“33.5)

54.1 (49.1ā€“59.1)

95.7 (93.1ā€“97.4)

Calprotectin ELISA and POC FIT extended model

ā€ƒā‰„2.5Ā %

92.8 (89.5ā€“95.6)

141

0

ā€“

ā€“

ā€“

ā€“

100.0 (97.3ā€“100.0)

18.8 (16.1ā€“21.8)

8.7 (5.3ā€“12.7)

100.0 (93.5ā€“100.0)

ā€ƒā‰„5.0Ā %

69.0 (64.8ā€“73.0)

132

9

1

ā€“

4

4

93.3 (87.7ā€“96.7)

23.6 (20.2ā€“27.3)

36.2 (31.6ā€“40.9)

96.3 (93.0ā€“98.1)

ā€ƒā‰„7.5Ā %

53.4 (49.2ā€“57.6)

124

17

1

2

7

7

87.9 (81.0ā€“92.7)

28.7 (24.4ā€“33.3)

53.9 (49.2ā€“58.5)

95.5 (92.7ā€“97.3)

  1. AA advanced adenoma, CEDAR Cost-Effectiveness of a Decision rule for Abdominal complaints in primary caRe, CI confidence interval, CRC colorectal cancer, D diverticulitis, ELISA enzyme-linked immunosorbent assay, FIT faecal immunochemical test for haemoglobin, IBD inflammatory bowel disease, NPV negative predictive value, POC point-of-care, PPV positive predictive value, SCD significant colorectal disease
  2. aThe percentage referred and the accuracy measures are each averaged over the 10 imputed datasets. Hence, it is possible that, e.g. 100.0Ā % sensitivity does not directly match with 100.0Ā % NPV
  3. bA patient with SCD was considered missed if his/her predicted SCD probability was below the respective threshold for referral in at least 5 of the 10 imputed datasets