International Journal of Hematology

, Volume 96, Issue 2, pp 214–221

Cutoff values of serum ferritin and TIBC saturation for the evaluation of gastrointestinal neoplasms in adult anemic patients

Authors

    • Division of Hematology-Oncology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine Seoul
  • Eugene Park
    • Division of Hematology-Oncology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine Seoul
  • Jongchan Lee
    • Department of StatisticsKorea University
  • So Young Kim
    • Division of Gastroenterology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine
  • Sung Yong Kim
    • Division of Hematology-Oncology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine Seoul
  • Sun Young Lee
    • Division of Gastroenterology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine
  • Yo Han Cho
    • Division of Hematology-Oncology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine Seoul
  • So Young Yoon
    • Division of Hematology-Oncology, Department of Internal Medicine, Konkuk University Medical CenterKonkuk University School of Medicine Seoul
Original Article

DOI: 10.1007/s12185-012-1129-3

Cite this article as:
Lee, M.H., Park, E., Lee, J. et al. Int J Hematol (2012) 96: 214. doi:10.1007/s12185-012-1129-3

Abstract

We conducted a retrospective study to find out the optimum values of serum ferritin and other hematologic indices in adult anemic patients who should be referred for thorough gastrointestinal (GI) endoscopic evaluation for GI neoplasms. 544 adult anemic patients were stratified into three groups according to the results of GI endoscopy: benign versus premalignant versus malignant. As compared to non-malignant groups, malignant group demonstrated statistically significant differences in terms of median values of ferritin and total iron-binding capacity (TIBC) saturation. By receiver operating characteristics curve analyses to find out optimum cut-off points of the serum ferritin and TIBC saturation which distinguish between non-malignant diseases and malignant diseases, the cut-off ferritin value of 44.33 ng/mL in male had 72.73 % sensitivity and 70.95 % specificity. The cut-off TIBC saturation value of 9.13 % in male had 73.33 % sensitivity and 70.92 % specificity. The cut-off TIBC saturation value of 6.16 % in female had 69.57 % sensitivity and 65.13 % specificity. It is recommended that adult male patients with anemia undergo thorough endoscopic evaluation to detect GI neoplasms when their serum ferritin levels are ≤44 ng/mL or TIBC saturation values are ≤9 %. For adult female, only TIBC saturation values less than 6 % may contribute to determining whether they undergo GI endoscopic evaluation.

Keywords

FerritinsAnemiaIron deficiencyGastrointestinal neoplasmsEndoscopyGastrointestinal

Introduction

The World Health Organization (WHO) defines anemia as a hemoglobin (Hb) level less than 13 g/dL in men and less than 12 g/dL in women [1]. Iron deficiency anemia (IDA) is characterized by microcytic hypochromic red blood cells (RBC), a low-serum iron (Fe), an increased serum total iron-binding capacity (TIBC), a reduced transferrin saturation and a low-serum ferritin. Serum ferritin is the most powerful one among the laboratory tests used in the diagnosis of IDA [2]. If an anemic patient does not have an accompanying infectious or inflammatory disease, a cutoff ferritin level 41 ng/mL or less provides optimal efficiency for diagnosing iron deficiency [3]. However, serum ferritin level tends to increase far beyond 41 ng/mL when chronic GI blood loss is associated with neoplasms. In the developed countries, chronic blood loss is one of the common causes of IDA in adults and often associated with an occult gastrointestinal (GI) neoplasms, especially in adult males and postmenopausal females [4]. Gastrointestinal endoscopy was warranted in anemic patients with ferritin below 100 ng/mL even in the absence of GI symptoms or documented bleeding [5]. Therefore, it is difficult to set the cutoff ferritin level for GI endoscopic evaluation in patients with anemia which is presupposed to result from GI neoplasms. To answer this question, we conducted a retrospective study to find out the prevalence of non-malignant and malignant GI lesions and the most reliable hematological index associated with GI neoplasms in adult anemic patients who were referred to our Gastrointestinal Disease Center for the evaluation of GI blood loss. Finally, we desired to find out the optimum values of serum ferritin and other hematologic indices in adult anemic patients to be referred for thorough GI endoscopic evaluation.

Patients and methods

Patients and study design

We reviewed retrospectively patients’ medical records. The subject of study was adult patients (n = 544) with anemia at Konkuk University Medical Center who underwent upper and lower GI endoscopy to search for the causes of GI blood loss from August 2005 to August 2009. All patients were older than 20 years. They met the WHO anemia criteria by Hb value less than 13 g/dL in males and less than 12 g/dL in women. The study excluded pregnant women and patients with the obvious causes of blood loss other than GI blood loss, such as heavy menstrual bleeding, hematuria, epistaxis and hemoptysis. All patients underwent sequential endoscopic procedures of esophagogastroduodenoscopy and colonoscopy with parenteral administration of midazolam and pethidine, while vital signs were closely monitored. For the upper endoscopic evaluation, malignant lesions included gastric carcinoma and esophageal squamous cell carcinoma. Premalignant lesions included chronic atrophic gastritis and tubular adenoma. All other lesions were considered benign ones, including no abnormality. For the lower endoscopic evaluation, malignant lesions included colonic carcinoma, rectosigmoid colonic carcinoma and rectal carcinoma. Premalignant lesions included colonic tubular adenoma. All other lesions were considered benign lesions, including no abnormality. Patients were stratified into three groups according to the results of GI endoscopy: benign group versus premalignant group versus malignant group.

Statistical analyses

The primary end points of this study were to define the optimum values of serum ferritin and other hematological indices in adult anemic patients to be referred for thorough GI endoscopic evaluation. The secondary end points were to determine the prevalence of non-malignant and malignant GI lesions and the hematological indices associated with GI malignancies, in adult anemic patients who were referred to our Gastrointestinal Disease Center for the evaluation of GI blood loss. For the primary end point, receiver operating characteristics (ROC) curve was used to find out the optimum cut-off points of serum ferritin and other hematologic indices which distinguish between non-malignant patients and malignant patients. And the most reliable levels of serum ferritin and other hematologic indices were selected that satisfied appropriate sensitivity and specificity in each factors. For the second end point, we described prevalence of GI malignant diseases and 95 % confidence intervals. Normality test was also done through Shapiro–Wilk’s test for continuous variables. Because all the continuous variables are proved not to be normal distribution, we summarized continuous variables by median (range) and categorical variables by n (%). Kruskall–Wallis test or Wilcoxon’s sign rank test was used to test group differences for the continuous variables (Hb, MCV etc.) and Chi-square test for the categorical variables (gender, biopsy). To evaluate diagnostic accuracy that was measured by the area under the ROC curve (AUC), we referenced a guideline as follows. One could distinguish among non-informative (AUC = 0.5), less accurate (0.5 < AUC ≤ 0.7), moderately accurate (0.7 < AUC ≤ 0.9), highly accurate (0.9 < AUC < 1) and perfect tests (AUC = 1) [6]. We used Medcalc software to depict ROC curves for serum ferritin and other hematologic indices used in the diagnosis of GI malignancy [7] and SAS version 9.12 (SAS Institute Inc., Cary, NC, USA) was used for the other statistical analyses. P values less than 0.05 was considered significant.

Results

Prevalence of GI malignant and premalignant diseases

During the specified study period, 244 men and 300 women met our inclusion and exclusion criteria. Of a total of 544 patients, benign, premalignant and malignant GI lesions were detected in 265, 220 and 59 patients, respectively (Table 1). The prevalence of GI malignant diseases in our study groups was 10.85 % (59/544) for all patients, 13.93 % (34/244) for male patients and 8.33 % (25/300) for female patients (Table 1). The prevalence of GI premalignant diseases was 40.44 % (220/544) for all patients, 40.57 % (99/244) for male patients and 40.33 % (121/300) for female patients (Table 1). Median age of the malignant group was significantly older than that of the non-malignant group (benign plus premalignant groups) in both genders (Tables 1, 2, 3). In addition, upper GI malignant lesions and lower GI malignant lesions were detected in 26 (4.78 %, 26/544) and 31 (5.70 %, 31/544) patients, respectively, and 2 patients (0.37 %, 2/544) were diagnosed to have both upper and lower GI malignant lesions simultaneously (Table 4).
Table 1

Results of laboratory tests in the three groups for all 544 patients

Variable

Benign (n = 265)

Premalignant (n = 220)

Malignant (n = 59)

P value

P value

Age (year)

56 (20–92)

66 (23–91)

70 (34–92)

<0.001

<0.001

Gender

 Both

265 (48.71 %)

220 (40.44 %)

59 (10.85 %)

  

 Female

154 (51.33 %)

121 (40.33 %)

25 (8.33 %)

0.0891§

0.0367§

 Male

111 (45.49 %)

99 (40.57 %)

34 (13.93 %)

  

Hb (g/dL)

8.75 (2.4–15)

9.1 (3.8–12.9)

8.35 (4.3–12.9)

0.1187

0.1705

MCV (fL)

86.7 (57.1–114.9)

88.2 (57.4–122.6)

78.6 (51.3–104.4)

0.0002

0.0005

RDW (%)

14.9 (3.07–34.9)

14.6 (11.8–27.7)

14.6 (12.2–22.7)

0.9339

0.8047

Hct (%)

26.65 (1.9–43.9)

27.4 (11.4–39.3)

26.05 (15.3–37.4)

0.2692

0.3947

Ferritin (μg/L)

69.39 (0.6–4238)

108.5 (1.13–4908)

21.7 (2.07–1285)

<0.001

0.0002

Fe (μg/dL)

35 (2–300)

38 (2–300)

13.5 (4–271)

<0.001

<0.001

TIBC (μg/dL)

279 (113–608)

259 (90–564)

320 (140–498)

0.0008

0.0024

Tsat (%)

13.07 (0.5–98.7)

14.78 (0.76–97.44)

4.64 (1.23–95.76)

<0.001

<0.001

Biopsy

 Done

152 (57.36 %)

161 (73.18 %)

59 (100 %)

<0.001§

<0.001§

 Not done

113 (42.64 %)

59 (26.82 %)

0 (0 %)

  

Gender was expressed as n (row %), biopsy as n (column %) and continuous variables were written as median (range)

MCV mean corpuscular volume, RDW red cell distribution width, Hct hematocrit, TIBC total iron-binding capacity, Tsat TIBC saturation

P values from Kruskall–Wallis test among the three groups (benign, premalignant and malignant)

P values from Wilcoxon’s sign rank test between the two groups (non-malignant, malignant)

§P values from Chi-square test

Table 2

Results of laboratory tests in the three groups for 244 male patients

Variable

Benign (n = 111)

Premalignant (n = 99)

Malignant (n = 34)

P value

P value

Age (years)

60 (20–65)

65 (30–91)

69 (34–91)

0.0028

0.016

Hb (g/dL)

9.45 (2.4–15.0)

9.0 (3.90–12.90)

8.20 (5.0–12.90)

0.3426

0.1715

MCV (fL)

89.15 (61.4–114.9)

89.85 (57.4–122.6)

77.9 (52.2–98.8)

0.0004

0.0001

RDW (%)

14.3 (11.6–34.9)

14.6 (12.1–27.7)

14.55 (12.2–22.7)

0.9055

0.9117

Hct (%)

27.6 (9.4–43.9)

26.6 (11.4–39.3)

25.0 (15.3–37.4)

0.7726

0.4816

Ferritin (μg/L)

98.80 (1.74–4238)

139.1 (1.96–4908)

17.51 (2.07–1285)

0.0004

0.0001

Fe (μg/dL)

49 (3–300)

42 (3–259)

14 (4–271)

0.0002

<0.001

TIBC (μg/dL)

260 (113–495)

253.50 (90–547)

324.5 (157–498)

0.0027

0.0006

Tsat (%)

20.15 (0.9–98.7)

17.48 (0.76–97.44)

4.63 (1.23–95.76)

<0.001

<0.001

Biopsy

 Done

70 (63.06 %)

78 (78.79 %)

34 (100 %)

<0.001§

0.0002§

 Not done

41 (36.94 %)

21 (21.21 %)

0 (0 %)

  

Categorical variable (biopsy) was expressed as n (%) and continuous variables were written as median (range)

MCV mean corpuscular volume, RDW red cell distribution width, Hct hematocrit, TIBC total iron-binding capacity, Tsat TIBC saturation

P values from Kruskall–Wallis test among the three groups (benign, premalignant and malignant)

P values from Wilcoxon’s sign rank test between the two groups (non-malignant, malignant)

§P values from Chi-square test

Table 3

Results of laboratory tests in the three groups for 300 female patients

Variable

Benign (n = 154)

Premalignant (n = 121)

Malignant (n = 25)

P value

P value

Age (yr)

53 (20–92)

68 (23–90)

73 (57–92)

<0.001

<0.001

Hb (g/dL)

8.60 (3.20–12.0)

9.20 (3.80–12.0)

8.40 (4.30–11.50)

0.0362

0.5621

MCV (fL)

84.1 (57.1–114.9)

86.8 (59.2–117.7)

81.3 (51.3–104.4)

0.0536

0.1737

RDW (%)

15.3 (3.07–24.9)

14.5 (11.8–26.3)

14.7 (12.4–21.5)

0.7001

0.7637

Hct (%)

26.45 (1.9–36.6)

28.1 (12.6–36.8)

27.4 (16.4–34.7)

0.0979

0.7025

Ferritin (μg/L)

35.12 (0.6–1962)

94.18 (1.13–4011)

25.29 (2.15–886.6)

0.0006

0.0707

Fe (μg/dL)

31 (2–299)

31 (2–300)

15 (5–203)

0.0339

0.0125

TIBC (μg/dL)

300 (143–608)

264 (96–564)

314 (140–467)

0.0175

0.3073

Tsat (%)

9.86 (0.5–92.4)

12.98 (0.89–83.72)

4.69 (1.24–63.44)

0.0400

0.0229

Biopsy

 Done

82 (53.25 %)

83 (68.6 %)

25 (100 %)

<0.001§

<0.001§

 Not done

72 (46.75 %)

38 (31.4 %)

0 (0 %)

  

Categorical variable (biopsy) was expressed as n ( %) and continuous variables were written as median (range)

MCV median values of mean corpuscular volume, RDW red cell distribution width, Hct hematocrit, TIBC total iron-binding capacity, Tsat TIBC saturation

P values from Kruskall–Wallis test among the three groups (benign, premalignant and malignant)

P values from Wilcoxon’s sign rank test between the two groups (non-malignant, malignant)

§P values from Chi-square test

Table 4

Results of laboratory tests between lower and upper GI neoplasm

Variable

Lower GI neoplasms (n = 31)

Upper GI neoplasms (n = 26)

P value

Age (year)

72 (46–92)

68.5 (34–91)

0.0681

Gender

 Female

15 (48.39 %)

9 (34.62 %)

0.2942

 Male

16 (51.61 %)

17 (65.38 %)

 

Hb (g/dL)

9.3 (4.3–12.9)

7.4 (5–11.5)

0.0778

MCV (fL)

78.2 (51.3–98.8)

79.3 (52.2–104.4)

0.5719

RDW (%)

14.6 (12.2–21.5)

14.6 (12.4–22.7)

0.9915

Hct (%)

29.4 (16.4–37.4)

24.4 (15.3–33.8)

0.0248

Ferritin (μg/L)

16.015 (2.15–1285)

29.135 (2.64–886.6)

0.4982

Fe (μg/dL)

18 (5–271)

13 (4–203)

0.3655

TIBC (μg/dL)

348 (168–498)

314.5 (140–476)

0.5902

Tsat (%)

4.86 (1.240–95.759)

4.361 (1.230–63.437)

0.5645

Gender was expressed as n (row %) and continuous variables were written as median (range)

MCV mean corpuscular volume, RDW red cell distribution width, Hct hematocrit, TIBC total iron-binding capacity, Tsat TIBC saturation

P values from Wilcoxon’s sign rank test between the two groups (lower and upper GI neoplasm)

P values from Chi-square test

Hematologic indices associated with GI neoplasms

For 544 whole patients, there were statistically significant differences in terms of median values of mean corpuscular volume (MCV), ferritin, Fe, TIBC and TIBC saturation between the non-malignant group (benign plus premalignant groups) and the malignant group (Table 1). However, Hb, hematocrit (Hct) and red cell distribution width (RDW) did not show significant differences between the two groups. For 244 male patients, the median values of MCV, ferritin, Fe, TIBC and TIBC saturation were significantly different between the non-malignant group and the malignant group (Table 2). These results of the male patients group were identical to those of the whole patients group. For 300 female patients, as compared to the non-malignant group, malignant group demonstrated statistically significant differences in the median values of Fe and TIBC saturation (Table 3). In contrast to the whole patients and the male patients groups, the female patients group failed to show significant differences in MCV, ferritin and TIBC; however, ferritin showed a strong trend of decrease in the malignant group (Table 3). In addition, there were no statistically significant differences in hematologic indices except hematocrit value between the upper GI malignant group and the lower GI malignant group (Table 4).

These results demonstrated that serum ferritin and TIBC saturation were worth being considered meaningful hematologic indices to distinguish the malignant group from the non-malignant group, especially in male patient group.

Cutoff values of serum ferritin for the diagnosis of GI neoplasms

For the whole patients, AUC was 0.65 and P value was 0.0002 that implies less accuracy (Table 5; Fig. 1). At the cutoff ferritin value of 74.5 ng/mL, the sensitivity was 77.59 % and the specificity was 53.72 % for the diagnosis of GI malignant diseases. For 244 male patients, AUC and P value were 0.705 and 0.0001, respectively, which implies moderate accuracy (Table 5; Fig. 1). At the cutoff ferritin value of 44.33 ng/mL, the sensitivity and specificity were 72.73 and 70.95 %, respectively. For 300 female patients, AUC was 0.609 and P value greater than 0.05 that implies less accuracy (Table 5; Fig. 1). And the cutoff ferritin value of 30.4 ng/mL provided 60 % of sensitivity and 64.23 % of specificity.
Table 5

ROC curve results of serum ferritin as a predictor for GI neoplasms

 

All patients (n = 544)

Male patients (n = 244)

Female patients (n = 300)

AUC

0.65

0.705

0.609

95 % CI of AUC

(0.608–0.690)

(0.644-0.762)

(0.551–0.665)

P value

0.0002

0.0001

0.0787

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

112.3

81.03

42.36

228

84.85

35.71

116.8

80

35.77

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

110

79.31

42.36 

72.74

78.79

61.9

68.51

72

50.36

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

95.68

79.31

46.69

66.5

75.76

63.33

63.21

64

52.19

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

74.5

77.59

53.72

44.33

72.73

70.95

30.4

60

64.23

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

72.15

75.86

54.13

40.69

63.64

70.95

26.67

52

65.33

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

66.75

72.41

55.79

31.6

57.58

74.76

25.29

48

65.69

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

63.94

70.69

56.61

23.29

51.52

77.14

20.1

44

68.69

Ferritin cut-off point (ng/mL)

Sensitivity (%)

Specificity (%)

45.49

67.24

61.98

15.71

45.45

84.29

16.32

36

69.71

ROC receiver operating characteristics, AUC area under ROC curve, CI confidence interval

https://static-content.springer.com/image/art%3A10.1007%2Fs12185-012-1129-3/MediaObjects/12185_2012_1129_Fig1_HTML.gif
Fig. 1

ROC curves for serum ferritin in the diagnosis of GI neoplasms AUC suggests that serum ferritin can be a useful tool in detecting GI neoplasms in male patients (male patients group, AUC = 0.705, P value = 0.0001)

Based on the above statistical analyses, serum ferritin 44 ng/mL or less was strongly suggestive of GI malignant diseases in male patients group.

Cutoff values of TIBC saturation for the diagnosis of GI neoplasms

For the whole patients, AUC was 0.691 and P value was 0.0001 that implies less accuracy (Table 6; Fig. 2). At the cutoff TIBC saturation value of 9.75 %, the sensitivity was 73.58 % and the specificity was 60.38 % for the diagnosis of GI malignant diseases. For 244 male patients, AUC and P value were 0.750 and 0.0001, respectively, which implies moderate accuracy (Table 6; Fig. 2). At the cutoff TIBC saturation value of 9.13 %, the sensitivity and specificity were 73.33 and 70.92 %, respectively. For 300 female patients, AUC and P value were 0.643 and 0.0262, respectively, which implies that ROC curve is less accurate but still useful statistically (Table 6; Fig. 2). And the cutoff TIBC saturation value of 6.16 % provided 69.57 % of sensitivity and 65.13 % of specificity.
Table 6

ROC curve results of serum TIBC saturation as a predictor for GI neoplasms

 

All patients (n = 544)

Male patients (n = 244)

Female patients (n = 300)

AUC

0.691

0.750

0.643

95 % CI of AUC

(0.649–0.731)

(0.688–0.805)

(0.584–0.699)

P value

0.0001

0.0001

0.0262

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

30.24

90.57

22.32

21.24

83.33

45.92

18.59

86.96

33.33

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

19.46

84.91

39.61

14.14

76.67

61.22

14.33

78.26

43.68

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

13.20

77.36

53.17

9.13

73.33

70.92

13.09

73.91

45.98

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

9.75

73.58

60.38

7.64

66.33

82.14

6.16

69.57

65.13

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

6.16

66.04

71.12

5.17

63.38

82.14

5.74

65.22

66.67

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

5.05

60.38

74.40

4.9

56.67

82.65

5.35

56.52

67.82

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

4.68

50.94

75.93

4.62

46.67

84.18

4.57

47.83

70.50

TIBC saturation cut-off point (%)

Sensitivity (%)

Specificity (%)

3.49

41.51

80.31

2.97

30.00

80.89

4.06

34.78

73.18

ROC receiver operating characteristics, AUC area under ROC curve, CI confidence interval

https://static-content.springer.com/image/art%3A10.1007%2Fs12185-012-1129-3/MediaObjects/12185_2012_1129_Fig2_HTML.gif
Fig. 2

ROC curves for TIBC saturation in the diagnosis of GI neoplasms AUC suggests that serum TIBC saturation can be a useful tool in detecting GI neoplasms in male patients (male patients group, AUC = 0.750, P value = 0.0001)

Based on the above statistical analyses, serum TIBC saturation ≤9 % in male patients group and ≤6 % in female patients group were suggestive of malignant diseases.

Discussion

The most common type of anemia among outpatients is known to be IDA, which results often from occult GI blood loss. Patients with IDA are recommended to undergo bidirectional endoscopy unless a definite etiology is identified. Unfortunately most of patients with GI neoplasms are asymptomatic until their neoplasms advance. Given the fact that prevalence of GI malignant diseases increases as the elderly population rises, the importance of GI endoscopic evaluation must be emphasized in the elderly anemic patients.

The previous studies reported that the prevalence of GI neoplasms in IDA patients was 8–13 % [811], except one study which revealed 51 % of prevalence of GI neoplasms [12]. 10.8 % of prevalence of GI neoplasms in our study was equivalent to those of the previous studies. In our study, the prevalence of GI neoplasms was higher in male patients than in female patients (13.9 vs. 8.3 %, χ2 = 4.366, P = 0.037).

Normal range of serum ferritin in males is between 29 and 248 ng/mL with an average of 100 ng/mL, while its normal range in females is between 10 and 150 ng/mL with an average of 30 ng/mL [13]. The normal value for serum ferritin varies according to age and gender of the individual [14]. Zanella et al. [15] reported that serum ferritin showed 82 % sensitivity with 95 % specificity for the diagnosis of iron deficiency; moreover, the sensitivity increased more in men and iron deficiency anemia. A diagnostic cutoff value of ferritin ≤30 ng/mL provides 92 % sensitivity and 98 % specificity for the diagnosis of IDA [16].

In some comorbidity conditions, serum ferritin is not reliable in the diagnosis of IDA because it can be raised by the various clinical conditions, such as infection [17], rheumatoid arthritis [18, 19], liver disease [2022], malignancy [23], and other inflammatory conditions. Ferritin is an acute phase reactant and inflammatory cytokines such as interleukin-1-beta or tumor necrosis factor-alpha to enhance ferritin secretion from hepatic cells [24]. Therefore, it has been proposed that serum ferritin level ≤100 μg/L was optimal for determining iron deficiency rather than its level of ≤50 μg/L cited in the textbooks in patients with inflammation, infection or malignancy [25]. Although the laboratory tests for serum transferrin receptor or the ratio of serum transferrin receptor to serum ferritin have been proposed to diagnose iron deficiency more accurately in the presence of comorbid illnesses, they also have certain limitations [26, 27]. In addition, serum ferritin level shows an age-related tendency to rise and it is higher in males than in females [14]. Kobune et al. [28] suggested that the increment of serum iron and % transferrin saturation at 120 min after sodium ferrous citrate administration was useful in distinguishing iron absorption between healthy volunteers, patients with IDA, and patients with anemia of chronic disorders.

Our study demonstrated that serum ferritin and TIBC saturation values were significantly lower in male anemic patients with GI neoplasms than those without GI neoplasms (Table 2, P values 0.0001 and <0.001, respectively), and it showed a strong trend of decrease in female anemic patients with GI neoplasms as compared with those without GI neoplasms (Table 3, P values 0.0707 and 0.0229, respectively).

We used ROC analysis for the selection of cut-off points of serum ferritin and TIBC saturation in adult anemic patients who required thorough GI endoscopic evaluation to search for GI neoplasms [7]. For male patients, AUC provided by ferritin in the identification of GI neoplasms was 0.705 and P value was <0.05, which indicates serum ferritin measurement to be a moderately accurate test. However, AUC in female patients was 0.609 and P value was 0.0787, indicating serum ferritin measurement to be a less accurate test. Therefore, we propose that serum ferritin measurement for the diagnosis of GI neoplasms be reliable for male anemic patients only.

In addition, AUC provided by TIBC saturation in the identification of GI neoplasms was 0.750 with P value <0.05 in male patients that indicates serum TIBC saturation measurement to be as reliable as serum ferritin measurement for a diagnostic tool of GI neoplasms in anemic male patients. And AUC in female patients was 0.643 with P value <0.05. These results imply that TIBC saturation can be used as a screening test to refer adult anemic patients for thorough GI endoscopic evaluation.

Through a discriminant analysis, a function of ferritin and TIBC saturation was considered a diagnosis of GI neoplasms. However, we did not describe the results because there were no significant results by a discriminant analysis.

In this study, the cutoff ferritin value of 44.33 μg/L provided 72.73 % sensitivity and 72.95 % specificity for the diagnosis of GI neoplasms in male anemic patients. The cutoff TIBC saturation value of 9.13 % provided 73.33 % sensitivity and 70.92 % specificity in male anemic patients, and that of 6.16 % provided 69.57 % sensitivity and 65.13 % specificity in female anemic patients for the diagnosis of GI neoplasms.

We demonstrate individual sensitivity and specificity on the ROC curves according to the various levels of serum ferritin and TIBC saturation for the purpose of providing a guideline for the future studies (Tables 5, 6).

In conclusion, this study proposes that male anemic patients with serum ferritin ≤44 μg/L or serum TIBC saturation ≤9 % be recommended to undergo thorough endoscopic evaluation to search for GI neoplasms. For female anemic patients, serum TIBC saturation ≤6 % is a reliable value for the endoscopic evaluation for GI neoplasms.

Conflict of interest

None.

Copyright information

© The Japanese Society of Hematology 2012