European Journal of Internal Medicine 26 (2015) 338–343

Contents lists available at ScienceDirect

European Journal of Internal Medicine journal homepage: www.elsevier.com/locate/ejim

Original Article

Predicting the emergence of anemia — A large cohort study Anat Gafter-Gvili a,b,d,⁎, Eytan Cohen c,d, Tomer Avni a,d, Alon Grossman a,d, Liat Vidal b,d, Moshe Garty c,d, Leonard Leibovici a,d, Ilan Krause b,d a

Department of Medicine E, Rabin Medical Center, Beilinson Hospital, Tel-Aviv University, Israel Institute of Hematology, Rabin Medical Center, Beilinson Hospital, Tel-Aviv University, Israel Department of Medicine F-Recanati, Rabin Medical Center, Beilinson Hospital, Tel-Aviv University, Israel d Sacker Faculty of Medicine, Tel-Aviv University, Israel b c

a r t i c l e

i n f o

Article history: Received 22 February 2015 Received in revised form 7 April 2015 Accepted 9 April 2015 Available online 23 April 2015 Keywords: Anemia Predictors Diabetes Triglycerides

a b s t r a c t Background and objectives: We aimed to find predictors for development of anemia in a large cohort of adults. Patients and methods: Cohort study of a large health database from a screening center at the Rabin Medical Center in Israel, between the years 2000–2013. We asked which variables, known at the first visit, would predict anemia at the last visit. Multivariable analysis was conducted using stepwise logistic regression analysis. Odds ratios (ORs) for anemia with 95% confidence intervals (CIs) were calculated. Results: Our cohort included 10,577 people. At baseline 4.4% were diagnosed with anemia and excluded. Therefore, 10,093 subjects, with a mean age of 42.3 ± 9 years comprised our study sample. At the end of follow-up of 4.7 ± 3.1 years, 307 developed anemia (3%). In men, independent predictors for development of anemia were diabetes mellitus (OR 3.00, 95% CI 1.41–6.39), age (OR 1.03, 95% CI 1.03–1.05, for 1 year increment), low MCV (OR 0.92, 95% CI 0.89–0.96, for every 1 fL unit increment) and elevated platelet count (OR 1.004, 95% CI 1.00– 1.01 for 1000/μL unit increment). For women, high total serum protein level was a strong predictor for anemia (OR 3.44, 95% CI 2.33–5.08 for 1 mg/dL increment) as well as low triglycerides (OR 0.996, 95% CI 0.993–1.000 for 1 mg/dL increment). Conclusions: Subgroups who are prone to develop anemia include men with diabetes, and women with an elevated serum protein level and low triglycerides. © 2015 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved.

1. Background Anemia is a common disorder considered as a global public health problem. According to global and regional data from the WHO Vitamin and Mineral Nutrition Information System for 1993–2005 [1], it may affect up to 25% of the world's population, especially in preschool-aged children and women. More recent global data from 187 countries, collected from 1990 to 2010, shows a higher prevalence of 32.9% in 2010 [2]. The prevalence is estimated at 9% in developed countries, but may reach 43% in low income countries [1]. A recent cohort study of about 30,000 Swedish participants aged 44–73 showed a prevalence of 3.8% [3], when anemia was defined according to the WHO criteria of hemoglobin less than 13 g/dL for men and less than 12 g/dL for women [4]. Anemia is a major determinant of morbidity. It may be associated with impaired oxygen delivery, decreased exercise tolerance and

⁎ Corresponding author at: Department of Medicine E and Institute of Hematology, Rabin Medical Center, Beilinson Hospital, Petah-Tikva 49100, Israel. Tel.: + 972 3 9376500/1; fax: +972 3 9376512. E-mail addresses: [email protected], [email protected] (A. Gafter-Gvili).

reduced quality of life [5]. It has also been shown to adversely affect cognitive function and cause fatigue and reduced work capacity [6,7]. Anemia has been shown in a recent population based cohort study to be an independent risk factor for all-cause mortality in the general population, cardiovascular and cancer-related mortality [3]. It is also associated with increased mortality in settings such as chronic kidney disease [8], heart failure [9,10], acute coronary syndrome [11], peri-operative [12, 13] and malignancy [14]. Predictors for anemia may help define subgroups prone to development of anemia, which may warrant a more intensive follow-up. We therefore aimed to define predictors for development of anemia in a large cohort of adults attending a screening program. 2. Methods 2.1. Study population We analyzed a large health database from a screening center at the Rabin Medical Center in Israel. This referral institute provides regular health assessments for employees of different companies. 24,000 people were assessed between the years 2000 and 2013. The population attending the screening center includes men and non-pregnant women,

http://dx.doi.org/10.1016/j.ejim.2015.04.010 0953-6205/© 2015 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved.

A. Gafter-Gvili et al. / European Journal of Internal Medicine 26 (2015) 338–343

with an age range between 20 and 80 years. Each person underwent a thorough medical history evaluation and a complete physical examination together with laboratory analysis, chest X-ray, electrocardiogram, exercise stress test and lung function test. Subjects had the opportunity to return once a year for a repeat investigation. We assessed the prevalence of anemia at study entry (visit 1). For the main analysis we included only those who did not have anemia at the first visit. Anemia was defined as hemoglobin level less than 12 g/dL in women and less than 13 g/dL in men, in accordance with World Health Organization (WHO) criteria [4]. We included only people who had at least two visits, in which hemoglobin level was taken. If there were more than two visits, then the visit at the latest follow-up was assessed. The study was approved by the Research Ethics Committee of the Rabin Medical Center.

2.2. Data collection The following data were collected: presence of hypertension, diabetes mellitus, impaired fasting glucose, hypertriglyceridemia, hyperuricemia, gout, current and past smoking, alcohol drinking, number of times of physical activity per week (all of which were self-reported). The following were measured at each visit: height, weight, body mass index (BMI), weight circumference, and blood pressure. A complete blood count including hemoglobin level, mean corpuscular volume (MCV), red blood cell distribution width (RDW), leukocyte count and differential, platelet count, full blood chemistry, lipid profile, erythrocyte sedimentation rate (ESR), high sensitive c-reactive protein (hsCRP), iron, transferrin, transferrin saturation, vitamin B12, folic acid, thyroid stimulating hormone (TSH), T3, T4, homocysteine and PSA were obtained.

2.3. Outcome We assessed anemia at the last follow-up as the outcome. Anemia was defined according to the WHO criteria.

2.4. Statistical analysis We looked for variables known at the first visit that would predict anemia at the last visit. The causes for anemia are probably different for adult men and women, and thus we have conducted the analyses separately for women and men. Dichotomous variables were examined by the Chi-square test. Continuous variables were described as means with standard deviation (SD), and compared using a T-test because of the large sample size [15]. We used the Breslow–Day test of homogeneity of the odds ratio to assess if there are significant gender differences. Independent risk factors for anemia at the end of follow-up were identified using a forward stepwise (by likelihood ratio) logistic regression analysis, including all variables associated with anemia on univariate analysis (p b 0.05) after removing those that were highly correlated (Spearman's correlation coefficient N 0.5). Missing values for laboratory values included in the model were imputed using linear regression multiple imputation analysis. However, no value was missing in more than 2% in women or 1.7% in men. Model fit was assessed using the Hosmer and Lemeshow Test and the predictive ability of the model was evaluated using the area under the curve (AUC) plotted on a receiver operating characteristic curve (ROC curve). Odds ratios (ORs) for anemia with 95% confidence intervals (CIs) are reported. We divided the probability calculated by the logistic model for anemia into ten percentiles, to assess the actual frequency of developing anemia. Statistical analysis was performed using or IBM SPSS Statistic 21 (IBM Corp, New York, NY).

339

3. Results During the years 2000–2013, 24,000 people were evaluated at the health center. Of these, 10,557 had at least two separate visits in which hemoglobin level was assessed. At first study entry, 464 (4.4%) were identified with anemia. Among people with anemia at baseline, 21% were men and 79% women. The baseline prevalence of anemia in men was 1.8% (134/7605 of the whole male population in the cohort) and in women 11.2% (330/2952 of the female population in the cohort). This group with anemia at baseline was excluded from further analysis. Therefore, 10,093 subjects without anemia at baseline comprised the study sample, and all analyses apply to this group. Of them, 7471 were men (74%) and 2622 were women (26%). At the end of followup 307 developed anemia (3%), 1.6% of men (118/7471) and 7.2% of women (189/2622) developed anemia. The baseline demographic characteristics of the study population according to people who developed anemia compared to those who did not develop anemia at the latest follow-up are shown in Table 1, the baseline hematological parameters in Table 2 and the baseline laboratory evaluation in Table 3. The mean age of the population was 42.3 ± 9.0 years. The mean baseline age was higher for men who developed anemia, 43.9 ± 10.1 years compared to men who did not develop anemia, 42.1 ± 8.9 years (p = 0.038) and was lower for women who developed anemia, 40.0 ± 7.7 years, compared to women who did not develop anemia, 42.9 ± 9.1 years, p b 0.001. The mean follow-up duration was 4.7 ± 3.1 years. It was longer for men who developed anemia (5.4 years for anemics compared with 4.8 years for non-anemics, p = 0.049) (Table 1). When assessing the occurrence of anemia in the study population according to four quartiles of age groups, the percentage of men with anemia increased with age. For men, the incidence of anemia was 1.2% in the first quartile (age b 35), 1.6% in the second and third quartiles (ages 35–41 and 41–48), and 2% in the fourth quartile (age N 48). For women, the percentage of anemia was 8.6% in the first quartile, 8.9% in the second quartile, and then decreased to 6.2% in the third quartile and to 3.9% in the fourth quartile (Table 1). Hematological parameters found to be associated with development of anemia on univariate analysis, for both men and women, included a low hemoglobin level at baseline (men: 14.0 g/dL in those who developed anemia vs. 15.1 in those who did not, p b 0.001, women: 12.7 vs. 13.3 p b 0.001), as well as a low MCV level (men: 83.1 vs. 85.3, p b 0.001, women: 85.7 vs. 86.6, p = 0.006) and low transferrin saturation (men: 37.3% vs. 40%, p = 0.001, women: 30.6% vs. 34.2%, p = 0.001). For men, an elevated RDW (13.3% vs. 13.0%, p = 0.003), an elevated platelet count (250.5 K/μL vs. 235.3 K/μL, p = 0.018), and a low iron level (95.72 μg/dL vs. 103.7 μg/dL, p = 0.011) were significantly associated with development of anemia as well. In women, an elevated transferrin level (282.9 mg/dL vs. 274.7 mg/dL, p = 0.01) was associated with anemia (Table 2). Metabolic parameters found to be associated with development of anemia in men included the presence of diabetes mellitus (8.5% of the male population with anemia were diabetic compared to 2.3% without anemia, p b 0.001) as well as a low mean triglycerides (120.4 mg/dL vs. 137.8 mg/dL, p = 0.033) and an elevated ESR (17.49 mm/h vs. 12.06 mm/h, p b 0.001). In females, other factors found to be significantly associated with development of anemia on univariate analysis included a low mean triglyceride level (90.72 mg/dL vs. 103.68 mg/dL), low mean cholesterol level (188.8 mg/dL vs. 197.6 mg/dL, p = 0.002), an elevated total protein level (7.5 g/dL vs. 7.4 g/dL, p b 0.001), an elevated globulin level (3.15 g/dL vs. 3.03 g/dL, p b 0.001) and a low BMI (24.2 kg/m2 vs. 25.1 kg/m2, p = 0.015) (Table 3). Results of the multivariable analysis are shown in Table 4. For men, the strongest predictor for development of anemia was the presence of diabetes mellitus (OR 3.00, 95% CI 1.41–6.39). Other factors included older age (OR 1.03, 95% CI 1.00–1.05, for 1 year increment), low MCV (OR 0.92, 95% CI 0.89–0.96, for every 1 fL unit increment) and elevated platelet count (OR 1.004, 95% CI 1.00–1.01 for 1000/μL unit increment).

340

A. Gafter-Gvili et al. / European Journal of Internal Medicine 26 (2015) 338–343

Table 1 Baseline demographic characteristics of the study population, according to appearance of anemia at the end of follow-up. Characteristic at baseline

Men

Age (years), mean (SD) Age b 35 years Age 35–41 Age 41–48 Age N 48 Hypertension Diabetes mellitus Impaired fasting glucose Hypertriglyceridemia Hyperuricemia Gout Current smoking Number of current smoking pack years, mean (SD) Past smoking Number of past smoking pack years, mean (SD) Alcohol drinking Number of times of physical activity per week Weight (kg) Height (cm) Waist circumference (cm) BMI (kg/m2) Follow-up duration in years

Women

Men with anemia N = 118

Men without anemia N = 7353

43.9 (10.1) 24 (1.2) 31 (1.6) 31 (1.6) 32 (2) 8 (6.8) 10 (8.5) 7 (5.9) 7 (5.9) 5 (4.2) 0 (0) 16 (13.6) 1.53 (6.3) 40 (33.9) 3.1 (11.1) 24 (20.3) 1.98 (1.8) 80.9 (14.5) 175.4 (6.4) 89.3 (8.6) 26.3 (4.4) 5.4 (3.1)

42.13 (8.9) 1918 (98.8) 1950 (98.4) 1950 (98.4) 1573 (98) 455 (6.2) 169 (2.3) 477 (6.5) 839 (11.4) 638 (8.7) 41 (0.6) 1084 (14.7) 1.91 (6.9) 2097 (28.5) 2.7 (7.9) 1585 (21.6) 2.03 (2.0) 83.3 (13.1) 175.9 (6.7) 91.9 (10.8) 26.9 (3.8) 4.8 (3.2)

p-Value 0.038

0.791 b0.001a 0.808 0.062 0.088 0.416 0.719 0.549 0.200 0.520 0.750 0.787 0.083 0.424 0.160 0.089 0.049

Women with anemia N = 189

Women without anemia N = 2433

p-Value

40.0 (7.7) 53 (8.6) 70 (8.9) 44 (6.7) 22 (3.9) 6 (3.2) 3 (1.6) 5 (2.6) 4 (2.1) 1 (0.5) 0 (0) 23 (12.2) 0.88 (3.8) 34 (18) 1.3 (5.48) 22 (11.6) 1.65 (1.8) 65.3 (12.7) 164.0 (6.6) 78.3 (10.8) 24.2 (4.2) 4.7 (3.1)

42.9 (9.1) 563 (91.4) 713 (91.1) 616 (93.3) 541 (96.1) 63 (2.6) 35 (1.4) 92 (3.8) 99 (4.1) 37 (1.5) 10 (0.4) 401 (16.5) 1.85 (6.9) 509 (20.9) 1.4 (5.3) 344 (14.1) 1.89 (2.0) 66.9 (12.9) 163.2 (6.2) 79.6 (11.0) 25.10 (4.7) 4.5 (3.0)

b0.001

0.628 0.890 0.425 0.183 0.272 0.256 0.121 0.058 0.338 0.829 0.340 0.103 0.111 0.112 0.342 0.015 0.406

For all dichotomous variables, n patients (%), with all patients assessed, unless otherwise indicated. For all continuous variables, mean (standard deviation, SD). a Breslow–Day Test of Homogeneity of the Odds Ratio between men and women significant, p b 0.05.

For women, high total serum protein level was a strong predictor for development of anemia (OR 3.44, 95% CI 2.33–5.08 for every increment of 1 mg/dL) as well as low triglycerides (OR 0.996, 95% CI 0.99–1.00, for every increment of 1 mg/dL). For both genders, a low hemoglobin at baseline was a strong predictor for anemia (men: OR 0.18, 95% CI 0.135–0.237; women: OR 0.19, 95% CI 0.15–0.26, for every increment of 1 g/dL), as well as a longer follow-up duration (men: OR 1.11, 95% CI 1.04–1.17; women: OR 1.06, 95% CI 1.01–1.12), for every increment of one year. We divided the probability calculated by the logistic model for anemia into ten percentiles, to assess the actual frequency of developing anemia. Fig. 1 shows the actual percentages of anemia in the percentiles of probability calculated by the logistic model. For women, the frequency of anemia increased steadily from 0.8% in the first percentile of probability of fulfilling the model, to 25.6% in the tenth percentile. For men, the frequency of anemia was 0% in the first percentile, ranged from

0.4% to 0.9% in the following seven percentiles, and reached 9.5% in the tenth percentile of probability (Fig. 1). The area under the ROC curve for the model showed good prediction (men: AUC 0.85, 95% CI 0.813–0.889, p = 0.000; women: AUC 0.792, 95% CI 0.761–0.823, p = 0.000).

4. Discussion In this very large cohort of a population attending a screening program (with an average age of less than 45 years), we found a prevalence of anemia of 4.4%, of whom 11.2% were women and 1.8% were men. In a follow-up of 4.7 years, 3% of the cohort which were not anemic on the first visit developed anemia, 1.6% of men and 7.2% of women. For men, diabetes mellitus, older age, low MCV and elevated platelet count were found to be predictors of anemia. For women, high total serum

Table 2 Baseline hematologic parameters of the study population, according to appearance of anemia at the end of follow-up. Characteristic at baseline

Hemoglobin (g/dL) MCV (fL) RDW (%) WBC (K/μL) Neutrophils (K/μL) Lymphocytes (K/μL) Monocytes (K/μL) Eosinophils (K/μL) Platelets (K/μL) Iron (μg/dL) Transferrin (mg/dL) Transferrin saturation (%) B12 (ng/L) Folic acid (ng/mL)

Men

Women

Men with anemia N = 118

Men without anemia N = 7353

p-Value

Women with anemia N = 189

Women without anemia N = 2433

p-Value

14.0 (0.8) 83.1 (6.1) 13.3 (0.9) 6.8 (1.6) 4.1 (1.4) 1.90 (0.5) 0.35 (0.1) 0.22 (0.2) 250.5 (68.6) 95.7 (33.7) 261.5 (33.9) 37.3 (14.1) 309.7 (134.7) 22.1 (8.1)

15.1 (0.87) 85.3 (3.9) 13.0 (0.6) 6.9 (1.6) 4.1 (1.3) 1.99 (0.5) 0.37 (0.4) 0.20 (0.1) 235.3 (53.4) 103.7 (33.3) 259.2 (31.9) 40 (14.3) 270.8 (122.5) 19.4 (9.4)

b0.001 b0.001 0.003 0.510 0.997 0.073 0.09 0.218 0.018 0.011 0.452 0.001 0.014 0.030

12.7 (0.56) 85.7 (4.6) 13.1 (0.7) 6.7 (1.7) 4.1 (1.3) 1.96 (0.6) 0.30 (0.1) 0.17 (0.1) 263.6 (60.9) 88.4 (35.2) 282.9 (46.2) 30.6 (13.6) 283.3 (126.8) 21.4 (8.1)

13.3 (0.74) 86.6 (4.1) 13.1 (0.7) 6.9 (1.7) 4.3 (1.4) 1.99 (0.5) 0.31 (0.1) 0.18 (0.1) 260.1 (59.1) 92.8 (35.1) 274.7 (41.3) 34.2 (14.2) 301.1 (147.3) 23.7 (17.6)

b0.001 0.006 0.592 0.178 0.314 0.373 0.126 0.358 0.421 0.098 0.01 0.001 0.252 0.220

For all continuous variables, mean (standard deviation). Abbreviations: MCV = mean corpuscular volume; RDW = red blood cell distribution width; WBC = white blood count.

A. Gafter-Gvili et al. / European Journal of Internal Medicine 26 (2015) 338–343

341

Table 3 Baseline laboratory parameters of the study population, according to appearance of anemia at the end of follow-up. Characteristic at baseline

Glucose (mg/dL) Creatinine level mg/dL Urea (mg/dL) Sodium (meq/L) Potassium (meq/L) Total protein (g/dL) Albumin (g/dL) Globulin (g/dL) Calcium (mg/dL) Phosphor (mg/dL) Uric acid (mg/dL) Bilirubin (mg/dL) Alkaline phosphatase (IU/L) Lactate dehydrogenase (LDH) (IU/L) GOT (IU/L) Creatinine kinase CPK (IU/L) Cholesterol total (mg/dL) Cholesterol-HDL (mg/dL) Cholesterol-LDL (mg/dL) Triglycerides (mg/dL) TSH (μU/mL) T3 (ng/dL) T4 (μg/dL) ESR (mm/h) hs-CRP (ng/mL) Homocysteine (μmol/L) PSA (ng/mL)

Men

Women

Men with anemia N = 118

Men without anemia N = 7353

p-Value

Women with anemia N = 189

Women without anemia N = 2433

p-Value

97.9 (22.3) 0.96 (0.1) 31.49 (6.9) 139.0 (2.1) 4.2 (0.2) 7.51 (0.4) 4.51 (0.3) 3.0 (0.4) 9.8 (0.4) 3.4 (0.4) 5.9 (1.2) 0.75 (0.3) 68.2 (17.387) 303.9 (61.2) 21.3 (6.0) 130.0 (103.9) 194.6 (35.3) 47.4 (9.5) 121.2 (29.2) 120.4 (68.6) 1.8 (1.4) 1.8 (0.5) 14.5 (3.5) 17.5 (9.6) 0.22 (0.3) 11.6 (3.4) 1.0 (1.4)

95.9 (16.9) 0. 99 (0.1) 30.95 (6.9) 140.0 (2.2) 4.2 (0.3) 7.48 (0.4) 4.54 (0.3) 2.9 (0.4) 9.8 (0.4) 3.3 (0.4) 6.1 (1.1) 0.81 (0.4) 69.91 (17.566) 306.5 (62.1) 22.6 (9.1) 150.7 (434.3) 197.8 (36.8) 46.3 (9.9) 123.9 (31.3) 137.8 (87.8) 1.9 (1.9) 1.8 (0.5) 16.6 (3.4) 12.1 (9.3) 0.26 (0.4) 13.9 (45.2) 0.9 (1.7)

0.352 0.055 0.406 0.499 0.504 0.507 0.295 0.100 0.457 0.017 0.009 0.049 0.292 0.651 0.148 0.607 0.344 0.260 0.392 0.033 0.918 0.911 0.138 b0.001 0.464 0.681 0.547

90.0 (10.7) 0.73 (0.1) 25.37 (5.9) 139.0 (2.1) 4.1 (0.273) 7.54 (0.4) 4.39 (0.3) 3.2 (0.4) 9.6 (0.4) 3.6 (0.4) 4.01 (1.0) 0.61 (0.3) 63.83 (20.436) 283.9 (56.7) 17.8 (6.1) 74.6 (37.3) 188.8 (29.2) 58.8 (13.6) 111.8 (25.2) 90.7 (46.9) 1.9 (1.9) 2.3 (1.4) 17.7 (11.7) 22.0 (11.8) 0.30 (0.5) 9.7 (3.5) NR

90.6 (13.6) 0.75 (0.1) 26.07 (6.6) 139.4 (2.2) 4.2 (0.3) 7.38 (0.4) 4.35 (0.3) 3.0 (0.4) 9.6 (0.4) 3.6 (0.4) 4.2 (0.9) 0.62 (0.3) 62.78 (18.186) 292.5 (57.0) 18.2 (5.7) 80.2 (66.8) 197.6 (37.3) 59.9 (13.8) 116.3 (31.7) 103.7 (65.1) 1.9 (1.5) 1.7 (0.6) 16.2 (3.6) 20.4 (11.8) 0.34 (0.5) 9.3 (3.1) NR

0.534 0.036 0.154 0.03 0.002 b0.001 0.058 b0.001 0.637 0.098 0.004 0.754 0.496 0.05 0.372 0.252 0.002 0.281 0.024 0.008 0.476 0.151 0.546 0.076 0.559 0.316 NR

For all continuous variables, mean (standard deviation). Abbreviations: BMI = body mass index; CPK = creatinine kinase; hs-CRP = high sensitive c-reactive protein; ESR = erythrocyte sedimentation rate; GOT = glutamic oxaloacetic transaminase; LDH = lactate dehydrogenase; PSA = prostate specific antigen; TSH = thyroid stimulating hormone.

protein level was a strong predictor for anemia, as well as low triglycerides. The baseline prevalence is consistent with reports of anemia in the same age range and population. Our cohort, selected from a screening center, represents a population largely of high socioeconomic status. The Third National Health and Nutrition Examination Survey (NHANES III) conducted in the United States during 1988 to 1994 demonstrated that in the 17- to 49-year-old age group, men have their lowest prevalence of anemia of about 1.5%, whereas women in their reproductive years have a prevalence greater than 12% [16], similar to what was found in our study. Of note, in the 50- to 64-year-old group the prevalence in men rises to 4.4% and the women's rates drop by half to 6.8% [16]. Our finding of 4.4% prevalence is also consistent with a recent cohort study of about 30,000 Swedish participants aged 44–73 which showed a prevalence of anemia of 3.8% [3], and a small Israeli cohort Table 4 Multivariable analysis for predictors of anemia. OR (95% CI) Men Follow-up duration in years Hemoglobin at baseline MCV Platelet count Diabetes mellitus Age Women Follow-up duration in years Hemoglobin at baseline Total protein Triglycerides Hosmer–Lemeshow test (for men) Hosmer–Lemeshow test (for women) Area under receiver operating curve (for men) Area under receiver operating curve (for women)

N = 7471 1.11 (1.04–1.17) 0.18 (0.14–0.24) 0.92 (0.89–0.96) 1.004 (1.001–1.007) 3.00 (1.41–6.39) 1.03 (1.00–1.05) N = 2622 1.06 (1.01–1.12) 0.20 (0.15–0.26) 3.44 (2.33–5.08) 0.996 (0.993–1.000) Chi-square 5.58, df = 8 Chi-square 8.37, df = 8 0.85 (0.81–0.88) 0.79 (0.76–0.82)

p value 0.001 0.000 0.000 0.001 0.004 0.023 0.025 0.000 0.000 0.037 0.694 0.398 0.000 0.000

of similar subjects who attended a periodic medical examination clinic and had a prevalence of anemia of 2.7%, in non-diabetics [17]. Data regarding longitudinal assessment of anemia development in the general population are lacking in the literature. After five years of follow-up, we found that for men the strongest predictor of anemia was the presence of diabetes mellitus. Previous studies have shown an association between anemia and diabetes mellitus. Anemia is common among diabetics with a prevalence that ranges from 10% in an Israeli cohort [17] to 25% in an Australian cohort [18], and to 45% in a Caribbean cohort [19]. Several studies have shown, as in our study, that diabetics have a significantly higher prevalence of anemia compared to non-diabetics [17,19]. As in ours, male diabetics had a significantly lower hemoglobin level than their non-diabetic counterpart, while the female diabetic and non-diabetic subjects had similar hemoglobin concentrations [16]. Our findings are unique in that we show not only increased prevalence of anemia in diabetic men, but that the presence of diabetes predicts the development of anemia. In our cohort, neither creatinine nor albumin level predicted anemia. Although diabetic nephropathy, renal insufficiency and macroalbuminuria are associated with anemia in diabetic patients [18,20], an increased risk for anemia is also observed in diabetics with normal renal function, consistent with our findings [17,21–23]. Age was also found to be a predictor for anemia in men. This is in accordance with previous reports that show that in the general population men's prevalence rates of anemia rise more rapidly than women's from middle age and onward [16]. Age was also shown to be an independent predictor for anemia in diabetic patients [17,23]. The low MCV and high platelet count, which represent iron deficiency, were also found to be predictors in men. Although more common in women of childbearing age, iron deficiency is the most common cause of anemia, and is responsible globally for about 50% of all cases of anemia [24]. In women, high total protein and low triglycerides were found to be predictors of anemia. Endocrine factors and lipid metabolism have been

342

A. Gafter-Gvili et al. / European Journal of Internal Medicine 26 (2015) 338–343

Fig. 1. The frequency of anemia according to the probability of fulfilling the model, grouped by ten percentiles. Legend: y axis — percentage of patients with anemia; x axis — probabilities, divided to ten percentiles, for developing anemia according to the prediction model. The highest percentile signifies the highest probability for fulfilling the prediction model.

shown to be involved in hematopoiesis [25]. In a recent cross-sectional study of 3519 Japanese subjects, low triglycerides and high total protein levels were significant factors that affected erythropoiesis [25]. In their study a negative correlation between triglycerides and adiponectin was shown. Adiponectin is a cytokine released by adipocytes that has been shown in basic studies to suppress hematopoiesis [26]. Other studies also showed a negative correlation between elevated adiponectin and both low hemoglobin and low triglycerides in patients with coronary artery disease [27], chronic kidney disease [28] and diabetic patients [29,30]. Moreover, adiponectin levels were found to be higher in diabetic women than men [30]. This may suggest that elevated adiponectin may be the link between low triglycerides and anemia in women our cohort, although it was not measured. This is not the first reported observation of an association between total protein level and anemia. The finding that an elevated serum protein level is predictive of anemia in women is in accordance with the study by Kohno et al. [25]. In their study, total protein was found to negatively affect erythropoiesis in both men and women. As shown in our study, an elevated transferrin level was associated with development of anemia in women in the univariate analysis. The elevated transferrin could partly explain the elevation in total protein. Total protein may also be influenced by elevated globulin levels, which were also found in the univariate analysis to be associated with anemia, and may possibly represent an underlying inflammatory state. Thus, we speculate that the elevated protein may be due to both elevation of transferrin and globulin. Low iron serum levels, low MCV and high platelet count were not found to be predictors for anemia in women, likely due to the exclusion criterion of anemia at baseline. The anemia at first visit is due to iron deficiency which is most common in women during the reproductive years due to menses and pregnancies [16]. With advancing years the causes of iron deficiency anemia diminish, allowing the unmasking of the link between low triglycerides, elevated protein and anemia. Several limitations of our study merit consideration. Our cohort is a convenience sample, rather than true sampling. Since we only evaluated

subjects who had at least two visits, we may have lost data regarding subjects who became severely anemic, and may have appointed with their primary physician, rather than the screening program. However, the people who did attend at least two visits probably considered themselves in good health. Regarding indicators for iron deficiency, the low MCV and elevated platelets which were entered into the model, displaced serum iron and transferrin saturation. This may be due to the fact that serum iron exhibits large diurnal variations, and its diagnostic specificity for iron deficiency is lost in the presence of inflammatory processes and certain other forms of chronic disease [31]. Unfortunately, ferritin, which is considered an accurate initial diagnosis test [32,33] was not taken routinely. Yet, ferritin is also an acute phase reactant and there is no single, reliable biochemical indicator that is consistently diagnostic of iron deficiency except the ‘gold standard’, bone marrow iron aspirates [34]. A low MCV has previously been shown to be highly sensitive for iron deficiency anemia, with a high likelihood ratio for development of anemia, and one strategy for anemia screening begins with the measurement of MCV [32]. Although we used self-reporting for the definition of diabetes mellitus, this is acceptable and has been shown to be highly specific for the diagnosis according to reference definitions compared with reference definitions using glucose and medication criteria [35,36]. 4.1. Implication for practice and research In conclusion, anemia is one of the world's most common preventable conditions, yet it may be often overlooked, especially in young people. Our prediction model defines several risk groups in which a closer follow-up for early identification of anemia is warranted. Diabetic patients, and especially men with diabetes, should be closely followed up for appearance of anemia. Women with low triglyceride levels and high protein levels are at risk to develop anemia. Future research should better elucidate the link between diabetes and anemia, mainly looking at preventable causes. The link between

A. Gafter-Gvili et al. / European Journal of Internal Medicine 26 (2015) 338–343

low triglyceride levels, high protein levels and anemia in women, reported by now in several studies, should be investigated and explained. Funding None. Conflict of interests The authors have no conflict of interest to declare. References [1] McLean E, Cogswell M, Egli I, Wojdyla D, de Benoist B. Worldwide prevalence of anaemia, WHO Vitamin and Mineral Nutrition Information System, 1993–2005. Public Health Nutr 2009;12:444–54. [2] Kassebaum NJ, Jasrasaria R, Naghavi M, Wulf SK, Johns N, Lozano R, et al. A systematic analysis of global anemia burden from 1990 to 2010. Blood 2014;123:615–24. [3] Martinsson A, Andersson C, Andell P, Koul S, Engstrom G, Smith JG. Anemia in the general population: prevalence, clinical correlates and prognostic impact. Eur J Epidemiol 2014;29:489–98. [4] World Health Organization. Nutritional anaemias. Report of a WHO scientific group, No 405. WHO Technical Report Series; 1968. p. 1–40 [Geneva, Switzerland]. [5] Thomas DR. Anemia and quality of life: unrecognized and undertreated. J Gerontol A Biol Sci Med Sci 2004;59:238–41. [6] Balarajan Y, Ramakrishnan U, Ozaltin E, Shankar AH, Subramanian SV. Anaemia in low-income and middle-income countries. Lancet 2011;378:2123–35. [7] Haas JD, Brownlie T. Iron deficiency and reduced work capacity: a critical review of the research to determine a causal relationship. J Nutr 2001;131:676S–88S [discussion 88S–90S]. [8] Foley RN, Parfrey PS, Kent GM, Harnett JD, Murray DC, Barre PE. Long-term evolution of cardiomyopathy in dialysis patients. Kidney Int 1998;54:1720–5. [9] Ezekowitz JA, McAlister FA, Armstrong PW. Anemia is common in heart failure and is associated with poor outcomes: insights from a cohort of 12 065 patients with new-onset heart failure. Circulation 2003;107:223–5. [10] Komajda M, Anker SD, Charlesworth A, Okonko D, Metra M, Di Lenarda A, et al. The impact of new onset anaemia on morbidity and mortality in chronic heart failure: results from COMET. Eur Heart J 2006;27:1440–6. [11] Sabatine MS, Morrow DA, Giugliano RP, Burton PB, Murphy SA, McCabe CH, et al. Association of hemoglobin levels with clinical outcomes in acute coronary syndromes. Circulation 2005;111:2042–9. [12] Ranucci M, Di Dedda U, Castelvecchio S, Menicanti L, Frigiola A, Pelissero G, et al. Impact of preoperative anemia on outcome in adult cardiac surgery: a propensitymatched analysis. Ann Thorac Surg 2012;94:1134–41. [13] Moral Garcia V, Angeles Gil de Bernabe Sala M, Nadia Diana K, Pericas BC, Nebot AG. Anemia as a surgical risk factor. Med Clin 2013;141(Suppl. 1):47–54. [14] Caro JJ, Salas M, Ward A, Goss G. Anemia as an independent prognostic factor for survival in patients with cancer: a systemic, quantitative review. Cancer 2001;91: 2214–21. [15] Fagerland MW. t-Tests, non-parametric tests, and large studies—a paradox of statistical practice? BMC Med Res Methodol 2012;12(78). [16] Guralnik JM, Eisenstaedt RS, Ferrucci L, Klein HG, Woodman RC. Prevalence of anemia in persons 65 years and older in the United States: evidence for a high rate of unexplained anemia. Blood 2004;104:2263–8.

343

[17] Grossman C, Dovrish Z, Koren-Morag N, Bornstein G, Leibowitz A. Diabetes mellitus with normal renal function is associated with anaemia. Diabetes Metab Res Rev 2014;30:291–6. [18] Thomas MC, Tsalamandris C, MacIsaac RJ, Jerums G. The epidemiology of hemoglobin levels in patients with type 2 diabetes. Am J Kidney Dis 2006;48:537–45. [19] Ezenwaka CE, Jones-Lecointe A, Nwagbara E, Seales D, Okali F. Anaemia and kidney dysfunction in Caribbean type 2 diabetic patients. Cardiovasc Diabetol 2008;7:25. [20] Angelousi A, Larger E. Anaemia, a common but often unrecognized risk in diabetic patients: a review. Diabetes Metab 2014;41:18–27. [21] Astor BC, Muntner P, Levin A, Eustace JA, Coresh J. Association of kidney function with anemia: the Third National Health and Nutrition Examination Survey (1988– 1994). Arch Intern Med 2002;162:1401–8. [22] Thomas MC, MacIsaac RJ, Tsalamandris C, Power D, Jerums G. Unrecognized anemia in patients with diabetes: a cross-sectional survey. Diabetes Care 2003;26:1164–9. [23] Goldhaber A, Ness-Abramof R, Ellis MH. Prevalence of anemia among unselected adults with diabetes mellitus and normal serum creatinine levels. Endocr Pract 2009;15:714–9. [24] Stoltzfus RJ. Iron deficiency: global prevalence and consequences. Food Nutr Bull 2003;24:S99–S103. [25] Kohno K, Narimatsu H, Shiono Y, Suzuki I, Kato Y, Fukao A, et al. Management of erythropoiesis: cross-sectional study of the relationships between erythropoiesis and nutrition, physical features, and adiponectin in 3519 Japanese people. Eur J Haematol 2014;92:298–307. [26] DiMascio L, Voermans C, Uqoezwa M, Duncan A, Lu D, Wu J, et al. Identification of adiponectin as a novel hemopoietic stem cell growth factor. J Immunol 2007;178: 3511–20. [27] Hara K, Uchida T, Takebayashi K, Sakai Y, Inoue T, Inukai T, et al. Determinants of serum high molecular weight (HMW) adiponectin levels in patients with coronary artery disease: associations with cardio-renal-anemia syndrome. Intern Med 2011; 50:2953–60. [28] Abe Y, Matsumae T, Eto S, Watanabe M, Ito K, Sasatomi Y, et al. Total and HMW adiponectin is independently associated with B-type natriuretic peptide and anemia in chronic hemodialysis patients. Intern Med 2012;51:3247–52. [29] Aso Y, Suganuma R, Wakabayashi S, Hara K, Nakano T, Suetsugu M, et al. Anemia is associated with an elevated serum level of high-molecular-weight adiponectin in patients with type 2 diabetes independently of renal dysfunction. Transl Res 2009; 154:175–82. [30] Aso Y, Yamamoto R, Wakabayashi S, Uchida T, Takayanagi K, Takebayashi K, et al. Comparison of serum high-molecular weight (HMW) adiponectin with total adiponectin concentrations in type 2 diabetic patients with coronary artery disease using a novel enzyme-linked immunosorbent assay to detect HMW adiponectin. Diabetes 2006;55:1954–60. [31] Cohen-Solal A, Leclercq C, Deray G, Lasocki S, Zambrowski JJ, Mebazaa A, et al. Iron deficiency: an emerging therapeutic target in heart failure. Heart 2014;100: 1414–20. [32] Guyatt GH, Oxman AD, Ali M, Willan A, McIlroy W, Patterson C. Laboratory diagnosis of iron-deficiency anemia: an overview. J Gen Intern Med 1992;7:145–53. [33] Killip S, Bennett JM, Chambers MD. Iron deficiency anemia. Am Fam Physician 2007; 75:671–8. [34] Clark SF. Iron deficiency anemia. Nutr Clin Pract 2008;23:128–41. [35] Goto A, Morita A, Goto M, Sasaki S, Miyachi M, Aiba N, et al. Validity of diabetes selfreports in the Saku diabetes study. J Epidemiol 2013;23:295–300. [36] Schneider AL, Pankow JS, Heiss G, Selvin E. Validity and reliability of self-reported diabetes in the atherosclerosis risk in communities study. Am J Epidemiol 2012; 176:738–43.

Predicting the emergence of anemia--A large cohort study.

We aimed to find predictors for development of anemia in a large cohort of adults...
325KB Sizes 0 Downloads 7 Views