Thyroid Dysfunction and Anemia: A Prospective Cohort Study and a Systematic Review

Abstract

Background:

Even though the association between thyroid dysfunction and anemia is commonly described, it is not known whether it is clinically relevant. This study set out to quantify the association of thyroid dysfunction on hemoglobin (Hb) concentration and risk of anemia. A systematic review (MEDLINE and EMBASE, from inception until May 15, 2017) was conducted to interpret the findings in context.

Methods:

Participants from the EPIC-Norfolk cohort with available baseline thyrotropin (TSH), free thyroxine (fT4), and Hb were included. Euthyroidism was defined as TSH 0.45–4.49 mIU/L (reference category), hypothyroidism as TSH ≥4.50 mIU/L (subclinical [SHypo] with normal fT4 or overt [OHypo] with low fT4), and hyperthyroidism as TSH ≤0.44 mIU/L (subclinical [SHyper] with normal fT4 or overt [OHyper] with elevated fT4). Anemia was defined as Hb <12 g/dL in women and Hb <13 g/dL in men. In the cross-sectional analyses, multiple linear regression was used to compare Hb across TSH categories. In the prospective analysis, participants with OHypo/OHyper at baseline were excluded, as it was assumed that they were treated for overt thyroid disease. A covariance model was used to determine change in Hb concentration from baseline to last follow-up, and multivariable Cox regression was used to analyze anemia risk.

Results:

In the cross-sectional population (n = 12,337), the adjusted Hb was 0.22 g/dL lower [confidence interval (CI) 0.07–0.38] in OHypo compared to euthyroids, and 0.08 g/dL lower [CI −0.23 to 0.38] in OHyper. In the prospective analysis, 460/7031 participants developed anemia over a median follow-up of 4.7 years. The adjusted mean Hb change over time was −0.04 g/dL in SHypo [CI −0.14 to 0.06] and 0.05 g/dL in SHyper [CI −0.10 to 0.20]. The adjusted hazard ratio for anemia was 0.99 [CI 0.67–1.48] in SHypo, and 0.52 [CI 0.23–1.16] in SHyper. The systematic review returned no other prospective studies on this association, but cross-sectional and case-control studies showed comparable results.

Conclusion:

In this first prospective population-based cohort, subclinical thyroid dysfunction was not associated with a change in Hb concentration during follow-up and was not an independent risk factor for developing anemia; variations in Hb concentration in patients with overt thyroid dysfunction were not clinically relevant.

Introduction

Acausal relationship between overt hypothyroidism (OHypo)/overt hyperthyroidism (OHyper) and anemia has been postulated for decades (1 –3). Researchers have suggested several pathophysiological mechanisms (4,5). Hyperthyroidism enhances erythropoiesis, but it also raises the basal metabolic rate. The latter may increase plasma volume (3), counteracting the effect of erythropoiesis and lowering hemoglobin (Hb) concentration, ultimately causing anemia. In hypothyroidism, a drop in basal metabolic rate and a decrease in cellular oxygen consumption may reduce erythropoietin secretion, which would also lower Hb concentration and, ultimately, cause either normocytic, microcytic, or macrocytic anemia, depending on comorbidities (1,6). Although several studies explored the pathophysiologic link between thyroid dysfunction and erythroid development (5,7), it is not known whether, or to what extent, thyroid dysfunction alone (in the absence of comorbidities) affects Hb concentration, or if the effect would be clinically relevant in large epidemiological studies.

Limited data exist on the association between thyroid dysfunction and anemia. Case series from the 1960s and 1970s had no control group (1,8,9). More recent cross-sectional studies did not consider potentially relevant confounders, such as renal function or C-reactive protein (CRP) (10,11), and included few patients with hypothyroidism (10 –13). Only two studies have reported on hyperthyroidism and anemia (11,14), and no prospective study has assessed the incidence of anemia in subclinical thyroid dysfunction.

Therefore, this study aimed to quantify the effect of thyroid dysfunction on Hb concentration in a large population-based cohort and to assess its association with anemia. Both a cross-sectional and prospective analysis were conducted, and a systematic review was completed.

Materials and Methods

This study is reported in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement (15). The results of the systematic review of observational studies are reported according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) statement (16). For the systematic review, one author (C.F.) searched OVID-MEDLINE and EMBASE (from inception to May 15, 2017) for studies that examined the link between thyroid function and anemia, and checked bibliographies of included articles. Studies with results on both Hb/anemia and thyroid dysfunction in adult, non-pregnant participants were included, and those without a euthyroid control group were excluded. The Supplementary Appendix (Supplementary Data are available online at www.liebertpub.com/thy) contains details of the search strategy. For quality assessment, each included study was rated using the Newcastle–Ottawa Quality Assessment Scale (NOS) for case-control studies or an adapted version for cross-sectional studies.

Setting and study population

Data were rated from the European Prospective Investigation of Cancer (EPIC) Norfolk cohort, a prospective observational study of 25,639 women and men, 40–79 years of age at enrolment (17,18). Everyone in this age range registered in the general practitioner databases in Norwich, United Kingdom, and surroundings, was invited from 1993 to 1997. The Norwich local research ethics committee (United Kingdom) approved the study, and all participants gave written informed consent. Participants were included in the cross-sectional analysis if measurements of their baseline thyrotropin (TSH), free thyroxine (fT4), and Hb were available at study entry. Those who self-reported thyroid dysfunction were excluded because it was assumed that most would already be in treatment for thyroid dysfunction. For the prospective analysis, the following were excluded: (i) participants with anemia at baseline (as anemia would be the prospective outcome), (ii) those with no available Hb measurments during follow-up, (iii) and those diagnosed with overt thyroid dysfunction at baseline, since it was assumed that they received thyroid therapy after diagnosis. Descriptive results on thyroid dysfunction and anemia were already reported in this population in a cross-sectional explanatory analysis without multivariable models to account for confounders through a rapid communication (19). The present study reports on new multivariable analyses and new prospective analyses that were not available in the previous rapid communication.

Definition of main outcomes, exposures, and potential confounders

The primary outcome of the cross-sectional analysis was Hb concentration. The secondary outcome was anemia, defined as Hb <12 g/dL for women and <13 g/dL for men (binary variable) (20). The main exposure variable was thyroid dysfunction, classified by TSH categories. To make the study easier to compare with others, guided by expert reviews (21,22), hypothyroidism was defined as TSH ≥4.50 mIU/L, either subclinical (SHypo), with normal fT4 or overt (OHypo) with fT4 below the reference range (23). Hyperthyroidism was defined as TSH ≤0.44 mIU/L, either subclinical (SHyper) with normal fT4 or overt (OHyper) with fT4 above the reference range (23). The reference category was euthyroidism (TSH 0.45–4.49 mIU/L and fT4 within the reference range 9.00–20.00 pmol/L) (23). A secondary exposure variable was fT4. In the prospective analysis, the primary outcome was the difference in Hb concentration between the last available follow-up Hb and baseline Hb (numerical variable, g/dL). The secondary outcome was incident anemia. The main exposure variable was thyroid dysfunction, classified by TSH categories (i.e., SHypo, SHyper, euthyroidism as reference). The secondary exposure variable was fT4 (categorized in quintiles), as in the cross-sectional analysis. The following potential confounders were measured at baseline: age (continuous variable); sex; body mass index (continuous variable); smoking (never, past, present); self-reported history of myocardial infarction, diabetes, stroke (binary variables; yes/no); ferritin (continuous variable); estimated glomerular filtration rate (eGFR; continuous variable); CRP (continuous variable); and mean corpuscular volume (MCV; continuous variable; as a surrogate parameter for B12/folic acid deficiency, since these measurements were unavailable in the EPIC-Norfolk study).

Statistical analyses

In the cross-sectional analysis, multiple linear regression was used to compare Hb across TSH categories (OHypo, SHypo, SHyper, OHyper, euthyroidism as reference category). Multiple linear regression was also used to compare Hb concentration across fT4 quintiles (mid-quintile as reference). To explore the association between thyroid dysfunction and anemia, logistic regression was used to compare the odds of anemia across TSH categories and fT4 quintiles. For all analyses, three models were used: unadjusted; age- and sex-adjusted; and fully adjusted considering the confounders already mentioned. The inclusion of confounders was based on a likelihood ratio test (p < 0.1 as cut-off) that indicated which confounders were most likely to confound the association between exposure and outcome in bivariable models.

In the prospective analysis, a covariance model (ANCOVA) was analyzed to determine change in Hb concentration from baseline to last follow-up, taking into account baseline Hb concentrations (24). Results are presented as difference in mean change of Hb between SHypo/SHyper and euthyroidism (24,25). Then, multivariable Cox regression models were used to compare the incidence of anemia in SHypo/SHyper to euthyroidism and, in a further analysis, the incidence of anemia across fT4 quintiles Again, three models were used: unadjusted; age- and sex-adjusted; and fully adjusted, similar to the cross-sectional analyses described above.

Multiple imputation (30 imputations) were used to complete missing values in potentially relevant confounders in cross-sectional and prospective analyses (for the cross-sectional population: body mass index 24 [0.19%] missing, smoking status 93 [0.75%] missing, ferritin 3851 [31.2%] missing, eGFR 3266 [26.5%] missing, and CRP 3311 [26.8%] missing; for the prospective population: body mass index 11 [0.16%] missing, smoking status 44 [0.63%] missing, ferritin 2068 [29.4%] missing, eGFR 1835 [26.1%] missing, and CRP 1855 [26.4%] missing), and further assessed and ruled out that iron status modifies the association between thyroid dysfunction and Hb (26). Finally, sensitivity analyses were conducted that used the original, unimputed data. Tests were two-sided, at a 0.05 level of significance. Stata v14.2 (StataCorp LP, College Station, TX) was used for all the analyses.

Results

After excluding participants with self-reported thyroid disease (n = 635), 12,337 participants were included in the cross-sectional analysis. Of these, 976 (7.9%) had anemia, most (11,174; 90.6%) were euthyroid; 644 (5.2%) had SHypo, 199 (1.6%) OHypo, 270 (2.2%) SHyper, and 50 (0.4%) had OHyper (Table 1). No relevant differences were observed in Hb concentration among TSH categories. In multivariable analyses, Hb was 0.22 g/dL [confidence interval (CI) 0.07–0.38] lower in OHypo than in euthyroidism, and 0.08 g/dL [CI −0.23 to 0.38] lower in OHyper (Table 2). The adjusted odds ratio (OR) for anemia was 1.96 [CI 1.29–2.98] in OHypo and 2.18 [CI 0.98–4.87] in OHyper, but no association was found in SHypo/SHyper (Supplementary Table S2).

Table 1.

Baseline Characteristics of the Cross-Sectional Study Population (N = 12,337)

	OHypo	SHypo	Euthyroidism	SHyper	OHyper	p-Value
Participants	199 (1.6)	644 (5.2)	11,174 (90.6)	270 (2.2)	50 (0.4)
Demographics
Age (years)	60.4 (9.2)	60.7 (9.0)	58.6 (9.5)	59.1 (10.0)	62.3 (9.9)	<0.001
Female	149 (74.9)	444 (68.9)	5,785 (51.8)	151 (55.9)	28 (56.0)	<0.001
Laboratory parameters
TSH (mIU/L)	25.66 (31.5)	7.78 (16.6)	1.84 (0.8)	0.27 (0.1)	0.16 (0.2)	<0.001
fT4 (pmol/L)	7.3 (1.6)	11.2 (1.5)	12.4 (1.8)	14.0 (2.2)	35.5 (19.6)	<0.001
Anemia^a	29 (14.6)	55 (8.5)	866 (7.8)	18 (6.7)	8 (16.0)	0.001
Hemoglobin (g/dL)	13.4 (1.2)	13.6 (1.4)	13.9 (1.3)	13.8 (1.3)	13.7 (1.6)	<0.001
MCV (fl)	89.4 (4.5)	88.9 (4.5)	89.3 (4.4)	89.2 (4.5)	88.5 (4.3)	0.196
Ferritin (ng/mL)	78.5 (71.9)	79.2 (63.6)	89.6 (75.2)	84.2 (77.2)	95.6 (101.4)	0.019
CRP (mg/L)	3.6 (6.5)	4.2 (12.0)	3.0 (5.6)	3.9 (11.2)	3.8 (6.1)	<0.001
eGFR (mL/min)	70.9 (20.0)	72.5 (19.5)	77.7 (24.9)	82.9 (35.7)	74.3 (25.9)	<0.001
BMI categories
<25.0 kg/m²	66 (33.2)	232 (36.0)	4,432 (39.7)	122 (45.2)	20 (40.0)	Overall 0.046
25.0–29.9 kg/m²	90 (45.2)	292 (45.3)	5,099 (45.6)	112 (41.5)	23 (46.0)
≥30.0 kg/m²	42 (21.1)	118 (18.3)	1,623 (14.5)	35 (13.0)	7 (14.0)
Smoking status
Never	88 (44.2)	342 (53.1)	5,055 (45.2)	107 (39.6)	25 (50.0)	Overall <0.001
Past	89 (44.7)	253 (39.3)	4,693 (42.0)	122 (45.2)	20 (40.0)
Present	19 (9.6)	41 (6.4)	1,348 (12.1)	38 (14.1)	4 (8.0)
Medical history
Myocardial infarction	4 (2.0)	18 (2.8)	342 (3.1)	3 (1.1)	0	0.216
Stroke	0	16 (2.5)	134 (1.2)	5 (2.2)	2 (4.0)	0.004
Diabetes mellitus	2 (1.0)	13 (2.0)	241 (2.2)	4 (1.5)	3 (6.0)	0.247

Results are presented as number (percentage) for categorical variables and as mean (standard deviation) for continuous variables. p-Values derived from one-way analysis of variance models in case of continuous variables and from chi-square tests in case of categorical variables.

Missing data: BMI (24; 0.19%), smoking status (93; 0.75%), ferritin (3851; 31.2%), eGFR (3266; 26.5%), and CRP (3311; 26.8%).

Definition of anemia: Hb <12 g/dL for women, Hb <13 g/dL for men.

OHypo, overt hypothyroidism; SHypo, subclinical hypothyroidism; SHyper, subclinical hyperthyroidism; OHyper, overt hyperthyroidism; TSH, thyrotropin; fT4, free thyroxine; MCV, mean corpuscular volume; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; BMI, body mass index.

Table 2.

Multiple Linear Regression: Hemoglobin Concentrations (g/dL) Compared Across Thyroid Dysfunction Categories in the Cross-Sectional Study Population (N = 12,337)

	OHypo	SHypo	Euthyroidism	SHyper	OHyper
Participants (%)	199 (1.6)	644 (5.2)	11,174 (90.6)	270 (2.2)	50 (0.4)
Crude analysis	−0.52 [−0.71 to −0.33]	−0.29 [−0.40 to −0.18]	Ref.	−0.10 [−0.26 to 0.06]	−0.16 [−0.53 to 0.21]
Age- and sex-adjusted analysis	−0.19 [−0.35 to −0.03]	−0.05 [−0.14 to 0.04]	Ref.	−0.04 [−0.18 to 0.10]	−0.11 [−0.43 to 0.20]
Multivariable^a analysis with imputed^b data	−0.22 [−0.38 to −0.07]	−0.03 [−0.11 to 0.06]	Ref.	−0.02 [−0.15 to 0.11]	−0.08 [−0.38 to 0.23]
Multivariable^a analysis with original^c data	−0.28 [−0.45 to −0.11]	−0.04 [−0.14 to 0.05]	Ref.	−0.01 [−0.17 to 0.16]	−0.42 [−0.82 to −0.03]

Data in square brackets are confidence intervals.

Adjustments: age, sex, BMI, smoking status, myocardial infarction, diabetes mellitus, stroke, ferritin, eGFR, CRP, MCV.

Imputation for BMI (24 [0.19%] missing), smoking status (93 [0.75%] missing), ferritin (3,851 [31.2%] missing), eGFR (3,266 [26.5%] missing), and CRP (3,311 [26.8%] missing). All other variables were complete.

N = 8275.

Ref., reference.

For the prospective analysis, participants with anemia at baseline (n = 976), overt thyroid dysfunction at baseline (n = 212), and missing follow-up data on Hb (n = 4118) were excluded. The latter group was similar to the included prospective study population with regard to thyroid (dys-)function, age, sex, and baseline Hb concentration. A total of 7031 participants (Table 3) were followed over 55,733 person years (median 4.7 years). Anemia developed during follow-up in 26/368 (7.1%) participants with SHypo, 428/6496 (6.6%) participants with euthyroidism, and 6/167 (4.0%) participants with SHyper. In the ANCOVA analysis, when SHypo/SHyper were compared with euthyroidism, no difference was found in adjusted mean Hb change (SHypo: -0.04 g/dL [CI −0.14 to 0.06]; SHyper: 0.05 g/dL [CI −0.10 to 0.20]; Table 4). When SHypo and SHyper were compared to euthyroidism, the adjusted hazard ratio for anemia was 0.99 for SHypo [CI 0.67–1.48] and 0.52 for SHyper ([CI 0.23–1.16]; Table 5).

Table 3.

Baseline Characteristics of the Prospective Study Population (N = 7031)

	SHypo	Euthyroidism	SHyper	p-Value
Participants	368 (5.2)	6496 (92.4)	167 (2.4)
Demographics
Age (years)	60.1 (8.8)	58.2 (9.0)	57.5 (10.0)	<0.001
Female	248 (67.4)	3302 (50.8)	95 (56.9)	<0.001
Laboratory parameters
TSH (mIU/L)	7.1 (4.5)	1.8 (0.9)	0.27 (0.13)	<0.001
fT4 (pmol/L)	11.2 (1.6)	12.4 (1.7)	13.9 (2.2)	<0.001
Hemoglobin (g/dL)	13.8 (1.1)	14.0 (1.2)	14.0 (1.3)	<0.001
MCV (fl)	89.0 (4.1)	89.5 (3.8)	89.6 (3.7)	0.057
Ferritin (ng/mL)	77.1 (59.4)	89.9 (74.8)	89.5 (84.7)	0.021
CRP (mg/L)	3.6 (11.8)	2.6 (4.7)	3.9 (13.4)	0.001
eGFR (mL/min)	72.2 (18.4)	77.7 (24.0)	82.1 (36.9)	<0.001
BMI categories
<25.0 kg/m²	143 (39.0)	2715 (41.9)	80 (47.9)	Overall 0.288
25.0–29.9 kg/m²	163 (44.4)	2926 (45.1)	66 (39.5)
≥30.0 kg/m²	61 (16.6)	845 (13.0)	21 (12.6)
Smoking status
Never	205 (56.3)	3114 (48.2)	70 (41.9)
Past	136 (37.4)	2681 (41.5)	75 (44.9)	Overall 0.008
Present	23 (6.3)	661 (10.2)	22 (13.2)
Medical history
Myocardial infarction	9 (2.5)	155 (2.4)	1 (0.6)	0.319
Stroke	9 (2.5)	58 (0.9)	1 (0.6)	0.011
Diabetes mellitus	7 (1.9)	112 (1.7)	3 (1.8)	0.966

Results are presented as number (percentage) for categorical variables and as mean (standard deviation) for continuous variables. Only participants without anemia or overt thyroid dysfunction at baseline and with available hemoglobin assessment during follow-up were included. p-Values derived from one-way analysis of variance models in case of continuous variables and from chi-square tests in case of categorical variables.

Missing data: BMI (11; 0.16%), smoking status (44; 0.63%), ferritin (2068; 29.4%), eGFR (1835; 26.1%), and CRP (1855; 26.4%).

Table 4.

Changes in Hemoglobin Concentration (g/dL) Compared Across Thyroid Dysfunction Categories in the Prospective Cohort (N = 7031)

	SHypo	Euthyroidism	SHyper
Participants (%)	368 (5.2%)	6496 (92.4%)	167 (2.4%)
Baseline hemoglobin, mean (SD)	13.8 (1.1)	14.0 (1.2)	14.0 (1.3)
Follow-up hemoglobin, mean (SD)	13.8 (1.2)	14.1 (1.3)	14.1 (1.1)
Difference in adjusted^a mean change of hemoglobin [CI], imputed^b data set	−0.04 [−0.14 to 0.06]	Ref.	0.05 [−0.10 to 0.20]

Only participants without anemia or overt thyroid dysfunction at baseline and with available hemoglobin assessment during follow-up were included.

Adjustments: baseline hemoglobin, length of follow-up in years, age, sex, BMI, smoking status, myocardial infarction, diabetes mellitus, stroke, ferritin, eGFR, CRP, and MCV.

Imputation for BMI (11 [0.16%] missing), smoking status (44 [0.63%] missing), ferritin (3,851 [31.2%] missing), eGFR (3,266 [26.5%] missing), and CRP (3,311 [26.8%] missing). All other variables were complete.

SD, standard deviation; CI, confidence interval.

Table 5.

Hazard Ratio for Anemia: Subclinical Thyroid Dysfunction Compared to Euthyroidism in the Prospective Study Population (N = 7031)

	SHypo	Euthyroidism	SHyper
Anemia, N (%)	26 (7.1)	428 (6.6)	6 (3.6)
HR for anemia [CI]
Crude analysis	0.94 [0.63–1.39]	Ref.	0.52 [0.23–1.16]
Age- and sex-adjusted analysis	1.00 [0.67–1.49]	Ref.	0.51 [0.23–1.13]
Multivariable^a analysis with imputed^b data	0.99 [0.67–1.48]	Ref.	0.52 [0.23–1.16]
Multivariable^a analysis with original data^c	0.86 [0.53–1.40]	Ref.	0.69 [0.29–1.68]

Data in square brackets are confidence intervals. Only participants without anemia or overt thyroid dysfunction at baseline and with available hemoglobin assessment during follow-up were included.

Adjustments: age, sex, BMI, diabetes mellitus, ferritin, and MCV.

Imputation for BMI (11 [0.16%] missing), smoking status (44 [0.63%] missing), ferritin (3851 [31.2%] missing), eGFR (3266 [26.5%] missing), and CRP (3311 [26.8%] missing). All other variables were complete.

N = 4955.

HR, hazard ratio.

All results (from cross-sectional and prospective analyses) were similar in the sensitivity analyses, which used the original unimputed data (Supplementary Tables S1–S4). Results for fT4 as the exposure variable in both the cross-sectional and the prospective analyses, with imputed and original data, yielded no relevant associations (Supplementary Tables S1–S4).

The literature search (see the Supplementary Appendix for detailed search strategy) identified 2721 articles in OVID-MEDLINE and EMBASE. After excluding duplicates, title and abstract were used to screen 2692 articles. A total of 2661 papers were excluded, and 31 full-text articles were assessed (Supplementary Fig. S1). A manual check of article bibliographies revealed 14 more potentially eligible studies. Of these 45 candidates, 37 articles were excluded after full-text evaluation (Supplementary Table S5), and eight articles finally met the inclusion criteria (Supplementary Table S6); four were cross-sectional studies (10,12 –14), four were case-control studies (7,11,27,28), and none had a prospective design.

Two of the four cross-sectional studies were population based and of fair methodological quality, scoring five out of seven points in the NOS (Supplementary Table S7) (10,13). Yet, both studies were small, included <100 patients with SHypo and/or OHypo, excluded participants with hyperthyroidism (both SHyper and OHyper), and did not report significant differences in Hb concentrations across exposure categories. After adjusting for relevant confounders, Bremner et al. found Hb concentration was 13.7 g/dL in participants with SHypo and 14.2 g/dL (p = 0.29) in participants with euthyroidism (13). Den Elzen et al. found Hb concentration was 12.8 g/dL in participants aged >85 years with OHypo, 13.1 g/dL in those with SHypo, and 13.0 g/dL in those with euthyroidism (10). In a retrospective analysis of 6534 consecutive female patients referred to a university hospital, Lippi et al. found similar Hb concentrations among participants with hypothyroidism (13.2 g/dL), euthyroidism (13.3 g/dL), and hyperthyroidism (13.1 g/dL) (14), but used uncommon definitions for thyroid dysfunction (hypothyroidism TSH >2.5 mIU/L; euthyroidism TSH 0.2–2.5 mIU/L; hyperthyroidism TSH <0.2 mIU/L) (14). These could have biased the results toward the null effect (23). Regarding methodological quality, the NOS score was moderate, with three out of seven points (Supplementary Table S7). In their cross-sectional analysis, however, Vitale et al. reported higher prevalence of hypothyroidism in hospitalized patients aged >65 years with anemia (20%) than in control patients without anemia (9.9%; p = 0.01) (12). In this study, Hb concentrations increased an average of 1.6 g/dL in nine patients with hypothyroidism and anemia after they were treated with thyroxine (12). It was not clear, however, how these nine patients with anemia, hypothyroidism, and no other potential cause of anemia were selected for thyroxine treatment. Overall, the methodological quality was moderate, with a NOS score of three out of seven points (Supplementary Table S7). Four case-control studies were included in the systematic review (7,11,27,28). Each reported slightly more pronounced differences in Hb concentrations across the categories of thyroid dysfunction. Mean Hb concentrations among euthyroid participants ranged from 12.8 to 14.7 g/dL; corresponding values were 10.8–12.7 g/dL in SHypo and 10.7–13.2 g/dL in OHypo. These studies included participants with more severe thyroid dysfunction (e.g., in Jafarzadeh et al., the 50 participants with OHypo had a mean TSH of 136.5 mIU/L) (11). They were of low methodological quality, scoring 2/10 points in the adapted NOS (Supplementary Table S8). These methodological limitations make it hard to determine if the slightly more pronounced differences in Hb concentrations were real (because thyroid dysfunction was more severe) or spurious (a product of bias).

Only two of the eight included studies considered hyperthyroidism (11,14). Both were negative. In the study by Lippi et al., Hb was 13.1 g/dL among hyperthyroid participants (SHyper and OHyper) and 13.3 g/dL among euthyroid participants (14). In the study by Jafarzadeh et al., Hb was 13.8 g/dL in OHyper and 14.0 g/dL in euthyroidism (11).

Discussion

In this first prospective population-based cohort, subclinical thyroid dysfunction was not associated with a change in Hb concentration during follow-up and was not an independent risk factor for developing anemia. OHypo and OHyper were associated with anemia in the cross-sectional analyses. This is congruent with pathophysiologic findings that thyroid hormone receptor alpha is expressed by hematopoietic and erythroid progenitors and regulates erythropoiesis (4,5,7). However, the differences in mean Hb concentrations between OHypo/OHyper participants and euthyroidism were not clinically relevant. In the cross-sectional and prospective analysis, no relevant association was found between fT4 and anemia. The systematic review did not identify other prospective cohort studies on the association between thyroid dysfunction and anemia/hemoglobin, and results from the included cross-sectional and case-control studies were comparable to the results of this cohort study.

This study has limitations. First, thyroid hormones were only measured once at baseline. Patients who had transient subclinical thyroid dysfunction, who developed overt thyroid over time, or who reverted to euthyroidism over time may have been misclassified. This single measurement may have biased the prospective results toward the null-effect, but the same is true of most large prospective cohorts that actually found associations with other outcomes such as coronary heart disease (23) and fractures (29). Second, no measurements for vitamin B12 and folic acid were available. Despite this, the daily intake of participants included the current recommended dose of both (vitamin B12 200–350 μg, folic acid 2–14 mg) (30 –32), and the study adjusted for MCV (a surrogate parameter for B12/folic acid deficiency). Third, although the EPIC-Norfolk cohort is population based, data could be analyzed from only 12,337 (48.1%) of the 25,639 eligible participants because no thyroid function or hemoglobin measurements were available for some participants. It is not possible to be sure that the study population represents the whole potentially eligible population. Fourth, in the prospective analyses, healthy participants could have been more likely to be available for follow-up examinations than sick participants (healthy cohort effect). If sick participants were both more anemic and more likely to suffer from thyroid dysfunction, then the association between subclinical thyroid dysfunction and anemia may have been underestimated. However, the consistent results between the cross-sectional and the prospective analyses suggest that any healthy cohort effect was not substantial. Fifth, participants who developed conditions during follow-up that could affect Hb (e.g., hematological malignancies, chronic diseases) could not be excluded, as this information was not available. As hematological malignancies are not associated with thyroid dysfunction, they should not be an important confounder. Chronic diseases could be associated with thyroid dysfunction, at least with subclinical hypothyroidism. In the prospective analyses, incident anemia cases may have been attributed to thyroid dysfunction when, in fact, anemia occurred due to chronic diseases. In this case, however, the association between thyroid dysfunction and anemia would have been overestimated, while significant large associations were not found. Finally, the observational design of the study precluded causal inference.

This study also has several strengths. It is the largest population-based study to examine the association between thyroid dysfunction and Hb concentration/anemia and to consider both hypo- and hyperthyroidism. It is the first prospective cohort study on the topic to have a long follow-up period (almost five years). The results are consistent between our cross-sectional, prospective, and sensitivity analyses, and the study was able to adjust for most potentially relevant confounders. The findings are strengthened by the systematic review that was conducted, since it enabled the results to be interpreted in the context of all available literature on the topic.

Conclusion

In this first prospective population-based cohort, subclinical thyroid dysfunction was not associated with a change in Hb concentration during follow-up and was not an independent risk factor for developing anemia. Variations in Hb concentration in patients with overt thyroid dysfunction were not clinically relevant.

Footnotes

Acknowledgments

This project was supported by grants from the Swiss National Science Foundation (SNSF 320030-150025 and 320030-172676 to N.R.; P3SMP3-155318, PZ00P3-167826 to T.-H.C.). We thank Kali Tal for her editorial assistance.

References

Horton

, Coburn

, England

, Himsworth

. 1976. The haematology of hypothyroidism. Q J Med, 45:101–123.

Fein

, Rivlin

. 1975. Anemia in thyroid diseases. Med Clin North Am, 59:1133–1145.

Ford

, Carter

. 1988. The haematology of hyperthyroidism: abnormalities of erythrocytes, leucocytes, thrombocytes and haemostasis. Postgrad Med J, 64:735–742.

Kendrick

, Payne

, Epis

, Schneider

, Leedman

, Klinken

, Ingley

. 2008. Erythroid defects in TRalpha–/– mice. Blood, 111:3245–3248.

Chute

, Ross

, McDonnell

. 2010. Minireview: nuclear receptors, hematopoiesis, and stem cells. Mol Endocrinol, 24:1–10.

Nightingale

, Vitek

, Himsworth

. 1978. The haematology of hyperthyroidism. Q J Med, 47:35–47.

Erdogan

, Kosenli

, Ganidagli

, Kulaksizoglu

. 2012. Characteristics of anemia in subclinical and overt hypothyroid patients. Endocr J, 59:213–220.

Tudhope

, Wilson

. 1960. Anaemia in hypothyroidism. Incidence, pathogenesis, and response to treatment. Q J Med, 29:513–537.

Rivlin

, Wagner

Jr . 1969. Anemia in hyperthyroidism. Ann Intern Med, 70:507–516.

10.

den Elzen

, de Craen

, Mooijaart

, Gussekloo

. 2015. Low thyroid function and anemia in old age: the Leiden 85-plus study. J Am Geriatr Soc, 63:407–409.

11.

Jafarzadeh

, Poorgholami

, Izadi

, Nemati

, Rezayati

. 2010. Immunological and hematological changes in patients with hyperthyroidism or hypothyroidism. Clin Invest Med, 33:E271–279.

12.

Vitale

, Fatti

, Prolo

, Girola

, Caraglia

, Marra

, Abbruzzese

, Gerli

, Mari

. 2010. Screening for hypothyroidism in older hospitalized patients with anemia: a new insight into an old disease. J Am Geriatr Soc, 58:1825–1827.

13.

Bremner

, Feddema

, Joske

, Leedman

, O'Leary

, Olynyk

, Walsh

. 2012. Significant association between thyroid hormones and erythrocyte indices in euthyroid subjects. Clin Endocrinol, 76:304–311.

14.

Lippi

, Montagnana

, Salvagno

, Guidi

. 2008. Should women with abnormal serum thyroid stimulating hormone undergo screening for anemia?. Arch Pathol Lab Med, 132:321–322.

15.

Vandenbroucke

, von Elm

, Altman

, Gotzsche

, Mulrow

, Pocock

, Poole

, Schlesselman

, Egger

, Initiative

. 2014. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Int J Surg, 12:1500–1524.

16.

Stroup

, Berlin

, Morton

, Olkin

, Williamson

, Rennie

, Moher

, Becker

, Sipe

, Thacker

. 2000. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA, 283:2008–2012.

17.

Hayat

, Luben

, Keevil

, Moore

, Dalzell

, Bhaniani

, Khawaja

, Foster

, Brayne

, Wareham

, Khaw

. 2014. Cohort profile: a prospective cohort study of objective physical and cognitive capability and visual health in an ageing population of men and women in Norfolk (EPIC-Norfolk 3). Int J Epidemiol, 43:1063–1072.

18.

Day

, Oakes

, Luben

, Khaw

, Bingham

, Welch

, Wareham

. 1999. EPIC-Norfolk: study design and characteristics of the cohort. European Prospective Investigation of Cancer. Br J Cancer, 80:95–103.

19.

M'Rabet-Bensalah

, Aubert

, Coslovsky

, Collet

, Baumgartner

, den Elzen

, Luben

, Angelillo-Scherrer

, Aujesky

, Khaw

, Rodondi

. 2016. Thyroid dysfunction and anaemia in a large population-based study. Clin Endocrinol, 84:627–631.

20.

World Health Organization 2011 Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. VMINS/Vitamin and Mineral Nutrition Information System (WHO/NMH/NHD/MNM/111). Available at: http://www.who.int/vmnis/indicators/haemoglobin/en/ (accessed April 23, 2018 ).

21.

Surks

, Ortiz

, Daniels

, Sawin

, Col

, Cobin

, Franklyn

, Hershman

, Burman

, Denke

, Gorman

, Cooper

, Weissman

. 2004. Subclinical thyroid disease: scientific review and guidelines for diagnosis and management. JAMA, 291:228–238.

22.

Helfand

. 2004. Screening for subclinical thyroid dysfunction in nonpregnant adults: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med, 140:128–141.

23.

Rodondi

, den Elzen

, Bauer

, Cappola

, Razvi

, Walsh

, Asvold

, Iervasi

, Imaizumi

, Collet

, Bremner

, Maisonneuve

, Sgarbi

, Khaw

, Vanderpump

, Newman

, Cornuz

, Franklyn

, Westendorp

, Vittinghoff

, Gussekloo

; Thyroid Studies Collaboration. 2010. Subclinical hypothyroidism and the risk of coronary heart disease and mortality. JAMA, 304:1365–1374.

24.

Vickers

, Altman

. 2001. Statistics notes: analysing controlled trials with baseline and follow up measurements. BMJ, 323:1123–1124.

25.

Vickers

. 2001. The use of percentage change from baseline as an outcome in a controlled trial is statistically inefficient: a simulation study. BMC Med Res Methodol, 1:6.

26.

Beard

, Borel

, Derr

. 1990. Impaired thermoregulation and thyroid function in iron-deficiency anemia. Am J Clin Nutr, 52:813–819.

27.

Srikrishna

, Girishbabu

, Ramesh

. 2015. Study of association of anemia in sub-clinical and overt hypothyroid patients. Res J Pharm Biol Chem Sci, 6:368–373.

28.

Bashir

, Bhat

, Farooq

, Majid

, Shoib

, Hamid

, Mattoo

, Rashid

, Bhat

, Wani

, Masood

. 2012. Comparison of hematological parameters in untreated and treated subclinical hypothyroidism and primary hypothyroidism patients. Med J Islam Repub Iran, 26:172–178.

29.

Blum

, Bauer

, Collet

, Fink

, Cappola

, da Costa

, Wirth

, Peeters

, Asvold

, den Elzen

, Luben

, Imaizumi

, Bremner

, Gogakos

, Eastell

, Kearney

, Strotmeyer

, Wallace

, Hoff

, Ceresini

, Rivadeneira

, Uitterlinden

, Stott

, Westendorp

, Khaw

, Langhammer

, Ferrucci

, Gussekloo

, Williams

, Walsh

, Juni

, Aujesky

, Rodondi

; Thyroid Studies Collaboration. 2015. Subclinical thyroid dysfunction and fracture risk: a meta-analysis. JAMA, 313:2055–2065.

30.

Park

, Nicolas

, Freisling

, Biessy

, Scalbert

, Romieu

, Chajes

, Chuang

, Ericson

, Wallstrom

, Ros

, Peeters

, Mattiello

, Palli

, Maria Huerta

, Amiano

, Halkjaer

, Dahm

, Trichopoulou

, Orfanos

, Teucher

, Feller

, Skeie

, Engeset

, Boutron-Ruault

, Clavel-Chapelon

, Crowe

, Khaw

, Vineis

, Slimani

. 2012. Comparison of standardised dietary folate intake across ten countries participating in the European Prospective Investigation into Cancer and Nutrition. Br J Nutr, 108:552–569.

31.

Olsen

, Halkjaer

, van Gils

, Buijsse

, Verhagen

, Jenab

, Boutron-Ruault

, Ericson

, Ocke

, Peeters

, Touvier

, Niravong

, Waaseth

, Skeie

, Khaw

, Travis

, Ferrari

, Sanchez

, Agudo

, Overvad

, Linseisen

, Weikert

, Sacerdote

, Evangelista

, Zylis

, Tsiotas

, Manjer

, van Guelpen

, Riboli

, Slimani

, Bingham

. 2009. Dietary intake of the water-soluble vitamins B1, B2, B6, B12 and C in 10 countries in the European Prospective Investigation into Cancer and Nutrition. Eur J Clin Nutr, 63:S122–149.

32.

de Bree

, van Dusseldorp

, Brouwer

, van het Hof

, Steegers-Theunissen

. 1997. Folate intake in Europe: recommended, actual and desired intake. Eur J Clin Nutr, 51:643–660.