Height and drinking status in relation to risk of anemia in rural adult healthy Japanese men: the Nagasaki Islands study

Abstract

Background: Several studies have reported that height is inversely associated with risk of cardiovascular disease but positively associated with cancer risk. On the other hand, evidence has been accumulating that anemia reflects poor health and increased vulnerability to poor outcomes in older persons. Moreover, alcohol consumption has also been reported to be associated with mortality. However, no studies have reported on a possible association between height and risk of anemia in relation to drinking status.

Methods: We conducted a cross-sectional study of 1287 men aged 40–89 years undergoing general health check-ups.

Results: Independent from classic cardiovascular risk factors, we found a significant inverse association between height and anemia for non-drinkers and a J-shaped association for drinkers. The multivariable odds ratio (ORs) of an increment of 1 SD (standard deviation) in height (6.68 cm) for anemia for non-drinkers was 0.59 (0.45–0.77). For drinkers, with the second quartile of height (Q2) as the reference group, the multivariable OR of anemia was 2.68(0.90–7.96) (p = 0.075) for the lowest height quartile (Q1), 2.73(0.92–8.08) for the third quartile (Q3) and 4.82(1.65–14.10) for the highest quartile (Q4) (p = 0.004).

Conclusion: Height was found to be associated with anemia for rural Japanese men and drinking status is likely to affect those associations.

• Anemia
• cross-sectional study
• drinking status
• height
• risk

Introduction

Previous prospective studies have shown that height is inversely associated with incidence of or mortality from cardiovascular disease [1–5], while other studies have confirmed that taller people are at greater risk of death from several organ-specific malignancies such as melanoma, cancer of the pancreas, prostate and colorectum [6–8]. Other studies have provided accumulating evidence that, although it was previously believed that a decline in hemoglobin levels might be a normal consequence of aging, anemia reflects poor health and increased vulnerability to poor outcomes in older persons [9,10]. However, no previous studies have reported on a possible association between height and anemia.

On the other hand, several studies have reported that alcohol consumption is also associated with mortality [11,12] and that hemoglobin levels of drinkers are higher than those of non-drinkers [13,14], even though alcohol consumption may induce hemorrhaging [15–18], which may lead to anemia. An investigation of associations between height and anemia should thus take drinking status into account.

However, women might not be suitable subjects for such an investigation because height is confounded by vertebral compression fractures, which are mainly caused by postmenopausal osteoporosis [19]. We, therefore, limited our investigation to Japanese men who participated in a general health check-up between 2005 and 2012.

Materials and methods

Participants

Written consent forms were available in Japanese to ensure comprehensive understanding of the study objectives, and informed consent was signed or thumb printed by the participants. This study was approved by the Ethics Committee for Use of Humans of Nagasaki University (project registration number 0501120073).

The survey population included 1520 men between 40 and 89 years old, all residents of the western rural community of the Goto Islands, who participated in this study between 2005 and 2012. A total of 86 individuals with missing data and 147 individuals with a history of cardiovascular disease were excluded, leaving 1287 men for enrolment in this study. The mean age of the study population was 65.2 years (±10.3 SD; range 40–89).

Data collection and laboratory measurements

Body weight and height were measured with an automatic body composition analyzer (BF-220; Tanita, Tokyo, Japan) at the time when blood was drawn.

Fasting blood samples were collected in an EDTA-2K tube and a siliconized tube. Samples from the EDTA-2K tube were used for measuring hemoglobin with the sodium lauryl surfate (SLS)-hemoglobin method. Mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH) and mean corpuscular hemoglobin concentration (MCHC) were calculated in the usual manner. Serum triglycerides, serum HDL-cholesterol, serum aspartate aminotransferase (AST), serum γ-glutamyltranspeptidase (GGT), HbA_1C and serum creatinine were measured with standard laboratory procedures.

Trained interviewers obtained information on smoking status, drinking status, medical history, use of antihypertensive agents and of medication for diabetes mellitus. HbA_1c (as defined by NGSP, the National Glycohemoglobin Standardization Program) was calculated with the following equation, which was recently proposed by a working group of the Japanese Diabetes Society (JDS): HbA_1c(NGSP) =HbA_1c(JDS) + 0.4% [20]. Presence of diabetes was defined as HbA_1c (NGSP) ≥6.5%, and/or initiation of glucose-lowering medication or insulin therapy [21].

Anemia was defined as hemoglobin <13 g/dL for men as previously reported [10,22–24], and normocytic normochromic anemia as hemoglobin <13 g/dL and MCV (80–98 fL), MCH (28–32 pg), MCHC (30–36%) were all within normal range.

Statistical analysis

Differences in age-adjusted mean values or prevalence of potential confounding factors by height quartile were calculated using covariance or generalized linear models, and logistic regression models were used for calculating odds ratios (OR) and 95% confidence intervals (CI) for determining the association of anemia and normocytic normochromic anemia with height. In addition, subjects were stratified by drinking status because alcohol consumption has been associated with mortality [11,12], and several studies also reported that hemoglobin levels of drinkers were higher than those of non-drinkers [13,14].

Three different approaches were used for making adjustments for confounding factors. First, the data were adjusted only for age (Model 1). Next we made further adjustment for actual birth year (Model 2). Finally, we included other potentially confounding factors, namely, smoking status (never smoker, former smoker and current smoker), alcohol consumption (non-drinker, current light to moderate drinker (1–6 times/week), current heavy drinker (every day)), body mass index (kg/m²),diabetes mellitus (no, yes), systolic blood pressure (mmHg), antihypertensive medication use (no, yes), serum triglycerides (mg/dL), serum HDL-cholesterol (mg/dL), serum AST (IU/L), serum GGT (IU/L), and serum creatinine (mg/dL) (Model 3).

All statistical analyses were performed with the SAS system for Windows (version 9.3; SAS Inc., Cary, NC). All p values for statistical tests were two-tailed, and values of <0.05 were regarded as statistically significant.

Results

Characteristics for study population

Of the 1287 men (631 non-drinkers, 656 drinkers), 147 (92 non-drinkers, 55 drinkers) were diagnosed with anemia.

Table 1 shows age-adjusted characteristics for this study population according to height. Current drinker and serum creatinine showed significant positive associations with height for total subjects. Stratification by drinking status of the results of the analysis showed significant positive associations for diastolic blood pressure, GGT and creatinine with height for non-drinkers, and a significant inverse association for HDL-cholesterol for drinkers.

Table 1. Age-adjusted mean values and proportions by quartiles of height.

	Total subjects					Non-drinker					Drinker
	Quartiles of height					Quartiles of height					Quartiles of height
	Q1 (low)	Q2	Q3	Q4 (high)	p	Q1 (low)	Q2	Q3	Q4 (high)	p	Q1 (low)	Q2	Q3	Q4 (high)	p
No. at risk	322	329	313	323		192	155	146	138		130	174	167	185
Age, years	70.0 ± 9.2	67.1 ± 8.9	63.9 ± 9.9	60.0 ± 10.0		70.5 ± 9.3	68.9 ± 8.8	64.3 ± 10.5	58.7 ± 10.0		69.1 ± 9.1	65.5 ± 8.8	63.5 ± 9.4	60.4 ± 10.0
Birth years, years	1941.7	1941.9	1941.9	1942.0	0.546	1940.8	1940.9	1941.2	1941.2	0.302	1942.6	1942.9	1942.6	1942.7	0.741
Sytolic blood pressure,mmHg	142	143	141	143	0.429	141	140	138	143	0.263	143	145	142	143	0.796
Diastolic blood pressure,mmHg	84	85	84	86	0.056	83	84	82	86	0.049	85	86	86	87	0.720
Antihypertensive medication use,%	28.1	27.3	26.3	27.4	0.959	26.5	23.1	20.7	27.7	0.439	30.9	31.0	31.0	27.1	0.798
Body mass index,kg/m^2	23.6	23.7	23.7	23.7	0.893	23.6	23.5	23.9	23.7	0.745	23.5	24.0	23.5	23.7	0.370
Diabetus melitus,%	8.6	9.8	11.2	9.9	0.761	8.5	10.4	13.1	12.6	0.559	8.7	9.1	9.7	8.0	0.957
Current smoker,%	26.9	24.6	24.6	23.4	0.795	24.3	18.3	22.3	24.2	0.543	29.8	30.0	26.8	23.5	0.520
Current drinker,%	41.7	53.4	53.0	55.8	0.003	−	−	−	−	−	−	−	−	–	–
Serum HDL-cholesterol,mg/dL	55	54	55	54	0.681	51	51	50	51	0.894	61	58	59	56	0.042
Serum triglycerides,mg/dL	123	125	127	129	0.877	123	122	124	133	0.653	122	128	130	126	0.864
Serum aspartate aminotransferase (AST),IU/L	25	26	25	24	0.089	24	26	25	23	0.312	26	26	26	24	0.239
Serum γ-glutamyltranspeptidase (GGT), IU/L	41	45	47	43	0.516	28	33	31	40	0.028	58	55	62	47	0.183
Serum creatinine,mg/dL	0.84	0.90	0.91	0.93	<0.001	0.82	0.92	0.90	0.94	<0.001	0.87	0.88	0.91	0.91	0.281

Age: mean ± standard deviation. p: p factor. Height levels for quartile were <158.8 cm, 158.8–163.0 cm, and 163.1–167.9 cm.

To identify differences in various associations according to drinking status, we used the age-adjusted characteristics by drinking status shown in Table 2. Compared with non-drinkers, drinkers showed significantly higher hemoglobin, MCV, MCH, MCHC, systolic blood pressure, diastolic blood pressure, HDL-cholesterol, AST, GGT levels and use of antihypertensive medication.

Table 2. Age-adjusted mean values and proportions by drinking status.

	Non-drinker	Drinker	p
Median value of height, cm	162	164
No.at risk	631	656
Age, years	66.1 ± 10.6	64.3 ± 9.8
Birth years, years	1942.0	1941.8	0.228
Hemoglobin (Hb), g/dL	14.3	14.6	<0.001
Mean Corpuscular Volume (MCV), fL	94.8	96.4	<0.001
Mean Corpuscular Hemoglobin (MCH), pg	31.0	31.7	<0.001
Mean Corpuscular Hemoglobin Concentration (MCHC), %	32.7	32.9	0.001
Sytolic blood pressure,mmHg	140	144	0.002
Diastolic blood pressure,mmHg	84	86	<0.001
Antihypertensive medication use, %	23.3	31.1	0.001
Body mass index,kg/m^2	23.7	23.6	0.752
Diabetus melitus,%	10.7	9.1	0.327
Current smoker,%	23.0	26.7	0.126
Serum HDL-cholesterol,mg/dL	51	58	<0.001
Serum triglycerides,mg/dL	127	126	0.812
Serum aspartate aminotransferase (AST),IU/L	24	25	0.048
Serum γ-glutamyltranspeptidase (GGT), IU/L	33	55	<0.001
Serum creatinine,mg/dL	0.89	0.90	0.503

Age: mean ± standard deviation. p: p factor.

Height levels in relation to drinking status

We used the general linear model for further investigations to determine the relationship between height and drinking status and found that height was also significantly associated with drinking status. The age-adjusted values for height (minimum mean square) and age (mean ± SD) were: 162.1 cm (66.5 ± 10.5) for non-drinkers, 163.7 cm (63.9 ± 10.3) for current light-to-moderate drinkers, and 163.8 cm (64.5 ± 9.5) for current heavy drinkers (p for trend <0.001). Since drinking status becomes a relevant variable only after childhood, height may affect drinking status in later life [25].

Association between height and anemia

The association between height and anemia for total subjects was U-shaped, while height and normocytic normochromic anemia showed an inverse linear association. And height of non-drinkers was inversely associated with both anemia and normocytic normochromic anemia. Drinkers showed a J-shaped association between height and anemia. However, an analysis of the association between height and normocytic normochromic anemia showed a weaker association (Table 3).

Table 3. Odds ratios (ORs) and 95% confidence intervals (CIs) for anemia in relation to height stratified by drinking status.

	Height quartile
	Q1 (low)	Q2	Q3	Q4 (high)	p for trend	1 SD increment in height (6.68 cm)
Anemia
Total subjects
No. at risk	322	329	313	323
No. of cases (percentage)	62 (19.3)	35 (10.6)	25 (8.0)	25 (7.7)
Model 1	1.00	0.61 (0.38–0.96)	0.54 (0.32–0.90)	0.71 (0.42–1.21)	–	–
Model 2	1.00	0.62 (0.39–0.98)	0.55 (0.33–0.92)	0.72 (0.42–1.22)	–	–
Model 3	1.00	0.58 (0.35–0.94)	0.56 (0.33–0.96)	0.71 (0.40–1.26)	–	–
Non–drinkers
No. at risk	192	155	146	138
No. of cases (percentage)	45 (23.4)	29 (18.7)	12 (8.2)	6 (4.3)
Model 1	1.00	0.82 (0.48–1.40)	0.39 (0.19–0.78)	0.27 (0.11–0.69)	<0.001	0.57 (0.45–0.74)
Model 2	1.00	0.84 (0.49–1.43)	0.40 (0.20–0.81)	0.28 (0.11–0.71)	<0.001	0.59 (0.46–0.76)
Model 3	1.00	0.76 (0.43–1.35)	0.41 (0.20–0.86)	0.27 (0.11–0.72)	0.001	0.59 (0.45–0.77)
Drinkers
No. at risk	130	174	167	185
No. of cases (percentage)	17 (13.1)	6 (3.4)	13 (7.8)	19 (10.3)
Model 1	3.25 (1.21–8.72)	1.00	2.83 (1.03–7.75)	4.90 (1.85–13.00)	0.001*	1.71 (1.08–2.70)*
Model 2	3.13 (1.16–8.45)	1.00	2.79 (1.02–7.67)	4.84 (1.82–12.84)	0.001*	1.70 (1.08–2.69)*
Model 3	2.68 (0.90–7.96)	1.00	2.73 (0.92–8.08)	4.82 (1.65–14.10)	0.004*	1.59 (0.95–2.65)*
Normocytic normochromic anemia
Total subjects
No. at risk	322	329	313	323
No. of cases (percentage)	37 (11.5)	17 (5.2)	11 (3.5)	7 (2.2)
Model 1	1.00	0.47 (0.26–0.28)	0.36 (0.18–0.73)	0.26 (0.11–0.62)	<0.001	0.55 (0.42–0.71)
Model 2	1.00	0.48 (0.26–0.89)	0.37 (0.18–0.74)	0.27 (0.11–0.63)	<0.001	0.56 (0.43–0.72)
Model 3	1.00	0.48 (0.26–0.90)	0.40 (0.19–0.84)	0.29 (0.12–0.70)	0.001	0.60 (0.46–0.79)
Non–drinkers
No. at risk	192	155	146	138
No. of cases (percentage)	28 (14.6)	14 (9.0)	7 (4.8)	3 (2.2)
Model 1	1.00	0.60 (0.30–1.18)	0.32 (0.13–0.78)	0.16 (0.04–0.55)	<0.001	0.52 (0.39–0.71)
Model 2	1.00	0.61 (0.31–1.21)	0.34 (0.14–0.82)	0.16 (0.05–0.58)	<0.001	0.54 (0.40–0.73)
Model 3	1.00	0.55 (0.27–1.13)	0.36 (0.14–0.89)	0.17 (0.05–0.61)	<0.001	0.58 (0.42–0.79)
Drinkers
No. at risk	130	174	167	185
No. of cases (percentage)	9 (6.9)	3 (1.7)	4 (2.4)	4 (2.2)
Model 1	3.11 (0.80–12.14)	1.00	1.62 (0.35–7.44)	1.79 (0.39–8.29)	0.457*	1.32 (0.58–2.97)*
Model 2	3.10 (0.79–12.16)	1.00	1.62 (0.35–7.45)	1.78 (0.38–8.32)	0.460*	1.32 (0.58–2.98)*
Model 3	2.91 (0.69–12.33)	1.00	1.79 (0.36–8.89)	1.83 (0.35–9.58)	0.476*	1.22 (0.50–2.98)*

Model 1: Adjusted for age group. Model 2: Adjusted further for age group and actual birth year. Model 3: Model 2 + systolic blood pressure, antihypertensive medication use, body mass index, smoking, alcohol intake, diabetes mellitus, serum triglycerides, serum HDL-cholesterol, serum aspartate aminotransferase (AST), serum γ-glutamyltranspeptidase (GGT) and serum creatinine. Height levels for quartiles were <158.8 cm, 158.8–163.0 cm, 163.1–167.9 cm, and >167.9 cm.

*Analyses were performed for height  ≥  158.8 cm. Anemia was defined as hemoglobin <13 g/dL and normocytic normochromic anemia as hemoglobin <13 g/dL, and MCV (80–98 fL), MCH (28–32 pg), MCHC (30–36%) were all within normal range.

We also identified a significant effect on anemia of interaction between height levels (158.8–163.0 cm, 163.1–167.9 cm, and  >167.9 cm) and drinking status (non-drinkers or drinkers) for participants among height ≥158.8 cm. The multivariable-adjusted p value of this effect of interaction on anemia was <0.001.

Discussion

A major finding of the present study was that height was associated with anemia for rural healthy Japanese men who were receiving an annual health check, and drinking status is likely to affect those associations. For non-drinkers, a significant inverse association between height and anemia was detected, whereas the association was J-shaped for drinkers.

Previous studies had found that height was inversely associated with risk of cardiovascular disease [1–5] and positively associated with risk of cancer [6–8].

A prospective cohort study with a 11.2-year follow-up of 5888 community-dwelling men and women aged ≥65 showed that anemia was independently associated with increased mortality [10]. However no study had been published on the possible association between height and anemia. Such an analysis should also take drinking status into account because other studies have reported that alcohol consumption is also associated with mortality [11,12].

A previous Japanese study reported that alcohol consumption leads to an increase in HDL-cholesterol, TG, AST and GGT levels in alcohol dehydrogenase 2 (ALDH2)-sufficient individuals and in AST and blood glucose levels in ALDH2-inefficient individuals [26]. Another study detected a positive association of alcohol consumption with blood pressure and serum creatinine [27]. In our study, drinkers showed significantly higher systolic blood pressure, diastolic blood pressure, HDL-cholesterol, serum AST, and serum GGT levels, as well as use of antihypertensive medication than did non-drinkers.

The mechanisms for short stature constituting a risk of anemia have not yet been elucidated. Side population hematopoietic stem cells in bone marrow decrease as individuals age [28,29], and this decline may be associated with an increase in the frequency of anemia and other hematopoietic disorders that are seen in the elderly [30]. Neumann reported as early as 1882 that active hematopoietic (red) bone marrow declined with age and changed into fatty (yellow) marrow from the periphery towards the axial skeleton. This has been referred to as Neumann’s law [31]. Since the volume of bone marrow is smaller in subjects with shorter than in those with taller stature, any reduction in the activity of hematopoietic bone marrow could be of crucial importance for individuals with shorter stature.

Further, Sonal et al. reported that increased bone marrow fat may be associated to a decrease with aging in the number of side population stem cells and levels of insulin-like growth factor-1 (IGF- 1) [32]. IGF-1 levels may also be an important determinant factor for anemia because they are reportedly positively associated with height in childhood [33] and inversely associated with risk of anemia for non-diabetic adults [34].

Birth weight and height correlate strongly with adult height [35], and low birth weight is known to be associated with altered renal shape, reduced renal volume, and fewer nephrons [36]. Thus, short stature may increase the chance of suffering kidney malfunction and hypertension in later life [37], which may then lead to higher risk of secondary anemia induced by renal dysfunction. In our study, however, the association height and anemia were found to be associated even after adjustment for serum creatinine levels.

Previous studies have reported that height and risk of stroke are inversely associated [1–5]. And our previous retrospective study reported that the inverse association between height and risk of stroke is limited to participants with lower BMI (<23 kg/m²) [2]. On the other hand, height was found to be inversely associated with carotid atherosclerosis for overweight (BMI ≥ 25 kg/m²) but not for non-overweight (BMI < 25 kg/m²) [38]. Therefore atherosclerosis should not be the main risk of stroke for shorter. Another previous study reported an association between a reduced number of circulating endothelial progenitor cell (CD34 + cells) and an increased number of infractions >5 mm in diameter, but no such association with atherosclerosis in carotid arteries of patients with ischemic stroke [39]. Since hematopoietic stem cell derived from bone marrow play a major role in vascular homeostasis, anemia might indicate the impaired vascular homeostasis by indicating reduced activity of bone marrow [40–42]. Our previous studies with non-anemic Japanese that reported hemoglobin is positively associated with atherosclerosis [22] and hypertension [23] among participants with non-overweight might partly explain this mechanism.

The risk of anemia for male taller drinkers has also not yet been elucidated. It was previously reported that hemoglobin levels of drinkers were higher than those of non-drinkers [13,14], and drinkers in our study showed slightly but significantly higher hemoglobin levels than did non-drinkers. Alcohol consumption thus may accelerate the production of hemoglobin. Since taller stature constitutes a risk of anemia for drinkers but not for non-drinkers, the main cause of anemia in taller subjects may feature a different mechanism from that of reduction in hemoglobin production.

More recent studies have confirmed that taller stature represents a higher risk of death from several malignancies [6–8]. On the other hand, angiogenesis is required for invasive tumor growth and metastasis and constitutes an important target for control of cancer progression [43] and such angiogenic vessels are known to be vulnerable and to rupture easily [44,45]. In addition to those mechanisms, alcohol induces reductions in platelet aggregation [15] and plasma fibrinogen levels [16], as well as enhancement of fibrinolysis [17], and heavy alcohol consumption increases not only blood pressure levels, but also blood pressure variability including morning surge [18]. Taller male drinkers thus may be at higher risk of hemorrhagic anemia. Our analysis of the associations between normocytic normochromic anemia and height for drinkers showed that such associations became non-significant, which seems to support this notion because hemorrhagic anemia may cause microcytic hypochromic anemia.

Possible limitations of this study warrant consideration. First, although significant positive associations between height and anemia were observed among taller drinker (height  ≥  158.8 cm) in our study, a wide confidence interval was resulted from the small number of anemia cases among drinkers. Second, since no medical data on history of cancer were available, we could not evaluate the effect of height on cancer. Since a previous study reported that the mean height for the age range 30–39 was lower for men born between 1936 and 1945 (163.8 cm) than for those born between 1961 to 1970 (170.6 cm) [5], this difference might be due to a change in diet and food consumption from during to after the second Word War. Since Q1 was the oldest age group, it was not unexpected that of anemia was most prevalent in Q1 and less prevalent in Q4, which was the youngest age group. However, our analysis further adjusted for birth year (Model 2) showed essentially the same association as for the model without adjustment for birth year (Model 1). Furthermore, the associations between height and anemia were the same even when we limited the analysis to older subjects who was born at or before 1945; the fully-adjusted OR of an increment of 1SD in height (6.40 cm) for anemia for non-drinkers was 0.60(0.45–0.80). With the second quartile of height (Q2: 157.4–161.5 cm) for drinkers as the reference group, the fully-adjusted OR of anemia was 3.80(1.08–13.43) (p = 0.038) for the lowest height quartile (Q1: < 157.4 cm), 3.09(0.85–11.21) for the third quartile (Q3: 161.6–165.5 cm) and 3.67(1.04–13.00) for the highest quartile (Q4: > 165.5 cm) (p = 0.050). We also identified a significant effect on anemia of interaction between height levels (157.4–161.5 cm, 161.6–165.5 cm, and >165.5 cm) and drinking status (non-drinkers or drinkers) for participants among height ≥157.4 cm. The fully-adjusted p value of this effect of interaction on anemia was 0.013.

Because creatinine clearance data were not available and estimated glomerular filtration rate is not an effective tool for evaluating kidney function for a comparison of associations with various body heights [2,38,46], we could not perform an analysis adjusted for exact renal function. However, our study showed that associations between height and anemia remained significant even after adjustment for serum creatinine. And because data for serum iron and vitamin concentrations were not available either, we could not evaluate the influence of those factors. However, inverse linear associations between height and normocytic normochromic anemia were also observed for total and non-drinking men. Finally, because this was a cross-sectional study, we could not establish any causal relationships.

Conclusion

In conclusion, our findings suggest that height was associated with risk of anemia for rural healthy Japanese men who were receiving an annual health check, and that drinking status was likely to affect this association.

Declaration of interest

This work was supported financially by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (No. 22370090).

The authors have no conflicts of interest with this paper.