Correlation of sleep-disordered breathing with anemia and microcytosis in premenopausal females: the Nagahama study

Abstract

Study Objectives:

While nocturnal hypoxia caused by sleep-disordered breathing (SDB) can result in polycythemia, the association between erythrocyte values and SDB may vary by sex and menopausal status. This study aimed to evaluate the correlation of erythrocyte variables with SDB in a large cohort categorized by sex and menopausal status from the general population.

Methods:

Sleep duration and SDB were assessed in community residents using a wearable actigraph and pulse oximeter. The actigraph-adjusted 3% oxygen saturation index was calculated by correcting the time measured by the pulse oximeter for sleep duration computed by the actigraph. SDB severity was defined by actigraph-adjusted 3% oxygen saturation index as normal (< 5 events/h), mild (5 to < 15 events/h), and moderate to severe (≥ 15 events/h).

Results:

Data were obtained from 6,836 participants (premenopausal females, 1,580; postmenopausal females, 3,058; and males, 2,198). Multivariate analysis revealed that actigraph-adjusted 3% oxygen saturation index was negatively correlated with hemoglobin level in premenopausal females (β = −0.13, P < .01), but no significant correlations were found in postmenopausal females and males. In addition, the presence of SDB was a risk factor for microcytic anemia (hemoglobin < 12 g/dL and mean corpuscular volume < 80 fL) in premenopausal females (odds ratio and 95% confidence interval: mild SDB, 1.93 [1.31–2.84], P < .01; moderate to severe SDB, 9.02 [3.02–26.88], P < .01).

Conclusions:

SDB was associated with microcytic anemia in premenopausal females. When SDB is diagnosed in a premenopausal female, we may need to evaluate her erythrocytic parameters.

Citation:

Murase K, Matsumoto T, Tabara Y, et al. Correlation of sleep-disordered breathing with anemia and microcytosis in premenopausal females: the Nagahama study. J Clin Sleep Med. 2025;21(7):1155–1164.

BRIEF SUMMARY

Current Knowledge/Study Rationale: While nocturnal hypoxia caused by sleep-disordered breathing (SDB) can result in polycythemia, the association of erythrocyte values with SDB can vary by sex and menopausal status. We performed the stratified analyses for their associations in a large population-based cohort.

Study Impact: No significant associations were found between SDB and erythrocyte values in males and postmenopausal females, but in premenopausal females, mild SDB as well as moderate to severe SDB were associated with microcytic anemia. This finding suggests a link between SDB and heavy menstrual bleeding and highlights the need to evaluate erythrocytic parameters when SDB is diagnosed in premenopausal females.

INTRODUCTION

Nocturnal hypoxic burden caused by sleep-disordered breathing (SDB) possibly stimulates erythropoiesis,^1–3 which results in polycythemia. Routine evaluation of SDB in individuals with polycythemia has been recommended.⁴ However, previous studies examining the correlation of erythrocyte values with SDB reported conflicting results.^5–9 The relatively small number of participants enrolled in these previous studies may have contributed to these discrepant results. In addition, while the association of erythrocyte values and SDB can vary by sex and menopausal status, this issue has not been assessed in a large cohort.^10–14 Even if there is a positive correlation between the severity of SDB and hemoglobin (Hb) levels in postmenopausal females and males, this relation may not be found in premenopausal females, who are more likely to have anemia due to menstrual bleeding.¹³ Here, we hypothesized that the significant association of SDB with erythrocyte values were found in postmenopausal females and males but not in premenopausal females and analyzed their correlation in the Nagahama study cohort, which is a large population-based cohort. The analyses were stratified by sex and menopausal status.

METHODS

Participants

Between 2013 and 2016, we recruited participants from residents in Nagahama City, located in the center of Japan, where there were approximately 125,000 residents. To be eligible, individuals had to be (1) 34–80 years of age, and (2) able to participate independently in health assessments. Our study protocol received approval from the ethics committee of Kyoto University Graduate School of Medicine and the Nagahama Municipal Review Board (No. G0278). All participants provided written informed consent.

We enrolled 9,850 individuals and collected information on their clinical background, which included smoking history, drinking habits, medications, and menstrual status through a structured questionnaire. Individuals who consumed alcohol more than 4 days a week were defined as frequent drinkers. We excluded participants who were pregnant, undergoing dialysis, receiving treatment for SDB or active malignancy, or were incapable of completing the questionnaire.

Erythrocyte measures

Venous blood samples were drawn during a health examination. Hb, red blood cell counts and hematocrit were measured using an automated analyzer (Sysmex XE-2100, Sysmex, Kobe, Japan). Mean corpuscular volume (MCV) was calculated as MCV (fL) = hematocrit (%) × 10/red blood cell (10¹²/L). Anemia was defined as Hb < 13 and Hb < 12 (g/dL) in males and females, respectively.¹⁵ Polycythemia was defined as Hb > 16.5 and > 16 g/dL in males and females, respectively.¹⁶ In addition, MCV was categorized as follows: microcytosis, < 80; normal, 80 to < 100; and macrocytosis, ≥ 100 (fL).¹⁷

Sleep duration and SDB

Participants were asked to measure their sleep duration and severity of SDB at home for a week with wearable devices immediately after they visited the examination site. Well trained research staff instructed participants to wear an actigraph (Actiwatch 2 or Actiwatch Spectrum Plus; Philips Respironics, Murrysville, PA) for a week (days 1–7) and a pulse oximeter (PULSOX-Me300; Konica Minolta Inc., Tokyo, Japan) for the first 4 nights (days 1–4). The participants were also requested to complete a sleep diary during the same period. Details of the data analysis of actigraphy and pulse oximetry were shown in our previous studies.^12,18,19 Briefly, in the process of editing actigraphy data, expert personnel manually identified and entered bed-entry and bed-exit times, consulting sleep diaries as a reference. Subsequently, specialized software automatically determined sleep initiation and wake-up times, along with intervals of wakefulness occurring after the onset of sleep, using the default algorithm provided by the manufacturer. Total sleep duration (from onset of sleep to wake-up) and the actual sleep duration (total sleep duration minus the time spent awake after sleep onset) were computed. To be eligible for analysis, data from ≥ 5 nights, including at least 1 weekend night, were required, and the values were subsequently averaged. In the evaluation of SDB, the 3% oxygen desaturation index (3%ODI) and the cumulative percentage of time during sleep with oxygen saturation below 90% (CT90%) were determined.

To ensure better comparability with results obtained by polysomnography, the actigraphy-modified 3%ODI (Acti-3%ODI) and actigraphy-modified CT90% (Acti-CT90%) were computed by adjusting the time measured by pulse oximetry with the actual sleep duration computed by the actigraph.¹² Our previous study, which enrolled 32 patients with SDB from a different cohort, showed that the correlation between Acti-3%ODI and the apnea-hypopnea index determined by polysomnography (r = 0.99, P < .001; apnea-hypopnea index = Acti-3%ODI * 1.04 + 1.45) was stronger than the correlation of the unmodified 3%ODI without adjustment of actigraphy (r = 0.92, P < .001; apnea-hypopnea index = 3%ODI * 1.27 + 2.06).¹⁸ Data from ≥ 2 nights were included in the analysis, and the values were averaged. The severity of SDB was defined based on Acti-3%ODI as follows: normal < 5, mild 5 to < 15, and moderate to severe ≥ 15 events/h.

Confounding variables

Serum creatinine was measured using an enzymatic method (Espa Cre Liquid II, NIPRO Corp., Tokyo, Japan) to estimate renal function, which affects erythropoiesis. Estimated glomerular filtration rate was computed by the equation proposed by the Japanese Society of Nephrology: estimated glomerular filtration rate = 194 × serum creatinine^−1.094 × age^−0.287 (× 0.739 for females).²⁰ In addition, since chronic inflammation causes anemia, C reactive protein was analyzed as a confounding factor with the Behring Nephelometer using anti-C reactive protein antibody latex particles.

Because menstrual bleeding, if occurring, may have an impact on Hb levels, the questionnaire asked premenopausal females to indicate whether they were in the menstrual phase at the time of testing. Furthermore, a history of childbirth and collagen diseases was also assessed using yes/no questions as possible factors influencing Hb levels.^21,22

Statistical analysis

Because the erythrocytic indices and the prevalence of SDB can exhibit notable variations depending on sex and menopausal status, we conducted analyses to investigate the correlations between SDB and erythrocyte values separately in premenopausal females, postmenopausal females, and males.^10–14 First, in each of these 3 cohorts, we compared Hb levels among groups categorized by the severity of SDB. Second, we performed multivariate regression analysis to explore whether the relationship between SDB indices and Hb was independent from confounding factors. Third, when the association of SDB indices with Hb remained significant in this multivariate analysis, we compared the prevalence of anemia and polycythemia among groups categorized by SDB severity and conducted multivariate logistic regression analysis to assess whether SDB was a risk factor for anemia or polycythemia.

We used analysis of variance or Kruskal-Wallis tests to compare continuous variables among groups and employed the chi-square test for categorical variables. Trend testing was performed using the Cochrane-Armitage trend test. In case of skewed distribution of variables in linear regression analysis, logarithmic transformations were applied. However, since Acti-CT90% had heavily skewed distribution, it was dichotomized by its 75th percentile of all participants (Acti-CT90% ≥ 0.8%) as in our previous study.¹⁹ We defined a 2-tailed P value of < .05 to indicate statistical significance. All statistical analyses were conducted using JMP Pro 17 (SAS Institute Inc., NC, USA).

RESULTS

Of 9,850 participants, 6,836 (premenopausal females, 1,580; postmenopausal females, 3,058; and males, 2,198) met the inclusion criteria and were enrolled in the analysis (Figure 1). Background information on the participants is shown in Table 1. Hb levels were 12.5 ± 1.3, 13.0 ± 0.9, and 14.6 ± 1.2 g/dL in premenopausal and postmenopausal females and males, respectively. No premenopausal females had polycythemia and only 0.1% (3 participants) of postmenopausal females had polycythemia. On the other hand, anemia was present in 26.0% of premenopausal females, 10.5% of postmenopausal females, and 7.5% of males. The prevalence of moderate to severe SDB was highest in males (premenopausal females vs postmenopausal females vs males, 1.7 vs 9.6 vs 23.5%).

Figure 1. Flow chart of study participants selection.

Table 1.

Background of analyzed participants.

	Premenopausal	Postmenopausal	Males
N	1,580	3,058	2,198
Age, years	44 ± 6	63 ± 8	60 ± 12
Current-smoker, %	5.8	2.1	22.0
Frequent-drinker, %	14.4	11.1	50.1
Body mass index, kg/m2	21.2 ± 3.2	22.0 ± 3.2	23.2 ± 3.0
Menstrual phase at the time of examination, %	14.8	—	—
History of childbirth, %	87.2	94.1	—
History of collagen diseases, %	1.1	3.1	1.2
eGFR, mL/minute/1.73 m2	85.7 ± 13.5	74.5 ± 13.2	74.0 ± 14.0
C reactive protein, ng/mL	208 (117–493)	349 (177–701)	420 (212–862)
Hemoglobin, g/dL	12.5 ± 1.3	13.0 ± 0.9	14.6 ± 1.2
Anemia, %	26.0	10.5	7.5
Polycythemia, %	0.0	0.1	3.9
Hematocrit (%)	38.1 ± 3.3	39.4 ± 2.6	43.3 ± 3.1
Red blood cell (1012/L)	4.28 ± 0.31	4.26 ± 0.33	4.62 ± 0.41
Mean corpuscular volume, fL	89.3 ± 6.9	92.8 ± 4.0	93.8 ± 4.6
Microcytosis, %	10.3	0.8	0.7
Macrocytosis, %	2.0	2.8	7.7
Sleep duration, minutes	352 ± 57	364 ± 54	361 ± 62
Acti-3%ODI, per hour	3.5 (2.4–5.0)	6.0 (3.8–9.6)	9.0 (5.7–14.4)
Severity of sleep-disordered breathing, %
Normal	74.6	39.6	19.1
Mild	23.7	50.8	57.4
Moderate to severe	1.7	9.6	23.5
Acti-CT90%, %	0.1 (0.0–0.2)	0.2 (0.1–0.7)	0.4 (0.1–1.5)
Acti-CT90% ≥ 0.8%, %	6.6	22.2	38.0

Values are expressed as mean ± standard deviation or median (first–third quartile). Consuming alcohol more than 4 days per week was defined as frequent drinking. Anemia was defined as blood hemoglobin level < 13 and < 12 g/dL in males and females, respectively. Polycythemia was defined as hemoglobin level ≥ 16.5 and ≥ 16 g/dL in males and females, respectively. Microcytosis and macrocytosis were defined as mean corpuscular volume < 80 and > 100 fL, respectively. Acti-CT90% = actigraphy-modified cumulative percentage of sleep time with oxygen saturation < 90%, Acti-3%ODI = actigraphy-modified 3% oxygen desaturation index, eGFR = estimated glomerular filtration rate.

Table 2 shows the background of groups categorized for severity of SDB. Age, body mass index, and C reactive protein increased as the severity of SDB increased in all 3 cohorts. Figure 2A shows the Hb level of groups categorized by the severity of SDB. In males, the Hb level did not differ significantly among the groups with normal, mild, and moderate to severe SDB (14.7 ± 1.1 vs 14.6 ± 1.1 vs 14.6 ± 1.3 g/dL, P = .18). In postmenopausal females, the group with mild SDB had the lowest Hb level (normal vs mild vs moderate to severe SDB: 13.1 ± 0.9 vs 13.0 ± 0.9 vs 13.1 ± 1.0 g/dL, P = .02). In premenopausal females, Hb decreased as the severity of SDB increased with significant differences among the groups categorized by the severity of SDB (normal vs mild vs moderate to severe SDB: 12.5 ± 1.2 vs 12.3 ± 1.6 vs 12.2 ± 1.6 g/dL, P = .01) (Figure 2A). Multivariate regression analysis including confounding factors revealed that Acti-3%ODI was significantly and negatively correlated with Hb in premenopausal females (β = −0.13, P < .01) but not in postmenopausal females and males (postmenopausal females, β = −0.04, P = .07; males, β = −0.05, P = .07) (Table 3). On the other hand, Acti-CT90% ≥ 0.8% showed no significant correlation with Hb in all 3 cohorts categorized by sex and menopausal status (premenopausal females, β = 0.03, P = .34; postmenopausal females, β = 0.02, P = .25; males, β = 0.03, P = .21) (Table 3).

Table 2.

Backgrounds of participants categorized by sex, menopausal status, and the severity of sleep-disordered breathing.

	Severity of SDB			P
	Normal	Mild	Moderate to Severe
Premenopausal females
n	1,178	375	27
Age, years	43 ± 5	44 ± 6	49 ± 10	< .01
Current smoker, %	6.3	4.3	3.7	.29
Frequent drinker, %	14.9	13.6	3.7	.15
Body mass index, kg/m2	20.6 ± 2.6	22.6 ± 3.8	27.1 ± 4.3	< .01
Menstrual phase at the time of examination, %	15.5	13.3	3.7	.09
History of childbirth, %	86.8	89.1	77.8	.20
History of collagen diseases, %	1.1	1.3	0.0	.69
C reactive protein, ng/mL	190 (107–409)	281 (137–699)	842 (504–2,050)	< .01
eGFR, mL/minute/1.73 m2	86.1 ± 13.5	84.3 ± 12.9	88.5 ± 17.7	.046
Sleep duration, minutes	355 ± 56	345 ± 59	334 ± 68	< .01
Acti-3%ODI, per hour	2.9 (2.2–3.8)	6.6 (5.7–8.4)	17.8 (15.9–22.7)	< .01
Acti-CT90%, %	0.0 (0.0–0.1)	0.2 (0.1–0.4)	1.8 (0.7–3.0)	< .01
Acti-CT90% > 0.8%, %	2.9	13.3	74.1	< .01
Postmenopausal females
n	1,212	1,554	292
Age, years	61 ± 7	65 ± 7	67 ± 7	< .01
Current smoker, %	2.9	1.6	1.4	.047
Frequent drinker, %	10.6	11.5	10.6	.71
Body mass index, kg/m2	20.8 ± 2.5	22.4 ± 3.1	24.7 ± 3.8	< .01
History of childbirth, %	92.1	95.6	94.2	< .01
History of collagen diseases, %	3.1	3.2	2.7	.91
C reactive protein, ng/mL	271 (153–536)	391 (196–790)	539 (291–1,088)	< .01
eGFR, mL/minute/1.73 m2	75.0 ± 12.2	74.6 ± 13.8	71.5 ± 13.2	< .01
Sleep duration, minutes	366 ± 54	363 ± 53	354 ± 61	< .01
Acti-3%ODI, per hour	3.4 (2.6–4.2)	7.9 (6.2–10.2)	18.8 (16.5–23.6)	< .01
Acti-CT90%, %	0.1 (0.0–0.1)	0.4 (0.2–0.8)	2.4 (1.3–4.2)	< .01
Acti-CT90% > 0.8%, %	2.8	25.3	86.6	< .01
Males
n	420	1,262	516
Age, years	55 ± 12	60 ± 12	64 ± 11	< .01
Current smoker, %	31.2	22.5	13.4	< .01
Frequent drinker, %	43.3	51.3	52.9	< .01
Body mass index, kg/m2	21.7 ± 2.3	23.1 ± 2.8	24.9 ± 3.4	< .01
History of collagen diseases, %	1.9	0.8	1.7	.10
C reactive protein, ng/mL	285 (160–630)	212 (126–406)	589 (287–1,218)	< .01
eGFR, mL/minute/1.73 m2	77.5 ± 13.0	73.9 ± 14.0	71.2 ± 14.1	< .01
Sleep duration, minutes	371 ± 63	363 ± 59	348 ± 66	< .01
Acti-3%ODI, per hour	3.8 (3.0–4.4)	8.6 (6.7–11.1)	21.4 (17.5–27.7)	< .01
Acti-CT90%, %	0.0 (0.0–0.1)	0.4 (0.1–1.0)	2.9 (1.3–5.9)	< .01
Acti-CT90% > 0.8%, %	3.8	30.2	85.1	< .01

Values are expressed as mean ± standard deviation or median (1st–3rd quartile). Consuming alcohol more than 4 days per week was defined as “frequent drinking.” Acti-CT90% = actigraphy-modified cumulative percentage of sleep time with oxygen saturation < 90%, Acti-3%ODI = actigraphy-modified 3% oxygen desaturation index, eGFR = estimated glomerular filtration rate.

Figure 2. Mean hemoglobin level and percentages of participants with anemia and polycythemia in groups categorized by the severity of SDB and sex/menopausal status.

(A) Mean hemoglobin level in groups categorized by the severity of SDB in premenopausal females, postmenopausal females and males. Black bar and white line in the bar represent mean value and standard deviation in each group. The numbers in parentheses indicate the number of participants in each group. (B) Percentages of participants with anemia or polycythemia in groups categorized by the severity of SDB in premenopausal females, postmenopausal females and males. Black, gray, and white bars represent percentage of participants with anemia, normal hemoglobin level, and polycythemia, respectively. Numbers in each bar represent the percentage of participants. SDB = sleep-disordered breathing.

Table 3.

Multivariate analysis to determine contributing factors to hemoglobin level.

	Premenopausal		Postmenopausal		Males
	β	P	Β	P	β	P
Age, years	−0.01	.68	−0.16	< .01	−0.29	< .01
Current smoker, %	0.06	.01	0.04	.03	0.10	< .01
Frequent drinker, %	0.02	.36	0.03	.13	0.06	< .01
Body mass index, kg/m2	0.05	.09	0.18	< .01	0.22	< .01
Menstrual phase at the time of examination	−0.10	< .01	—	—	—	—
History of childbirth	−0.11	< .01	0.02	.16	—	—
History of collagen diseases	−0.05	.06	−0.07	< .01	−0.01	.83
eGFR, mL/minute/1.73 m2	−0.10	< .01	0.01	.98	0.05	.02
C reactive protein, ng/mL (log-transformed)	0.12	< .01	−0.01	.78	0.03	.14
Sleep duration, minutes	0.06	.02	0.06	< .01	0.05	.01
Acti-3%ODI, per hour (log-transformed)	−0.13	< .01	−0.04	.07	−0.05	.07
Acti-CT90% > 0.8%	0.03	.34	0.02	.25	0.03	.21

Acti-CT90% = actigraphy-modified cumulative percentage of sleep time with oxygen saturation < 90%, Acti-3%ODI = actigraphy-modified 3% oxygen desaturation index, eGFR = estimated glomerular filtration rate.

The prevalence of participants with anemia and polycythemia are shown in Figure 2B. In premenopausal females, the prevalence of anemia increased as the severity of SDB increased (normal 23.8%, mild SDB 32.0% and moderate to severe SDB 37.0%, P < .01) (Figure 2B). In addition, multivariate analysis showed that mild and moderate to severe SDB significantly raised the odds ratio for anemia (odds ratio and 95% confidence interval as follows; mild SDB, 1.74 [1.32–2.29], P < .01; moderate to severe SDB, 3.85 [1.51–9.78], P < .01) (Figure 3A).

Figure 3. Adjusted odds ratio of SDB for anemia and microcytic anemia in premenopausal females.

(A) Adjusted odds ratio for anemia. (B) Adjusted odds ratio for microcytic anemia. Results were adjusted by age, body mass index, smoking history, drinking habit, menstrual phase at the time of examination, history of childbirth, history of collagen diseases, estimated glomerular filtration rate, C reactive protein, actual sleep duration and Acti-CT90% ≥ 0.8%. Numbers in parentheses under bars are the number of the participants in each of the 3 groups. Acti-CT90% = actigraphy adjusted the cumulative percentage of time during sleep with oxygen saturation below 90%, SDB = sleep-disordered breathing.

Because Acti-3%ODI correlated significantly with Hb in premenopausal females, we evaluated the association of SDB with MCV to investigate which type of anemia (microcytosis or macrocytosis) was associated with SDB. MCV decreased with increasing severity of SDB and significant differences were found among the 3 groups according to severity (normal vs mild vs moderate to severe SDB: 89.9 ± 6.5 vs 87.6 ± 7.7 vs 85.6 ± 9.0 fL, P < .01) (Figure 4A). In addition, multivariate analysis showed a significant negative correlation of Acti-3%ODI with MCV (β = −0.15, P < .01) (Table 4). Among the examined confounding factors, smoking history, childbirth history, C reactive protein level and sleep duration were significantly associated with both Hb levels and MCV in the premenopausal females (Table 3 and Table 4).

Figure 4. MCV and percentages of participants with microcytic, normocytic, and macrocytic anemia in groups categorized by the severity of SDB in premenopausal females.

(A) MCV. Black bar and white line in the bar represent mean value and standard deviation in each group. (B) Percentages of premenopausal females with microcytic, normocytic, and macrocytic anemia. Black, gray, and white bars represent the percentages of participants with microcytic anemia, normocytic anemia, and normal hemoglobin levels, respectively. Macrocytic anemia was found in only 0.3% (3/1,178) of participants in the no SDB group. Number in each bar represents the percentage of the participants. MCV = mean corpuscular volume, SDB = sleep-disordered breathing.

Table 4.

Multivariate analysis to determine contributing factors to mean corpuscular volume in premenopausal females.

	β	P
Age, years	0.03	.27
Current smoker, %	0.07	< .01
Frequent drinker, %	0.10	< .01
Body mass index, kg/m2	−0.11	< .01
Menstrual phase at the time of examination	−0.04	.13
History of childbirth	−0.09	< .01
History of collagen diseases	−0.01	.77
eGFR, mL/minute/1.73 m2	−0.02	.38
C reactive protein, ng/mL (log-transformed)	0.06	.02
Sleep duration, minutes	0.05	.04
Acti-3%ODI, per hour (log-transformed)	−0.15	< .01
Acti-CT90% > 0.8%	0.05	.05

Acti-CT90% = actigraphy-modified cumulative percentage of sleep time with oxygen saturation < 90%, Acti-3%ODI = actigraphy-modified 3% oxygen desaturation index, eGFR = estimated glomerular filtration rate.

Lastly, we categorized premenopausal females into 4 groups according to their Hb levels and MCV: normal Hb level, microcytic anemia, normocytic anemia and macrocytic anemia. The percentages of participants with each type of anemia categorized by the severity of SDB are presented in Figure 4B. Macrocytic anemia was found in only 0.3% (3/1178) of participants in the no SDB group. The rate of microcytic anemia increased in a stepwise manner as the severity of SDB increased (normal, 8.0%; mild SDB, 13.6%; and moderate to severe SDB, 29.6%; P for trend < .01). Conversely, no significant association was found between the rate of normocytic anemia and severity of SDB (normal, 15.5%; mild SDB, 18.4%; and moderate to severe, SDB 7.4%; p for trend = 0.57). Multivariate analysis demonstrated that SDB was significantly correlated with increased risk of microcytic anemia (odds ratio [95% confidence interval]: mild SDB, 1.93 [1.31–2.84], P < .01; moderate to severe SDB, 9.02 [3.02–26.88], P < .01) (Figure 3B).

DISCUSSION

The present analysis showed that higher Acti-3%ODI was associated with lower Hb values in premenopausal females. In addition, SDB was a significant risk factor for anemia, especially for microcytic anemia in this group. Conversely no significant correlation between Acti-3%ODI and Hb was found in postmenopausal females and males. To the best of our knowledge, this is the first study to examine the relationships between SDB and erythrocyte measures in a large sample from the general population. Although previous studies have focused on the relationship between SDB and polycythemia, this is the first study to show that SDB can present a risk for anemia in premenopausal females.^5,8,9,23 Interestingly, even mild SDB was significantly associated with microcytic anemia. Therefore, these findings suggest the need to check for anemia in premenopausal females with mild SDB as well as moderate to severe SDB.

Although the reason for the correlation of SDB with anemia and microcytosis is unclear, given that these correlations were found only in premenopausal females, SDB may be related to heavy menstrual bleeding. Liu et al showed that exposure of female rats to periodic hypoxia causes inflammation of the endometrium and other previous study demonstrated that nuclear factor κB, which is a transcription factor that responds to hypoxia, plays an important role in endometrial shedding and repair in the menstrual cycle.^24,25 However, another study revealed that hypoxia is required for normal endometrial repair in the menstrual cycle.²⁶ Further investigations are needed to clarify the impact of SDB on the endometrium and menstrual bleeding. As another possible mechanism underlying the correlation of SDB with anemia, previous studies suggested that hypoxia induced proliferation of uterine leiomyoma, which is one of the major causes of heavy menstrual bleeding.^27,28 In addition, these studies showed that hypoxia inducible factor 1α, which is the master regulator of the cellular response to hypoxia, mediated the response of the leiomyoma cells to hypoxia. SDB possibly leads to proliferation of uterine leiomyoma via its hypoxia inducible factor 1α-mediated response to hypoxia.

Alternatively, anemia and iron deficiency are associated with restless legs syndrome.²⁹ Patients with restless legs syndrome frequently exhibit involuntary limb movement during sleep, and arousals accompanied by limb movement can lead to hyperventilation followed by central apnea.^30–33 Investigating whether the relationship between SDB and microcytic anemia observed in premenopausal females varies depending on the presence of restless legs syndrome symptoms could help to elucidate the mechanisms linking SDB and anemia. In this study, because we only assessed nocturnal oxygen levels by pulse oximetry, it was not possible to determine whether each respiratory event was obstructive or central. Polysomnographic data, including electromyogram of the legs, arousals on electroencephalogram and types of respiratory events, might help to clarify some of the mechanisms by which anemia and SDB are related in premenopausal females. In addition, a history of childbirth was significantly associated with lower Hb levels and MCV in premenopausal females in this cohort. Given that SDB, anemia and restless legs syndrome can be exacerbated during pregnancy, examining the relationships among these 3 factors in pregnant females could provide insights into the mechanism of the association between SDB and anemia observed in this study.^34,35

On the other hand, although Acti-CT90% is an index of nocturnal hypoxia, as is Acti-3%ODI, this was not associated with Hb in all 3 groups. While reasons for the discrepancy between 3%ODI and CT90% in relation to Hb are unclear, effects of hypoxia on tissue oxygen profiles and metabolic outcomes can vary depending on the frequency and duration of hypoxia and target organs.³⁶ The mechanism by which SDB correlates with anemia in premenopausal females may be more strongly related to the frequency rather than the duration of hypoxemia. In addition to examining whether a similar relationship between SDB and anemia is found in premenopausal females in other large cohorts, the association between SDB and menstrual bleeding and uterine leiomyoma may also need to be evaluated in the future studies.

In the present study, the association of SDB with higher Hb was not found in all 3 cohorts categorized by sex and menopausal status. On the other hand, a recent meta-analysis demonstrated an increased prevalence of polycythemia in patients with obstructive sleep apnea.²³ There are some possible reasons for these discrepant findings. First, while the participants in the studies included in the meta-analysis were those who visited the hospital for diagnosis of SDB, participants in the present study are from the general population. Therefore, the present study included fewer participants with severe SDB than these previous studies. Because it has been reported that erythropoiesis can be stimulated only in patients with severe SDB but not in mild to moderate SDB, the correlation of SDB with erythrocyte indices may vary from study to study depending on the SDB severity of the analyzed cohort.² Whether the relationship between SDB and microcytic anemia observed in premenopausal females in this study remains consistent in hospital-based cohorts should be confirmed. Second, the prevalence of SDB and iron deficiency varies by ethnicity, and the association between SDB and anemia might depend on the ethnicity of the cohorts.^37,38 Because all participants in this study were East Asians, we could not evaluate the influence of the ethnicity on the relationship between SDB and anemia. The existence of an association between anemia and SDB in premenopausal females should be investigated in large cohorts of different ethnicities. Third, Song et al showed that obstructive sleep apnea may promote hematopoiesis via an increase in erythropoietin and simultaneously cause hemolysis by increased blood reactive oxygen species and hepcidin.³⁹ In this cohort, the effects of obstructive sleep apnea on hematopoiesis and hemolysis may have been offset, resulting in no positive association between Hb and indices of nocturnal hypoxemia. Further investigations are needed to clarify how the balance between the effects of obstructive sleep apnea on hematopoiesis and hemolysis varies from individual to individual.

Some limitations should be noted in the present study. First, since this was a cross-sectional study, we could not clarify the causal relationships between SDB and anemia and microcytosis in premenopausal females. Future prospective studies examining the changes of erythrocyte variables when SDB is treated in premenopausal females are warranted. In addition, we have not collected data that would provide clues as to the cause of microcytic anemia, such as menstrual bleeding patterns and iron dynamics, because of the difficulty of objective evaluation and cost issues. These data could help to elucidate the mechanisms underlying the association of SDB with anemia. Second, we evaluated the severity of SDB with pulse oximetry and could not obtain data about sleep duration and architecture by polysomnography. However, in order to ensure that results regarding variables associated with hypoxia in the present study were comparable to those obtained through polysomnography, we adjusted the duration of pulse oximetry measurements to match the actual sleep duration recorded by actigraphy. Consequently, the parameters of nocturnal hypoxia measured in this study are expected to closely resemble those observed during polysomnography.¹⁸ Therefore, the significant correlation between indices of hypoxia and erythrocyte variables observed in the present study would not change when analyses were done by polysomnography. Third, some variables, such as smoking history, alcohol consumption, and clinical history, were self-reported, potentially introducing recall bias.

In conclusion, the present study demonstrated that SDB was not necessarily correlated with higher Hb levels in a general population. In addition, not only moderate to severe but mild SDB was associated with microcytic anemia in premenopausal females. When SDB is diagnosed in a premenopausal female, even if it is mild SDB, we may need to evaluate her erythrocytic parameters. The pathophysiologic mechanism underlying the correlation of SDB with microcytic anemia in premenopausal females should be investigated in future studies. Because the risk of SDB is lower in young females than in males or older adult females, the impact of SDB on other diseases in young females has not been fully investigated.⁴⁰ However, when SDB is found in young females, it might have a larger impact on comorbidities and health outcomes than observed in males and older adult females, as typified by the relationship between SDB and anemia observed in this study.¹² Further studies are warranted investigating the impact of SDB on health outcomes in young females.

DISCLOSURE STATEMENT

All authors have read and approved the final version of the manuscript. This was not an industry-supported study. This study was supported by a university grant; Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (Grant numbers: 25293141, 26670313, 26293198, 17H04182, 17H04126, 17H04123, 18K18450, 19K17674, 20H03690, 23K07638, and 24K11363) and grants from the Center of Innovation Program and the Global University Project from the Japan Science and Technology Agency, Japan Agency for Medical Research and Development (Grant numbers: dk0207006, dk0207027, ek0109070, ek0109283, ek0109196, ek0109348, kk0205008, ek0210066, ek0210096,ek0210116, ek0210150, le0110005, and wm0425018). This research was also supported by grants from Ministry of Health, Labour and Welfare of Japan (Grant numbers, Intractable Respiratory Diseases and Pulmonary Hypertension Research Group: H29-intractable diseases-general-027 and JPMH20FC1027, Special Health Check-Up Research Group: R3-Comprehensive Research on Life-Style Related Disease including Cardiovascular Diseases and Diabetes Mellitus-21FA1004, Regional Medical Care Research Group: H28-iryo-ippan-016 and H30-iryo-ippan-009). The Takeda Medical Research Foundation, the Mitsubishi Foundation, Daiwa Securities Health Foundation, the Sumitomo Foundation, Japan Respiratory Foundation, and the Japan Research Foundation for Healthy Ageing supported the present study by providing their grants. Kimihiko Murase and Satoshi Hamada report that the Department of Advanced Medicine for Respiratory Failure of Kyoto University Graduate School of Medicine is funded by grants from Teijin Pharma to Kyoto University. Naomi Takahashi, Hironobu Sunadome and Susumu Sato report that the Department of Respiratory Care and Sleep Control Medicine of Kyoto University is funded by grants from Philips Japan, Fukuda Denshi, Fukuda Lifetec Keiji, and ResMed Japan to Kyoto University. Kazuo Chin reports that the Department of Sleep Medicine and Respiratory Care, Division of Respiratory Medicine, Department of Internal Medicine of Nihon University is funded by grants from Philips Japan, Fukuda Denshi, Fukuda Lifetec Tokyo, and ResMed Japan to Nihon University. Other authors have nothing to disclose.

ABBREVIATIONS

Acti-CT90%

actigraphy-modified cumulative percentage of time during sleep with oxygen

Acti-3%ODI

actigraphy-modified 3% oxygen desaturation index

CT90%

cumulative percentage of time during sleep with oxygen saturation below 90%

Hb

hemoglobin

MCV

mean corpuscular volume

3%ODI

3% oxygen desaturation index

SDB

sleep-disordered breathing