Menstrual blood loss is an independent determinant of hemoglobin and ferritin levels in premenopausal blood donors.

Katja van den Hurk conceived the original idea. Sofie Ekroos, Mikko Arvas, and Katja van den Hurk designed the study. Sofie Ekroos and Jan Karregat performed the analyses. Sofie Ekroos, Mikko Arvas, and Katja van den Hurk interpreted the results with input from all authors. Sofie Ekroos wrote the manuscript with a critical review from all authors. Elena Toffol and Johanna Castrén provided medical insight. Katja van den Hurk and Mikko Arvas supervised. All authors approved the final version for submission.

All participants gave written informed consent. Ethical approval was provided by the Medical Ethical Committees of Arnhem‐Nijmegen (DIS) and Academic Medical Center, Amsterdam (DIS‐III) on July 25, 2014 (protocol number 2014_124#B2014664).

This study was partially funded by the MD‐PhD program at the University of Helsinki, the Finnish Red Cross Blood Service (FRCBS), and The Medical Society of Finland. Data collection, processing, and supervision were enabled by Product and Process Development Grants (PPOC‐14‐028 and PPOC‐19‐02) from the Sanquin Blood Supply Foundation.

Complete information was obtained from 1005 (74%) of 1367 participants. A total of 964 participants were included in the final sample after exclusions, with 473 (49%) premenopausal and 491 postmenopausal women. Table  2 summarizes sample characteristics. Ferritin and Hb levels were lower in premenopausal women, and 61 (13%) of premenopausal women had HMB based on PBAC score. Almost half of premenopausal women and 13% of postmenopausal women reported using some form of hormonal contraception, with oral contraceptives (OCs) being the most common among premenopausal women and LNG‐IUDs among postmenopausal women, if any. Study characteristics of all 981 participants included in the study after exclusion. Note : All covariates except for current smoking and the number of days since last whole blood donation were significantly different between pre‐ and postmenopausal women. Medians are presented with interquartile ranges (IQR) for non‐normally distributed variables and means with standard deviations (SD) for normally distributed variables. There was a significant and consistent negative association between PBAC score and Hb and ferritin levels (Figure  1 ). Mean ferritin was highest in LNG‐IUD users (50.33 μg/L), followed by users of other hormonal contraception types (44.59 μg/L), oral contraception users (33.51 μg/L), and non‐users (31.23 μg/L). Mean Hb levels were highest in all forms of hormonal contraception use (LNG‐IUD 136.7 g/L, other type 135.8 g/L, oral contraception 135.2 g/L) compared to non‐users (131.7 g/L). Mean PBAC scores were highest in non‐users of hormonal contraception (95.40), followed by oral contraception users (59.34), users of other hormonal contraception types (55.00), and LNG‐IUD users (7.173). Association of PBAC score to Hb and ferritin in premenopausal users and non‐users of hormonal contraception. Marginal histograms to the right show the number of hormonal contraception users and mean Hb/ferritin for the different types of use. The upper marginal histogram similarly describes hormonal contraception use and mean PBAC score among users and non‐users. The blue vertical line represents the cutoff for HMB. Anemia and ID cutoffs are represented by the red horizontal lines in the Hb and ferritin scatter plots, respectively. A simple linear regression line of Hb/ferritin explained by PBAC score is represented by the solid blue line. HMB, heavy menstrual bleeding; IUD, levonorgestrel‐releasing intrauterine device; PBAC, pictorial blood assessment chart. Smoking and PBAC score were associated with both ferritin and Hb (Figure  2 ). Additionally, postmenopausal status, number of blood donations in the past 2 years, and number of days since the last whole blood donation were associated with ferritin but not with Hb. Number of pregnancies, blood volume, and age did not show a significant association in either model. Although most variables were not consistently found to have an effect with 95% probability in both models, directions of explanatory variables were similar between the models. Removing the PBAC score variable (Figure  S2 ) caused only slight changes to the models. Multivariate model coefficient estimates with corresponding univariate estimates are available in Table  3 . Multivariate Bayesian linear regression coefficients for ferritin (A) and Hb (B). Black horizontal bars represent 95% credible intervals of estimates of standardized coefficients and the circles show the distribution medians. A 95% credible interval has a 95% probability to contain the actual coefficient, given the data. Filled circles highlight variables whose effect distributions do not contain 0 with at least 95% probability. As age was divided by 5 and PBAC score by 50, a unit increase corresponds to 5 kg and 50 points, respectively. Univariate and multivariate linear regression analysis. −0.71 ‐ 0.48 1.45 ‐ 8.85 1.05 ‐ 1.13 Note : Bolded beta coefficients are statistically significant. Based on variable importance analysis, PBAC score accounted for most of the explained variance for Hb and was the second most important factor for variance in ferritin, only after the number of days since last whole blood donation (Figure  3 ). Relative importance analysis (RIA) of a linear model of ferritin (A) and Hb (B). RIA estimates the average percentage of variance in the outcome variable explained by each covariate. A positive value represents a positive correlation, and a negative value represents a negative correlation. The bootstrapped 95% CIs are characterized by the black lines. In premenopausal women, 25.8% [21.9, 30.0] ( n  = 122) were iron deficient and 8.25% [5.93, 11.1] ( n  = 39) were anemic. In contrast, only 9.37% [6.94, 12.3] ( n  = 46) of postmenopausal women were iron deficient and 1.63% [0.710, 3.19] ( n  = 8) were anemic. Ninety‐five percent confidence intervals were calculated using binomial distribution using the Pearson‐Clopper method. Results of the logistic Bayesian model for premenopausal women who reported menstruating ( n  = 402, 85% of premenopausal women) are shown in Figure  4 . ID was positively associated with the number of whole blood donations in the past 2 years, while smoking, higher number of days since last donation, and LNG‐IUD use was negatively associated with ID. HMB, duration of flow, and number of pregnancies were positively associated with anemia while higher number of years since menarche was negatively associated with anemia. None of the other variables studied, including the use of other types of hormonal contraception and regularity of menstrual cycles, were associated with ID or anemia. When replacing the HMB variable with PBAC score, MBL was positively associated with both ID and anemia (Figure  S3 ). Multivariate model coefficient estimates with corresponding univariate estimates are available in Table  3 . To further validate our results, we tested the logistic model variables in a linear model with PBAC score as the outcome measure (Figure  S4 ). Despite not being associated with iron deficiency in the final logistic model, oral contraception was negatively associated, and duration of menstrual bleeding was positively associated with MBL, suggesting some of the effects of the single independent variables in the final models may be lost due to sample size. Associations of menstruation‐related variables and donor characteristics on ID (A) and anemia (B). Models are Bayesian logistic regression models. Black horizontal bars represent 95% credible intervals of estimates of standardized coefficients and the circles show the distribution medians. A 95% credible interval has a 95% probability to contain the actual coefficient, given the data. Filled circles highlight variables whose effect distributions do not contain 1 with at least 95% probability. A unit increase is equivalent to a doubling of the number of days since last blood donation. A unit increase in age corresponds to 5 years.

In this study, we found that volume of menstrual blood loss is consistently and significantly negatively associated with ferritin and Hb in female blood donors. Blood loss due to whole blood donation and menstruation were the two most important determinants for ferritin and Hb levels in the female blood donor population. HMB and duration of menstrual blood flow were associated with anemia and, though not significantly, with ID in menstruating blood donors. LNG‐IUD use was negatively associated with ID, and increased number of years since menarche was negatively associated with anemia. Our study is supported by earlier results on menstruation and iron status of women in the general population. Our finding that MBL is associated with Hb and ferritin levels is in line with results from Harvey et al., who using the alkaline hematine method showed that MBL accounted for the majority of variance in ferritin. 29 Napolitano et al. also showed a significant correlation between MBL and iron loss as well as a 5–6 times higher iron loss in women with HMB compared to women with normal MBL. 30 Previous studies of whole blood donors in Denmark, New Zealand, and the USA have already shown the duration of menstruation to be significantly associated with ferritin levels and ID (Rigas et al., 16 Beck et al., 17 Cable et al. 18 ). Our results add that the amount of blood loss is associated with ferritin and Hb, and the effect of volume of loss is even more strongly associated with anemia than the duration of menses. Although HMB was not significantly associated with ID in our model, not every variable that influences MBL extends its effect to ferritin/Hb as our supplementary results (Figure  S4 ) show. This suggests that the effect of a single independent variable used in the logistic model might be too small to be captured or that our cohort size is too small to see full effects. It is also possible that some variables are too similar, that is duration of bleeding and HMB both consider PBAC score, although both duration and extent of MBL have been associated with ferritin/Hb in previous studies (Beck et al., 17 Heath et al. 31 ). Ferritin levels were significantly higher and Hb levels somewhat higher in LNG‐IUD users, suggesting that LNG‐IUDs have a protective effect in female blood donors. Additionally, although none of the other types of hormonal contraception showed an association, LNG‐IUDs were found to have a protective effect against ID but not anemia. As women in our study were only asked about the use of OCs, not type, we are not able to differentiate between combined estrogen‐progesterone and progestin‐only pills nor type of treatment regimen (i.e., cyclic or continuous use). This limitation could partially explain why we did not see any effect of OCs in this study, yet OCs are known to decrease MBL. Few contemporary studies from industrialized countries exist on the extent to which different forms of hormonal contraception affect ID/anemia, but studies regarding the use of LNG‐IUDs are in line with the results from this study. A recent Cochrane review 32 concluded that LNG‐IUDs appear superior to OC in the treatment of HMB and both a clinical trial (Mawet et al. 33 ) and literature review (Lowe and Prata 34 ) showed ferritin and Hb levels improving as MBL decreased. However, in studies of other types of hormonal contraception, results are conflicting. No association was found for the use of OC on ID or ferritin by Beck et al. 17 and Alfaro‐Magallanes et al. 35 with colleagues. Heath et al. 31 and colleagues found the use of OC protective against ID, but the effect disappeared after controlling for MBL. Contrarily, a study by Milman et al. 36 found higher ferritin levels in women using oral contraception. With the exception of Alfaro‐Magallanes et al.'s study, which only included monophasic OCs, all other studies mentioned grouped all forms of OCs into one group which could explain contrary results. Although the effect of LNG‐IUDs likely is due to a decrease in PBAC score, our logistic regression results suggest that LNG‐IUD use has an independent effect on ID risk. As PBAC is known to underestimate blood loss, it appears more likely that this underestimation is captured by the LNG‐IUD variable rather than LNG‐IUDs having an independent effect. As ferritin is an active phase protein, inflammation caused by the LNG‐IUD as a foreign object could also explain higher ferritin in LNG‐IUD users. In contrast, a large study by Toffol and colleagues 37 found Finnish LNG‐IUD users to have reduced inflammation markers, making a ferritin increase due to inflammation a less likely explanation. In 2013, 12% of premenopausal women in the Netherlands used an LNG‐IUD 38 compared to 17.9% of premenopausal women in this study. A protective effect of LNG‐IUD use could be an explanation for why LNG‐IUD use is higher in blood donors compared to the general population, as this could create a form of self‐selection, allowing the donor to tolerate the donation better. As many factors other than bleeding patterns influence the prescriber when selecting the type of hormonal contraception, including patient preference, it is possible there are other reasons behind higher LNG‐IUD use in this cohort. This is, to our knowledge, the first time PBACs have been used to study the effect of MBL in blood donors. MBL was the most important determinant of iron levels and difference in average Hb between the pre‐ and postmenopausal groups is likely to stem from menstruation. Thus, our results suggest that in addition to age, questions regarding menstruation and hormonal contraception use might help inform recommended donation intervals. Due to practical reasons, incorporating PBAC into donor selection may not be possible and female donors may not be able to correctly answer a question related to suffering from HMB due to not being aware that their MBL is classified as HMB. 23 However, including questions associated with clinical features of HMB 24 , 39 could be beneficial in identifying high‐risk women. These include asking about menstrual product use (using double protection, rate of required change, changing protection during night‐time), flooding episodes, clot sizes, and quality of life during menstruation. Alternatively, an optional PBAC could be included in a blood donor app, allowing the blood donor to register the volume of MBL during a cycle. The PBAC score could then be included in an algorithm calculating risk for ID/anemia, allowing for more personalized donation intervals. Importantly, donors and staff at blood services should be informed about the effects of menstruation on iron levels. Although causality should not be determined due to the cross‐sectional nature, we believe our study provides much‐needed insight into the causes of ID/anemia in female blood donors. A limitation of our study is that PBAC scores were collected during one single menstrual cycle, and it is thereby possible that exogenous factors affected blood loss during the time of collection. However, due to low interindividual variation in PBAC score as described by Janssen and colleagues, 20 additional cycles are not expected to meaningfully add to results. Sanquin does not recommend or supply donors with iron supplements, and only a total of three donors representing 0.003% of donors in this sample (two premenopausal and one postmenopausal) reported using iron supplements after data filtering. As such, further studies should investigate the effect of donor iron supplementation with a focus on donors who report current menstruation, particularly those with HMB. Further prediction models for ID/anemia and focus group discussions are needed to ascertain the usefulness of questions related to menstruation in blood donors, and whether women are willing to provide information about their reproductive health to blood services. Finally, to determine whether oral contraceptives affect iron stores in blood donors further research needs to compare drugs by the Anatomical Therapeutic Chemical (ATC) classification system.

Blood services are responsible for protecting the health of blood donors. One challenge is identifying donation intervals that allow donors with greater capacity to donate more frequently and extending intervals for donors needing longer time to recover between donations. As our results suggest that blood donors could benefit from donation intervals based at least partially on information of menstruation and hormonal contraception use, accounting for these may be an important piece to this puzzle.

In the past decades, concern has risen over the global decrease in blood donation. The recent SARS‐CoV‐2 pandemic worsened the problem and caused an acute shortage of blood products. Blood transfusion often cannot be deferred, such as in the case of obstetric pathologies like postpartum hemorrhage. 1 Despite donor selection processes to ensure donor safety, frequent whole blood donors are at risk of developing iron deficiency (ID), which can progress to iron deficiency anemia, a condition detrimental to donor health if left untreated. According to World Health Organization guidelines for women, anemia is defined as hemoglobin (Hb) levels <120 g/L and ID is defined as serum ferritin levels <15 μg/L. 2 The only legislated donor deferral criterion in the European Union is a cut‐off point for Hb <125 g/L. 3 Additionally, some blood services base donation interval‐ and iron supplementation policies on age, sex, and ferritin screening. 3 Research shows that a temporary blood donation deferral has a negative impact on future willingness to donate blood, especially among new donors, and retaining current donors is preferable to recruiting new donors. 4 Due to menstrual blood loss and pregnancies, premenopausal women are at increased risk of ID/anemia. 5 Although studies have shown lower ferritin and Hb at the beginning of the follicular phase compared to the luteal phase, 6 more recent studies have not been able to show a significant inter‐individual variation in Hb 7 or change in Hb mass caused by menstrual cycle phase. 8 Whereas women, on average, lose around 40 mL of blood during one menstrual cycle, heavy menstrual bleeding (HMB) is defined as menstrual blood loss (MBL) of ≥80 mL. 9 Several structural (PALM) and non‐structural causes (COEIN) can underly HMB, as classified by PALM‐COEIN classification system (polyp; adenomyosis; leiomyoma; malignancy and hyperplasia; coagulopathy; ovulatory dysfunction; endometrial; iatrogenic; not yet classified). 10 Women who suffer from HMB face greater risk of ID and anemia. 11 Although lifetime prevalence is estimated to be 10%–40%, HMB remains underdiagnosed. 12 , 13 Treatment of HMB, including use of hormonal contraception such as levonorgestrel‐releasing intrauterine devices (LNG‐IUD) or tranexamic acid to decrease MBL, 14 and subsequent correction of anemia has been shown to be associated with improved quality of life. 2 Despite women being at higher risk compared to men for low hemoglobin deferral and donation‐related adverse events such as fatigue and vasovagal reactions, women are highly motivated to donate blood, and young women <40 years are proportionally overrepresented among whole blood donors. 15 To protect the health of premenopausal blood donors and to prevent blood donor deferral, it is crucial to understand the extent to which different factors underly the risk of ID and anemia in the donor population. Although blood services recognize premenopausal women to be at higher risk of ID and anemia, and previous studies suggest menstruation is associated with lower iron status in blood donors, 16 , 17 , 18 to our knowledge, none of them utilize information on volume of blood loss or any other factors related to female reproductive health in donor selection or interval guidelines. Thus, it is expected that volume of blood loss during menstruation will affect ferritin and Hb in female blood donors, and that HMB will put female donors at higher risk of ID and anemia. Moreover, use of hormonal contraception that decrease MBL is expected to decrease risk of ID and anemia in blood donors. In this study, our goal was to investigate if and to what extent MBL is associated with variations in ferritin and Hb levels in female blood donors, and whether other female reproductive health‐ and menstruation‐related variables, such as hormonal contraception use and HMB, are associated with risk of ID or anemia.

The authors have declared no conflicts of interest.

The alkaline hematine method is the most reliable way to quantify MBL, but it is unpractical for clinical use. 19 Pictorial blood assessment chart (PBAC) is a semiquantitative method that allows menstruating individuals to evaluate the number of used menstrual pads and tampons, their degree of staining, and number and size of blood clots present. The scoring system allows for a total score, which can be used to estimate menstrual blood volume. Although self‐reported and prone to progressive underestimation of MBL as the volume increases and the menstrual item becomes saturated, PBAC is easy to incorporate into clinical practice and has a lower cost compared to the alkaline hematine method. 19 , 20 Several variations of PBACs exist. 19 A modified Janssen PBAC (Appendix  S1 ) does not take into consideration number or size of blood clots present, as a validation study suggests only minor predictive value. 20 In the modified Janssen PBAC, depending on if the menstrual product is lightly, moderately, or heavily stained, each menstrual pad is given 1, 5, or 20 points and each tampon is given 1, 5, or 10 points, respectively. MBL of ≥80 mL corresponds to a total PBAC score of ≥150. 9 Donor InSight (DIS) is a prospective longitudinal nationwide cohort study of blood donors from the Netherlands, consisting of three questionnaire rounds, the last one (DIS‐III) also including blood sampling. Data from complete blood counts, ferritin measurements, and genetic arrays are included in DIS‐III (2015–2016, n  = 3046). The cohort data are linked to the Sanquin donor database, containing details on donations and health questionnaires. The cohort has been described extensively elsewhere. 21 In line with donor selection criteria, participants had to meet general donation health requirements. A pre‐donation screening consisting of a donor health questionnaire, an interview to screen for pathologies (including, but not limited to; bleeding disorders, cancer, insulin‐dependent diabetes mellitus, active inflammatory bowel disease), and high‐risk behavior (such as drug use and travel) that would cause either temporary or permanent deferral from blood donation, and blood pressure and Hb measurement were performed before donation. As per Sanquin policy, participants were not issued postdonation iron supplements. An invitation letter for DIS‐III including a link to an online questionnaire was sent to 6140 randomly selected participants of DIS‐I and/or DIS‐II. Females were provided a modified Janssen PBAC and asked questions regarding menstruation, contraceptive use, and pregnancy. Out of 3046 DIS‐III participants who completed the questionnaire and/or provided blood samples, 2552 donors in total provided necessary blood samples, completed at least part of the questionnaire, and gave informed consent. 21 A total of 1367 (54%) donors were female whole blood donors. Peripheral blood was collected from the sampling pouch during successful blood donation. Separate venous samples were drawn if the donor was deferred or unwilling to provide a full donation. All samples were sent to Amsterdam and processed within 24 h after donation. Full blood counts (XT‐2000, Sysmex, Kobe, Japan) were performed from EDTA full blood immediately after processing. Plasma samples from lithium heparin tubes (Architect Ci8200, Abbott Laboratories, IL, USA) were stored at −80°C and used to measure ferritin within 12 months after sampling. Women who reported menstruating were assigned as premenopausal. Women who reported not menstruating due to menopause, hysterectomy or bilateral oophorectomy and women ≥55 years old who did not answer the question whether they still menstruated (as by this age, menopause likely would have occurred) were assigned as postmenopausal. Women under the age of 55 years old who did not answer the question whether they still menstruated were excluded. Exclusion criteria and information of participants excluded due to missing data are presented in Table  1 . Additionally, a subpopulation consisting of only menstruating women ( n  = 402) of the cohort was created to test the effect of variables related to female reproductive health and menstruation on ID/anemia. Flow diagram of inclusion and exclusion criteria. A subpopulation of only menstruating premenopausal women was created to test the secondary research question as explained in “aims.” As outcome variables for the linear models, we used Hb levels or log transformed ferritin levels measured at the time of DIS‐III. Anemia (Hb <120 g/L) and ID (ferritin <15 μg/L) were used as outcome variables in the logistic models. For linear analysis, in addition to PBAC score, we selected explanatory variables previously associated with ferritin levels in a cohort of Finnish whole blood donors. 22 Based on results from the linear models and literature on factors related to MBL/HMB, 9 , 23 , 24 we selected variables available in the questionnaire for the logistic models. For categorical variables, population characteristics were described as proportions (number and percentage). Normally distributed continuous variables were described by mean and standard deviation (SD) and non‐normally distributed variables by median and interquartile range (IQR). All analyses were performed using R Statistical Software (v4.1.2; R Core Team 2021). The code used is available on GitHub. 25 We used three methods to analyze data. First, we studied the association of variables to log(ferritin) or Hb in all cohort participants with Bayesian multivariable linear regression models. Secondly, we measured the average variance explained of log(ferritin) or Hb by each variable with relative importance analysis (RIA). The “relaimpo” package estimates the proportionate contribution of each variable to the total model R 2 by adding variables to the model sequentially and averaging over all possible orders. 26 Thirdly, we measured the association of variables to risk of ID or anemia in the cohort participants reporting menstruation with a Bayesian multivariable logistic regression. We used “brms” 27 for fitting Bayesian models and “bayesplot” 28 for MCMC diagnostics, allowing us to visualize possible divergent transitions, autocorrelation, and Rhat estimates. For linear models, we used PBAC as a measure of MBL, dividing by 50 to simplify coefficient interpretation. PBAC scores for women who reported not menstruating were recorded as 0. For logistic models, we used the categorical variable heavy menstrual bleeding, assigned as PBAC score ≥150. To quantify blood loss due to blood donation behavior we used number of days since the last whole blood donation (base‐2 logarithm) and number of whole blood donations in the past 2 years along with a quadratic term to account for non‐linear relationship. Age was divided by 5 to simplify coefficient interpretation. Nadler's equation was used to estimate blood volume. Hormonal oral contraceptives included both combined oral contraceptives and progestin‐only oral contraceptives. All hormonal intrauterine devices were 52 mg levonorgestrel‐releasing systems. All other forms of hormonal contraception (contraceptive implants, vaginal rings, patches, and injections) were put into a separate group due to the low number of users. All data transformations and further information on explanatory variables are available in the supplement (Table  S1 ). We included an interaction term for smoking and being postmenopausal in the final linear model, as results from running the models separately for pre‐ and postmenopausal women suggested different associations between smoking and Hb levels (Figure  S1 ). Additionally, we ran the linear model without the PBAC variable (Figure  S2 ) and replaced the HMB variable with PBAC score in the logistic model (Figure  S3 ) to check model stability. We also tested the association of variables used in the logistic model on PBAC score (Figure  S4 ).

Appendix S1. Figure S1. Figure S2. Figure S3. Figure S4. Table S1.