{"paper_id":"b7d28281-13dc-44e4-b70e-0d199a88df8f","body_text":"Material for this study was selected from 2 population-based surveys conducted in Finland, namely the Health 2000 and the Health 2011 studies ( 16 ,  17 ). The Health 2000 was a nationally representative comprehensive survey consisting of interviews, questionnaires, and extensive health examinations conducted across 80 regions in Finland in 2000-2001; overall, more than 8000 people aged 30 years and over were invited to participate. The Health 2011 was a follow-up study conducted in 2011-2012. Over 8000 individuals who had participated in the Health 2000 survey and an additional sample of nearly 2000 young adults (18-29 years) were invited to participate in the follow-up study by means of health examination (inclusive of blood sampling), interviews, and/or questionnaires. The surveys were approved by the Coordinating Ethics Committee of the Helsinki and Uusimaa Hospital District (HUS Regional Committee on Medical Research Ethics). All participants gave written informed consent.\nThe population for this study included all fertile-aged (18-49 years) respondents with available (self-reported) information on their current and past use of HC and metabolomics data obtained from serum samples drawn in connection with the health examination.\nIn each survey, the following exclusion criteria were applied: current pregnancy, menopause (“Did your periods end?”, with answers “naturally with the menopause” or “because of an operation or radiotherapy”), current use of menopausal hormonal therapy, hysterectomy, and current use of oral contraceptive pills (or of vaginal ring or transdermal patch, information available only in Health 2011), resulting in a final population of 1767 individuals with available metabolomics data (341 currently using a LNG-IUD and 1426 HC non-users). Detailed sample size and background information for each survey are reported in Supplementary Table S1 ( 18 ) and Supplementary Fig. S1 ( 18 ).\nOf this population, 240 of the women with metabolomics data who were using LNG-IUD or no HC in 2000 had also information on their LNG-IUD use and metabolomics data available at the Health 2011 and thus contributed to the sample used for the longitudinal study. Women who were using oral contraceptives in 2000 or 2011 were excluded from this part of the study.\nInformation on current and previous use of HC (pills and LNG-IUD) and its duration was obtained through questions during the health interview. All surveys additionally inquired into current use of nonhormonal IUDs; the Health 2011 also included information on the use of the contraceptive vaginal ring, patch, or other HC. The 52 mg LNG-IUD was the only hormonal IUD commercially available in Finland until 2013, with an approved maximum duration of use in Finland at the time of 5 years.\nIn addition to age, covariates obtained from the health examination, self-administered questionnaires, and clinical measurements included the following: season of sampling; alcohol use (yes, no/quit); frequency of smoking (every day, sometimes, never); physical activity (regular physical activity vs no regular physical activity); current use (during the past 7 days) of prescription drugs as listed in  Table 1 ; chronic/severe diseases (heart diseases, hypertension, venous thrombosis, diabetes, psychological or mental illnesses, rheumatoid arthritis, osteoarthritis, cancer); and body mass index (BMI) based on the measured height and weight.\nBackground characteristics of the study population (Pooled Health 2000 and Health 2011 data, n = 1767)\nAbbreviations: BMI, body mass index; LNG-IUD, levonorgestrel-releasing intrauterine device.\na \n Range: 30-49 years in Health 2000; 18-49 in Health 2011.\nb \n Insulin, oral hypoglycemics, antithrombotic therapy, lipid medications, combination products of estrogens and progestins, estrogens, progestogens, systemic corticosteroids, systemic antibiotics, systemic antimycotics, cytostatics, anti-inflammatory analgesics, opioids, other analgesics (no nonsteroidal anti-inflammatory drug group), migraine medications, epilepsy medications, antipsychotics, anxiolytics, hypnotics and sedatives, antidepressants, psycholeptics and psychoanaleptics in combination, long- and short-acting beta agonists, combination product of beta agonist and corticosteroids, inhaled corticosteroids, anticholinergics.\nc \n Heart diseases, hypertension, venous thrombosis, diabetes, psychological or mental illnesses, rheumatoid arthritis, osteoarthritis, cancer.\nIn each survey, the metabolomics measures were obtained from blood samples drawn in connection with the health examination, after fasting and without drinking on the same day, and after a minimum of 4-hour fasting. The serum and plasma samples were centrifuged at 1600 to 1800 g for 10 minutes and immediately frozen to −20 °C on site, normally within 45 to 60 minutes but no later than 90 minutes from sampling; thereafter, they were transferred to their final storage location (−70 °C), no later than 1 to 2 weeks after sampling. Serum sections were analyzed with a high-throughput serum nuclear magnetic resonance metabolomics platform ( 1 H NMR Spectroscopy, Nightingale Health 2018-2019). Biomarkers were quantified independently for each serum sample. Nightingale's biomarker analysis technology applies a single experimental setup, and spectral information is converted to absolute concentrations (in molar units) of the metabolic measures. The platform allows for the simultaneous quantification of 250 metabolic biomarkers per sample, including 12 lipid measures of 14 lipoprotein subclasses [6 very-low-density lipoproteins, 4 high-density lipoproteins (HDLs), 3 low-density lipoproteins (LDLs), intermediate-density lipoprotein], other detailed molecular information on serum lipids (eg, sphingomyelin, fatty acids, etc.), or low molecular weight metabolites (eg, amino acids). An additional set of metabolite ratios, recognized as proxies for enzymatic activity and/or a biologically separate entity of measures, is also computed to obtain further clinical insight ( 19-21 ).\nThe study consists of a cross-sectional and a longitudinal part. After preliminary inspections of the metabolomics data in each cohort, metabolites with more than 100 missing observations were excluded from the analyses, resulting in a total of 212 metabolic measures; metabolic measures with value “zero” were replaced with 0.25 × the minimum observed value for that metabolite, and, after log transformation, remaining missing data were imputed through random forest imputation ( 22 ). Given the rather long time interval between the 2 surveys, in the cross-sectional part of the study, single observations were treated as independent, and the 2 datasets were pooled together on common variables. Analyses of associations were carried out via linear regression models, with each metabolic measure as the outcome variable and use of a hormonal IUD (vs current non-use of any HC) as the predictor of interest using the “ggforestplot” R-package ( 23 ). To exclude dependency issues related to some women providing data to both Health 2000 and Health 2011 surveys, analyses were repeated with generalized estimating equation (GEE) models (with the “gee” R-package, with exchangeable correlation structure). Three models were fitted: model 1, controlled for age, BMI, season of sampling, and study cohort; model 2, which is model 1 further controlled for diseases and medication use, which have a potential impact on both the choice of HC and the metabolic status; and model 3, which is model 1 further adjusted for lifestyle factors that likely affect metabolic status and choice of HC use, such as alcohol use, smoking, and physical activity. Direct acyclic graphs illustrating covariate selection for each model are reported in Supplementary Fig. S2 ( 18 ).\nRegression model 3 was additionally repeated in age-stratified groups (18-39 years vs 40-49 years). As mentioned, metabolomics data of women younger than 30 years were available only in the Health 2011 survey; in addition, at the time of data collection, LNG-IUD was prescribed mostly to older and parous reproductive-age women. For these reasons, our overall sample of LNG-IUD users was underrepresented in the young age group. Hence, analyses were also repeated after exclusion of women younger than 30 years.\nIn order to allow the comparison across multiple measures, association magnitudes are reported in SD units of difference in biomarker concentration compared to the reference group. Results of analyses carried out separately in each of the 2 surveys (ie, Health 2000 and Health 2011) are reported as supplementary material (Supplementary Figs. S3-S5) ( 18 ).\nMoreover, to examine the effect of duration of use on the metabolic profiles of hormonal IUDs, linear regression analyses (model 3) were repeated in the group of current users only, comparing intermediate-term (2 to 5 years) and long-term (more than 5 years) use of LNG-IUDs to short-term cumulative use (up to 2 years).\nIn order to test if the possible associations of LNG-IUD use with specific metabolic patterns are detectable after discontinuation of its use, additional linear regression models (adjusted for age, BMI, season of sampling, alcohol use, smoking, and physical activity) compared the associations of current and previous use of LNG-IUDs and all the metabolic measures with never-users of LNG-IUD as the reference category. Because information on past use of LNG-IUD was not available in the Health 2011, these analyses were performed in the Health 2000 dataset only.\nFor the longitudinal part of the study, the 240 users of LNG-IUD and non-users of any HC in 2000 who also had available information on HC use (LNG-IUD vs no HC) and metabolomics data in 2011 were selected. Four groups were identified: (1) continuers, ie, women who used LNG-IUD at both time points; (2) stoppers, ie, women who used LNG-IUD in 2000 but no HC in 2011; (3) starters, ie, women who used no HC in 2000 but used LNG-IUD in 2011; and (4) never-users, who did not use any HC in 2000 or 2011. Changes in levels of each metabolite were computed as the difference between levels in 2011 and 2000. Linear regression models were further performed to compare changes in each metabolite level in the 4 groups of users of LNG-IUD, with never-users of HC being the reference category. Analyses were controlled for age at baseline and change in BMI.\nTo take the multiple testing into account, we applied the false discovery ratio procedure, which considers the expected proportion of false discoveries among the rejected hypotheses. All the analyses were performed with R software version 4.2.3 ( 24 ).\n\nTable 1  reports the baseline characteristics of the 1767 participants with available information on their current use of LNG-IUD and metabolomics data. A total of 341 individuals (19.3%) were using a LNG-IUD at the time of the surveys; they were older than those not using any HC (mean ± SD age 41.5 ± 6.1 vs 39.5 ± 5.1 years) and more likely to be employed (88.0% vs 76.1%), to be parous (97.9% vs 86.2%), and to use alcohol (96.2% vs 90.1%).\nIn linear regression analyses adjusted for age, BMI, season of sampling, and study cohort (model 1), 102 out of the 212 metabolic measures differed between LNG-IUD users and non-users of HC (median difference in biomarker concentration: −0.12 SD), in particular fatty acids concentrations and ratios, cholesterol, triglycerides, and other lipids, as well as particle concentration, cholesterol, triglycerides, total lipids, and phospholipids in lipoproteins ( Fig. 1 ). The pattern of associations remained substantially unchanged after adjustment for disease and medication use (model 2); adjustment for lifestyle habits mitigated the associations with cholesterol and other lipids, while the associations of triglycerides and fatty acids remained mostly unchanged (model 3) ( Fig. 1 ). Analyses conducted with GEE models to take correlation of observations of the same individual into account provided similar results (data not shown). In GEE, point estimates remain the same, but standard errors are increased. The results did not change after exclusion of women younger than 30 years of age (n = 47) (Supplementary Fig. S6) ( 18 ). The patterns of associations varied in different age groups: in particular, associations with lower levels of inflammation marker, fatty acids, and other lipids were mostly driven by the oldest age group (40-49 years, n = 972), while associations with lower levels of particle concentration, and in detail of cholesterol, triglycerides, free cholesterol, total lipids, and phospholipids in lipoproteins, were mostly evident in younger women (18-39 years, n = 795) ( Fig. 2 ).\nAssociations between the use of a levonorgestrel-releasing intrauterine device and 212 metabolic measures. Analyses conducted in pooled Health 2000 and Health 2011 surveys. Model 1 is adjusted by age, body mass index, season of sampling, and study cohort; model 2 is model 1 further adjusted for disease and medication use; model 3 is model 1 further adjusted for lifestyle habits (alcohol use, smoking, and physical activity). Results are in SD units of difference in metabolite concentrations; bars indicate 95% confidence interval; reference category is non-users of hormonal contraceptives. Closed circles indicate significant associations at  P -value adjusted for false discovery rate.\nAge-stratified associations between the use of a levonorgestrel-releasing intrauterine device and 212 metabolic measures. Analyses conducted in pooled Health 2000 and Health 2011 surveys. Analyses are adjusted by age, body mass index, season of sampling, study cohort, alcohol use, smoking, and physical activity. Results are in SD units of difference in metabolite concentrations; bars indicate 95% confidence interval; reference category is non-users of hormonal contraceptives. Closed circles indicate significant associations at  P -value adjusted for false discovery rate.\nTo examine the impact of duration of use on the associations between the use of LNG-IUD and metabolic profiles, we performed regression analyses in the subgroup of current users only, comparing the metabolic profile of intermediate-term (n = 129) and long-term (n = 154) use with short-term use (n = 58). No significant differences were found between the 3 groups ( Fig. 3 ).\nAssociations between duration of use of a levonorgestrel-releasing intrauterine device and 212 metabolic measures. Analyses conducted in the group of current hormonal intrauterine device users only in the pooled Health 2000 and Health 2011 surveys. Analyses are adjusted by age, body mass index, season of sampling, study cohort, alcohol use, smoking, and physical activity. Results are in SD units of differences in metabolite concentrations; bars indicate 95% confidence interval; reference category is short-term (up to 2 years) use of a hormonal intrauterine device. Closed circles indicate significant associations at  P -value adjusted for false discovery rate.\nWe further compared the metabolic profiles of current (n = 168) and previous users (n = 85) of LNG-IUDs with those of never-users (n = 938) in the Health 2000 Survey (Supplementary Table S2) ( 18 ). A model adjusted for age, BMI, season of sampling, alcohol use, smoking, and physical activity did not show any significantly different profiles in current or previous users compared to never-users of LNG-IUD (Supplementary Fig. S7) ( 18 ).\nIn the longitudinal analyses of the 11-year change in metabolic profiles of starters (n = 59), continuers (n = 33), and stoppers (n = 6) of LNG-IUD compared to never-users of HC (n = 142) ( Table 2 ), no significant differences were found ( Fig. 4 ).\nEleven-year changes in molecular concentrations of 212 metabolic measures by starting, continuing, and stopping use of a levonorgestrel-releasing intrauterine device. Analyses are adjusted by age at baseline and 11-year change in body mass index. Results are in SD units of change in metabolite concentrations; bars indicate 95% confidence interval; reference category is never-users of hormonal contraception. Closed circles indicate significant associations at  P -value adjusted for false discovery rate.\nBackground characteristics of the 240 participants included in the longitudinal study\nAbbreviation: BMI, body mass index.\na \n Insulin, oral hypoglycemics, antithrombotic therapy, lipid medications, combination products of estrogens and progestins, estrogens, progestogens, systemic corticosteroids, systemic antibiotics, systemic antimycotics, cytostatics, anti-inflammatory analgesics, opioids, other analgesics (no nonsteroidal anti-inflammatory drug group), migraine medications, epilepsy medications, antipsychotics, anxiolytics, hypnotics and sedatives, antidepressants, psycholeptics and psychoanaleptics in combination, long- and short-acting beta agonists, combination product of beta agonist and corticosteroids, inhaled corticosteroids, anticholinergics.\nb \n Heart diseases, hypertension, venous thrombosis, diabetes, psychological or mental illnesses, rheumatoid arthritis, osteoarthritis, cancer.\n\nThe main finding of this study is that the metabolic profile of current users of LNG-IUD differed in a relatively large number of metabolites compared to that of non-users of HC, but the magnitude of the associations was mostly from low to moderate. The detected associations were indicative of a reduced arterial cardiovascular risk, with lower levels (compared to non-users of HC) of fatty acids concentrations and ratios, triglycerides and other lipids, total lipids, and phospholipids in lipoproteins. The metabolic profile appeared not to vary with the duration of use, nor did we find any significant metabolic pattern related to 11-year changes in LNG-IUD use.\nThe overall results of this study are in line with our previous study ( 15 ), showing that the current use of LNG-IUD might be related to benign metabolic alterations. However, the magnitude of these alterations is likely to be of marginal importance. In the current study, the median difference in biomarker concentration between the users of LNG-IUD and non-users of HC was −0.12 SD. In line with the results of our previous study based on larger groups, we did not observe any difference in relation to the duration of use of LNG-IUD. In age-stratified analyses, this pattern of associations with lower levels of cholesterol, triglycerides, free cholesterol, total lipids, and phospholipids in lipoproteins was mostly evident in younger women. Although possibly explained by a selection bias, with women who were wealthier, more health-conscious, and in a long-term relationship being more likely to purchase LNG-IUD, altogether, these observations likely reflect the ongoing trend for LNG-IUD to be used at an increasingly younger age in Finland.\nTo the best of our knowledge, only 1 previous study has examined the metabolic effects of LNG-IUD in a longitudinal pattern, and it found no metabolic differences between starters, stoppers, and persistent users of LNG-IUD compared to persistent non-users ( 14 ). Similarly, in our study we found no metabolic changes related to changes in the use of LNG-IUD. Given the cross-sectional associations described here, the lack of a significant impact of starting or stopping LNG-IUD is somehow counterintuitive, although in line with the study by Wang et al ( 14 ). As such, it is likely that the null finding of our study, as well as that of Wang et al ( 14 ), is attributable to a lack of power consequent to the small subgroup sizes. However, because the detected cross-sectional associations are in the direction of protective effects, longitudinal associations in a larger sample, if any, were expected to be in the same direction. Taken together, the cross-sectional protective associations and the null finding in the longitudinal study point to a relative metabolic safety of LNG-IUD use.\nAs in our previous work, the majority of the associations were related to the absolute rather than relative concentrations of lipids and lipids in lipoproteins ( Fig. 1 ). In detail, absolute levels of cholesterol and triglycerides in lipoproteins were especially lower (compared to non-users) in the XL- and XXL-very-low-density lipoproteins, and to a lesser extent, in the LDL and intermediate-density lipoproteins subclasses, which altogether carry the highest risk for myocardial infarction ( 25 ) and other cardiovascular events ( 26 ,  27 ). Similarly, we observed a general decrease in absolute levels of all fatty acids. On the other hand, the results for fatty acid proportions of the total fatty acids were more difficult to interpret. Fatty acids ratios are biological meaningful measures that better reflect the biology of individual fatty acids than the absolute concentrations and have proven distinct population distributions ( 19 ,  20 ). In our study, we found a lower ratio of monounsaturated fatty acids but higher ratio of omega 6, linoleic acid, and polyunsaturated fatty acids (PUFA) in users of LNG-IUD (suggesting relative higher proportion of omega 6 than omega 3 within PUFA). This observation is rather counterintuitive, as both reduced and increased cardiovascular risk (mostly explained by its proinflammatory effect) has been shown in relation to omega 6 PUFA and its precursor linoleic acid ( 28 ,  29 ). On the other hand, high levels of omega 3 have a clearer protective role in terms of cardiovascular risk ( 30 ), while effects of monounsaturated fatty acids are less clear ( 31 ).\nAlthough we observed reduced levels of total HDL cholesterol associated with the use of LNG-IUD, we were able to obtain a detailed profile of HDL subfractions by means of metabolomics. In contrast to our previous work, where HDL2 appeared slightly more reduced than smaller and denser HDL3, we currently found lower levels of HDL3 but not total HDL or HDL2 in users of hormonal IUDs. A reason for these different findings could be the possible inclusion of users of 13.5 mg and 19.5 mg LNG-IUDs, along with the much larger sample size, in our previous study. As low levels of HDL2 are related to dyslipidemia and high levels of HDL3 to an increased risk of coronary heart disease ( 32 ), the current findings further support the potentially beneficial effects of LNG-IUD.\nFinally, the protective metabolic profile of LNG-IUD users is further confirmed by the additional finding of reduced levels of glycoprotein acetyls, an inflammation marker that has emerged as a predictor for diabetes, cardiovascular diseases, and all-cause mortality ( 33 ), in users of a LNG-IUD, as also previously shown in our study on hormonal IUD ( 15 ).\nThe lack of differences between current, previous, and never users of LNG-IUD is opposite to findings of our previous study, conducted in a larger population, showing a consistently different profile in current vs never users of LNG-IUD. As the analyses were performed in a smaller sample inclusive of only Health 2000 data, it cannot be excluded that our current finding is a false negative due to the small size of the 3 subgroups. However, we expect that, if differences were to be found, they would have been in the same direction as the main findings and as such be suggestive of a nonharmful metabolic profile related to LNG-IUD use.\nTaken together, our findings support the notion of the relative safety of LNG-IUD in terms of metabolic effects. Not only did the use of LNG-IUD, either short term or long term, appear to not carry an increased risk of adverse metabolic profiles, but it seemed to be related to a potentially cardioprotective pattern. These results confirm previous evidence, indicating no difference in serum concentrations of total and LDL cholesterol, triglycerides, insulin, glucose, and C-reactive protein ( 34 ,  35 ), or a reduction of total cholesterol, APO-A1, and APO-B ( 34 ,  36 ) in relation to LNG-IUD use. Additionally, findings from a randomized controlled trial ( 36 ) and a population-based cross-sectional study ( 37 ) indicate that short-term LNG-IUD use is associated with a decrease of HDL cholesterol, while long-term use is related to unchanged or increased levels of HDL cholesterol. In future longitudinal studies, the relation of LNG-IUD use and its protective metabolic profile to long-term clinical cardiometabolic events, as well as whether there are long-term beneficial metabolic effects of LNG-IUD use in menopause, should be further addressed.\nThis study has some limitations. The first is related to the small sample size especially in subgroup and longitudinal analyses. Thus, it is likely that the absence of differences in subgroup analyses is due to a lack of statistical power. However, the null findings, eg, in relation to the duration of use of LNG-IUD, are in line with the results of our previous study, conducted in a similar population but with larger groups, suggesting that it is likely to be a true finding. The generalizability of the findings is further limited, considering that the majority of users of LNG-IUD in our population were older than 30 years. Additionally, because only 52 mg LNG-IUD was available in Finland until 2013, our results cannot be generalized to users of low-dose LNG-IUDs (13.5 mg and 19.5 mg). However, for this same reason, our results are free from confounding due to different hormonal doses. Although we did not have complete information on conditions possibly affecting the type and use of contraception and the metabolic status, such as endometriosis and polycystic ovary syndrome, our results were controlled for covariates covering a large set of diseases and confounding conditions. Because of the likely high correlations between several metabolites, interpretation of the implications and importance of single metabolites may have been biased. The data of this study derived mostly from self-administered questionnaires, potentially introducing a recall bias. However, a self-report of contraception use has been shown reliable in specifically focused studies ( 38 ,  39 ). Additionally, reliability of the data was confirmed through consistency throughout multiple related questions. While our previous ( 15 ) and current study are similar in the basic aim of examining the relationship between the use of LNG-IUD and metabolic biomarkers and use the same metabolomics platform, the overall underlying methodology and design are different, as the Health 2000 and Health 2011 include a much more extensive physical examination and comprehensive interview and questionnaire-based data collection. Even though the current study population resulted in a smaller sample size compared to our previous work, the follow-up data allowed us to include a longitudinal design in the current analysis plan.\nStrengths of the study include the large number of available metabolomics measures and the extensive information on possible confounders, such as lifestyle and health characteristics. Additionally, the utilized metabolomic platform provides a rather complete assessment of metabolic status, inclusive of absolute and relative metabolite measures with proven separate biological meaning ( 20 ,  21 ). The clinical significance of the different molecular ratios has been previously shown ( 19-21 ).\nFurther strength of the study is given by the availability of repeated measures for a subgroup of women, which allowed a longitudinal examination of the previously detected associations.\nIn spite of the relatively small sample size and consequently limited power, especially for subgroup and longitudinal analyses, there is a general lack of metabolomics studies on the topic, and the clinical studies that have to date evaluated metabolic outcomes of LNG-IUD are based on similar samples as ours. Additionally, it is not feasible to conduct randomized controlled trials comparing the effects of LNG-IUD vs non-use of HC on clinical cardiovascular outcomes. Thus, although our results have to be taken with caution, they represent important preliminary knowledge in the field that calls for future studies with larger samples.\n\nOur findings confirm previous observations on the use of LNG-IUD being associated with many multiple moderate metabolic changes, mostly suggestive of reduced arterial cardiometabolic risk. These associations appeared mostly to be independent of the duration of use, and changes in LNG-IUD use were not related to long-term metabolic changes. Taken together, our results support the relative metabolic safety of LNG-IUD.","source_license":"CC-BY-4.0","license_restricted":false}