Longitudinal pre-diagnostic samples allow early osteoporosis diagnosis

Longitudinal pre-diagnostic samples allow early osteoporosis diagnosis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Brief Communication Longitudinal pre-diagnostic samples allow early osteoporosis diagnosis Palle Villesen, Johan Lassen, Kirstine Nielsen, Lotte Hindhede, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4642034/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Biomarker discovery for degenerative diseases is challenging due to low statistical power, selection bias, and biological variability. To address these problems, we introduced pre-diagnostic longitudinal sampling using samples from the Danish Blood Donor Study. We obtained up to six longitudinal metabolomics profiles using one-year intervals with the latest profile within one year before osteoporosis diagnosis, including 99 cases and 99 controls. We matched the patients with controls based on sex, age, sampling site, disease history, body mass index, analytical batch, and sample storage time. Our longitudinal model of molecular changes improved the signal from non-significant in single-sample modeling between patient cases and controls to an area under the curve (AUC) of 0.75. This pilot study demonstrates the advantages of longitudinal data in biomarker research, including robustness to day-to-day biological variance, inter-individual variance, and post-diagnostic biases. Biological sciences/Systems biology/Time series Health sciences/Biomarkers/Diagnostic markers Health sciences/Health care/Quality of life Health sciences/Diseases Biological sciences/Biochemistry/Metabolomics Figures Figure 1 Figure 2 Main Early diagnosis of degenerative diseases can improve treatment outcomes and reduce healthcare costs. However, finding reliable biomarkers is difficult due to several factors, such as low statistical power, selection bias, and biological variability 1 . Studies seeking to identify disease biomarkers often use cross-sectional designs, comparing samples from patients and healthy controls at a single time point 2 . This approach has several limitations, including selection bias, which may lead to non-validating studies and is caused by bias in age, lifestyle, medication, or comorbidities. Studies utilizing post-diagnosis samples often introduce biases due to treatment effects, lifestyle changes, or psychological stress, which complicate both the definition of the case group and the construction of suitable control groups. Biomarkers identified in symptomatic patients might reflect non-specific inflammation rather than distinct disease pathology, raising the question of whether control groups should include individuals with other conditions. Finally, day-to-day variability, influenced by factors such as diet, lifestyle, and circadian rhythm, contributes to inter-individual variance, further impeding biomarker discovery. To overcome these challenges, we introduce longitudinal and pre-diagnostic samples from the Danish Blood Donor Study (DBDS), focusing on participants later diagnosed with osteoporosis 3 . The DBDS encompasses over 2.7 million plasma samples from more than 165,000 healthy donors included since 2010, and has previously been used for longitudinal studies 4 . By linking the DBDS data with national health registries, we identified participants who were diagnosed with osteoporosis after their first donation. We then selected their pre-diagnostic samples, as well as samples from matched controls who did not receive an osteoporosis diagnosis within one year of the last donation. We matched cases and controls based on sex, age, sampling site, analytical batch, and plasma sample storage time, excluding participants with certain disease and medication histories before becoming donors. For each participant, we used up to six samples spaced 8 to 14 months apart (Fig. 1 a). The latest sample was collected within one year before diagnosis for the cases, and within the same period for the controls. The demographic characteristics of the participants are summarized in Table S1 . In total, we included 78 participants (39 cases and 39 controls) with longitudinal data and 120 participants (60 cases and 60 controls) with single-sample data. Approximately 80% of the cases continued to donate blood after diagnosis, and around 96% of these were diagnosed without a fracture event. We performed untargeted metabolomics on the plasma samples using liquid chromatography-mass spectrometry (LC-MS) in both positive and negative ionization modes. This approach allowed us to capture the signatures of thousands of metabolites without prior knowledge of their identity or function, enabling a comprehensive and unbiased exploration of the metabolome. The LC-MS metabolomics analysis yielded metabolomic profiles of 3221 and 1429 features in positive and negative ionization after data cleaning and normalization (Fig. S1 - S2 ). Principal component analysis (PCA) of the metabolomics data revealed no strong patterns of osteoporosis in the first two principal components or the following four components (Fig. 1 b and Fig. S3). Consistent with previous findings, we did not expect osteoporosis to be a major source of variance in the metabolome 5 , 6 . Additionally, we found that within-participant variance was on the same scale as between-participant variance in the first two principal components, impeding identification of between-participant differences (Fig. 1 c). Notably, later components explained the between-individual variance in the negative ionization data (Fig. S4). We then built predictive models based on two different strategies: a single-sample model and a longitudinal model. The single-sample model used the most recent sample from each participant (n = 198 participants) and compared the metabolite levels between cases and controls. The longitudinal model used all the available samples from each participant (n = 78 participants, 420 samples) and compared the metabolite changes over time (Fig. S5). The metabolite changes were represented as feature slopes for each individual rather than the relative feature abundance, thus normalizing inter-individual effects on abundance. We used elastic net regression and random forest to evaluate the effects of non-linearity and interactions. The elastic net model performed the best in all settings, but its performance was comparable to random in the single-sample model (Fig. 2 a, Fig. S6). This indicated insignificant metabolite differences between cases and controls at the latest time point. However, the longitudinal elastic net model yielded significant AUCs of 0.75 and 0.68 in negative and positive ionization mode (Fig. 2 b). This suggests that the longitudinal setup identifies metabolites that change over time and captures the dynamic signal of osteoporosis progression (Fig. 2 c-d). The longitudinal elastic net model selected 52 features in positive ionization mode and 42 features in negative ionization mode, having non-zero coefficients (Extended Data 1). We identified and annotated 24 of the selected features across both datasets (Fig. 2 c-d and Extended Data 1). Here, several compounds including amino acids (derivates), bisphenol A, cortisol, and hippuric acid have previously been found to associate with bone metabolism and osteoporosis 7 – 12 . For hippuric acid, we observed a time dependent increase in the heatlhy controls, and a decrease in the cases. This corresponds with the litterature, which states that hippuric acid increases with age but remains low for individuals with degenerative diseases or frailty 11 . Also, hippuric acid has been shown to inhibit osteclast differentiation, resulting in decreased bone resorption and higher bone mineral density (BMD) 12 . Bisphenol A binds to the estrogen receptor and is an exogenous driver of osteoporosis. In the longitudinal setting, the contribition of bisphenol A might be explained by its deposition in adipose fat tissue 8 , 9 , 13 . Hence, depositioned exogenous compounds might serve as biomarkers, if they prove stable to day-to-day variance. Finally, amino acid metabolites have previously been identified in other studies as associated with osteoporosis in agreement with our identified amino acid derivates 6 , 7 , 14 . These findings suggest that the longitudinal model uses biologically relevant features for predicting osteoporosis. Some of the metabolites remained unknown or unconfirmed, and further efforts should be made to elucidate their identity and function. Comprising 78 participants with longitudinal samples and several thousand untargeted molecules, our pilot study faces challenges due to its sample size. Although the vast number of molecules leads to overfitting, we found that the method validates across independent batches in hold-one-batch-out cross-validation with an AUC of 0.68 in the negative ionization dataset (Fig. S7). We attribute the small performance loss to reduced training data when more data is held out. The validation indicates that the method is stable and suggests that more data might improve performance. However, we cannot conclude whether the predictors and models will pass biological validation on new cohorts. The method validation is partially explained by the longitudinal setup, which improves the chances of selecting robustly measured features across individuals. Untargeted metabolomics is susceptible to technical variance (Fig. S1 ), meaning that any method improving stability also enhances the likelihood of identifying novel biomarkers. We found that batch effects and storage time accounted for the most variance in the data, but we regressed out these effects (Fig. S2 ). Other methods, such as proteomics, where platforms like OLINK have matured, might also provide results robust to technical variation 15 . The difficulty in distinguishing between cases and controls could be explained by similar pathogensis progression among controls. A majority of participants in the case group (80%) remained healthy enough to continue to give blood after receiving a diagnosis of osteoporosis. As age increases, BMD decreases, and with a lifetime risk of osteoporosis exceeding 10% in the Danish population, it is likely that some controls also experience osteoporosis progression without being diagnosed 16 , 17 . This suggests that BMD might not differ substantially between cases and controls. Unfortunately, BMD measurements were not available. Given the progressive nature of osteoporosis, classification models cannot represent the true progression state. We suggest interpreting the model probabilities as osteoporosis acceleration scores rather than binary outcomes. Acceleration scores offer a means for individual feedback on healthy aging and reduce the model performance demands. Established markers such as cholesterol and blood pressure guide lifestyle modifications to prevent cardiovascular diseases 18 . Similarly, osteoporosis progression scores could promote physical activity to preserve BMD in the aging population 19 . Despite our thorough study design, the findings may not generalize to high-risk patients assessed in an outpatient clinic. Instead, the identified biomarkers in the current study, which used a design with asymptomatic cases and controls, propose a potential use in screening an asymptomatic population. In such a setting, diseases covarying with osteoporosis might bias the model, meaning disease specific acceleration scores should also rely on appropiate controls with e.g. orthopedic or degenerative diseases. To our knowledge, only one other study has performed deep longitudinal sampling using untargeted metabolomics, aiming to describe aging profiles 20 . Other longitudinal studies have used few timepoints or short sampling periods to understand e.g., diabetes induced polyneuropathy 21 , Alzheimer’s 22 , and general health profiles 23 – albeit with great success, we believe more samples and follow-up years may provide deeper insights into the disease progression. We conclude that this study warrants more research in the Danish Blood Donor Study, which has thousands of participants, some with longitudinal data for over ten years. Methods Participants and data collection Cases included Danish Blood Donor Study participants aged 18–75 years, diagnosed with osteoporosis (ICD10-DM80 and ICD10-DM81), who had given consent for research participation. Eligible participants had at least one sample collected a year prior to their first diagnosis and up to six years of donation history. All samples, both for cases and controls, were collected between 2010 and 2022 and had not undergone any freeze-thaw cycles. See C. Erikstrup et al. for technical details about the cohort, storage, and sample collection 3 . Exclusion criteria included missing data on BMI and smoking status, diagnoses (ICD and ATC codes in Table S2 ) of eating disorders, alcoholism, or gonad dysfunctions, and absence of osteoporosis medication during the donation period. Medications considered included Bisphosphonates (M05BA01, M05BA04, M05BA06, M05BA07), bisphosphonates combinations (M05BB01, M05BB03), strontiumrenalate (M05BX03), denosumab (M05BX04), raloxifene (G03XC01), teriparatide (H05AA02), and parathyroid hormone (Preotact, H05AA01). This resulted in 70 cases, 56 of which continued donation post-diagnosis, and 54 were diagnosed without a pathologic fracture (DM81). Due to resource constraints, 39 longitudinal and 60 single time point cases were selected and paired with matched controls. Controls were chosen based on the absence of osteoporosis diagnosis and matched based on birth year, region of inclusion, sex, and donation date (+/- 60 days) to ensure similar storage times. The study comprised 540 samples and 198 participants in total. Matched pairs were analyzed within the same analytical batches. Ethics DBDS was approved by the Scientific Ethical Committee of Central Denmark (M-20090237) and the Zealand Region (SJ-740) and by the Danish Data Protection Agency (P-2019–99). Chemicals Acetonitrile (LC-MS grade), methanol (LC-MS grade), formic acid (FA, LC-MS grade), amphetamine- d 5 , cocaine- d 3 , diazepam- d 5 , and phenopbarbital- d 5 were acquired from Merck (Darmstadt, Germany). MilliQ water was prepared using a Milli-Q IQ 7000. Plasma sample extraction Plasma sample extraction was performed in 2mL 96 well plates using a Tecan FREEDOM EVO 200 (Tecan Trading AG, Switzerland). In brief, 200µL plasma were added 100µL 0.1M HCl followed by 134 µL of an internal standard mix containing amphetamine- d 5 (60ng/mL), cocaine- d 3 (60ng/mL), diazepam- d 5 (60ng/mL), and phenopbarbital- d 5 (1.2µg/mL) in methanol. The mixture was shaken at 1650 rpm for 30 sec following each addition. Acetonitrile was added three times at a volume of 67 µL and shaken 30 sec in between. Finally, 200µL acetonitrile was added, and the mixture was left to stand for 15 min. The sample was centrifuged at 5000xg for 5 min and 500µL was transferred to an AcroPrep Advance 96 Well 1mL 30K Omega filter plate (Pall Corporation, NY, USA). The filter plate was centrifuged at 2000xg for 5 min and the filtrate was transferred to a 1mL glass insert in a 96 well plate that was evaporated under a stream of nitrogen. The pellet was redissolved in 100µL MilliQ water containing 5% acetonitrile and 0.1% FA. LC-HR-MS analysis The samples were subject to untargeted metabolomics analysis using an ACQUITY I-Class UPLC system (Waters Corporation, Milford, MA, USA) coupled to a Bruker maXis Impact QTOF mass spectrometer (Bruker Daltonics, Bremen, Germany). The mass spectrometer was operated in both positive and negative electrospray ionization (ESI) mode. Chromatographic separation employed an ACQUITY UPLC HSS T3 C18 column (2.1mm x 100mm, 1.8µm, Waters) with gradient elution. Mobile phase A comprised MilliQ water with 0.1% FA, while mobile phase B consisted of acetonitrile with 0.1% FA. The column temperature was maintained at 50°C, with a flow rate of 0.4mL/min. The gradient elution commenced at 0% B (0-2min), followed by linear increases to 40% B (2-6min), 60% B (6-6.5min), 88% B (6.5-11min), 100% B (11-11.5min), held at 100% B (11.5-17min), linear decrease to 0% B (17-18min), and finally, equilibration at 0% B (18-21min). The injection volume was 5µL in both positive and negative ESI. The temperature in the autosampler was kept at 6°C. MS scans were conducted in full scan mode at a sampling rate of 4Hz across a mass range of 50-1000 m/z. The nebulizing gas pressure was maintained at 1.2 bar, with a capillary voltage of 4.0 kV in positive ESI and 2.5 kV in negative ESI. The drying gas flow rate was 8.0L/min at a temperature of 220°C. MS/MS scans using data-dependent acquisition (auto-MS/MS) followed similar parameters, with collision energies set at 10, 20 or 30 eV and a sampling rate of 10Hz. Internal calibration was conducted at the conclusion of each run using sodium formate in both ionization modes. The samples were extracted and analyzed in seven batches, with all time-series samples from the same participant analyzed within the same batch. Matched cases and controls were additionally analyzed within the same batch. A control sample containing 37 standard compounds was analyzed in the beginning and the end of each batch for quality assessment of instrument performance. The system was equilibrated with four injections of a separate blood sample not part of the study (QC-sample). The QC-sample was additionally injected for every 10th sample during each batch to ensure instrument stability during the run. Samples within a batch were analyzed in randomized order. Eight samples within a batch were analyzed twice as short technical replicates. Additionally, four samples from a previous batch were analyzed within each new batch as batch replicates. A sample containing 20 different pooled samples from the study were used for MS/MS fragmentation analysis for annotation/identification of metabolites. Pre-processing of LC-MS data The mass spectrometry data from UPLC-HR-QTOF were converted to mzML . file format using msConvert from ProteoWizard ( http://proteowizard.sourceforge.net ). XCMS was used to create the feature table 24 . Here, Centwave (ppm = 12, snthresh = 6, peakwidth = 4–25) was used for peak identification, peak density (min fraction = 0.3) was used for peak grouping and CAMERA was used for isotope annotation. Metid (part of tidymass R package 25 ) was used to automatically annotate the MS/MS scans, at level 2 identification according to the Metabolomics standard initiative 26 . The following databases were used: Massbank, Mona, NIST, msdatabase (Snyder lab), HMDB, orbitrap, and the Fiehn Hilic database. Automatic matches were manually evaluated and compared against an in-house database containing approximately 600 compounds for level 1 identification when available. Statistical preprocessing The data was log-transformed, and features with more than 10% missing values were removed. Next, features were processed with robust row normalization as described in J. Lassen et al 27 . The features were then normalized using the setup from Richie et al. 28 except all corrections were done simultaneously using the following model: $$In{t}_{conf}=Region+Batch+InjOrder+Storage+batch\bullet InjOrder+batch\bullet Storage$$ $$Intensit{y}_{corrected}=Intensity-In{t}_{conf}$$ Where \(In{t}_{conf}\) is the intensity explained by confounders, \(InjOrder\) is the injection order during an LC-MS batch, and \(Intensit{y}_{corrected}\) is the corrected intensity after regressing out the confounding variables. After normalization, features were removed based on technical replication correlations (remove if 0.1). For positive ionization, 8851 features were removed, and for negative ionization, 3047 features were removed. The slopes of the longitudinal data were calculated for all selected features in each participant where technical replicates were included. Modeling Two models were trained: one based on the latest donation before diagnosis (or no diagnosis) and the other based on the feature slopes calculated from the longitudinal data. Caret was used for modeling. For the latest donation model custom cross-validation splits for leave-one-participant-out validation were used to avoid data leakage. An elastic net logistic regression was fitted using the glmnet model with a TuneLength of 5 (hyperparameter tuning). Random forest was also evaluated using a TuneLength of 5. The hyperparameter set returning the best predictions was used to assess the model performance and train a final model for feature importance estimation. The same setup was carried out for the slope model, again using elastic net and random forest at a TuneLength of 5. The slope model was also evaluated with leave-one-participant-out cross-validation. Important features of the elastic net logistic regression model were extracted from Caret. To build null distributions of the model performance, each dataset was permuted, hyperparameter tuned ( TuneLength 5), and evaluated using elastic net logistic regression (the best performing model). The performance exceeds 0.5 (AUC) by random because the best performing model is always selected in hyperparameter tuning. In total the procedure was performed 100 times to calculate a mean and a 95% confidence interval of the null distribution in the ROC curve. References Moqri M et al (2024) Validation of biomarkers of aging. Nat Med 30:360–372 Mann CJ (2003) Observational research methods. Research design II: cohort, cross sectional, and case-control studies. Emerg Med J 20:54 Erikstrup C et al (2023) Cohort Profile: The Danish Blood Donor Study. Int J Epidemiol 52:e162–e171 Jørgensen S et al (2020) The value of circulating microRNAs for early diagnosis of B-cell lymphoma: A case-control study on historical samples. Sci Rep 10:9637 Zhang X et al (2021) Metabolomics Insights into Osteoporosis Through Association With Bone Mineral Density. J Bone Miner Res 36:729–738 Wang J et al (2019) Discovery of potential biomarkers for osteoporosis using LC-MS/MS metabolomic methods. Osteoporos Int 30:1491–1499 Yang T-L et al (2020) A road map for understanding molecular and genetic determinants of osteoporosis. Nat Reviews Endocrinol 16:91–103 Hwang JK et al (2013) Bisphenol A reduces differentiation and stimulates apoptosis of osteoclasts and osteoblasts. Life Sci 93:367–372 Maduranga Karunarathne WAH, Choi YH, Park SR, Lee C-M, Kim G-Y (2022) Bisphenol A inhibits osteogenic activity and causes bone resorption via the activation of retinoic acid-related orphan receptor α. J Hazard Mater 438:129458 Osella G et al (2012) Cortisol secretion, bone health, and bone loss: a cross-sectional and prospective study in normal nonosteoporotic women in the early postmenopausal period. Eur J Endocrinol 166:855–860 De Simone G, Balducci C, Forloni G, Pastorelli R, Brunelli L (2021) Hippuric acid: Could became a barometer for frailty and geriatric syndromes? Ageing Res Rev 72:101466 Chen J-R et al (2021) GPR109A mediates the effects of hippuric acid on regulating osteoclastogenesis and bone resorption in mice. Commun Biology 4:53 Wang L, Asimakopoulos AG, Kannan K (2015) Accumulation of 19 environmental phenolic and xenobiotic heterocyclic aromatic compounds in human adipose tissue. Environ Int 78:45–50 Panahi N et al (2022) Association of amino acid metabolites with osteoporosis, a metabolomic approach: Bushehr elderly health program. Metabolomics 18:63 Eldjarn GH et al (2023) Large-scale plasma proteomics comparisons through genetics and disease associations. Nature 622:348–358 Warming L, Hassager C, Christiansen C (2002) Changes in Bone Mineral Density with Age in Men and Women: A Longitudinal Study. Osteoporos Int 13:105–112 Willers C et al (2022) Osteoporosis in Europe: a compendium of country-specific reports. Archives Osteoporos 17:23 D’Agostino RB et al (2008) General Cardiovascular Risk Profile for Use in Primary Care. Circulation 117:743–753 Kai MC, Anderson M, Lau EM (2003) Exercise interventions: defusing the world's osteoporosis time bomb. Bull World Health Organ 81:827–830 Ahadi S et al (2020) Personal aging markers and ageotypes revealed by deep longitudinal profiling. Nat Med 26:83–90 Määttä LL et al (2024) Longitudinal Change in Serum Neurofilament Light Chain in Type 2 Diabetes and Early Diabetic Polyneuropathy: ADDITION-Denmark. Diabetes Care 47:986–994 Libiger O et al (2021) Longitudinal CSF proteomics identifies NPTX2 as a prognostic biomarker of Alzheimer's disease. Alzheimer's Dement 17:1976–1987 Dodig-Crnković T et al (2020) Facets of individual-specific health signatures determined from longitudinal plasma proteome profiling. eBioMedicine 57 Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak GXCMS (2006) Processing Mass Spectrometry Data for Metabolite Profiling Using Nonlinear Peak Alignment, Matching, and Identification. Anal Chem 78:779–787 Shen X et al (2022) TidyMass an object-oriented reproducible analysis framework for LC–MS data. Nat Commun 13:4365 Sumner LW et al (2007) Proposed minimum reporting standards for chemical analysis. Metabolomics 3:211–221 Lassen J, Nielsen KL, Johannsen M, Villesen P (2021) Assessment of XCMS Optimization Methods with Machine-Learning Performance. Anal Chem 93:13459–13466 Ritchie SC et al (2023) Quality control and removal of technical variation of NMR metabolic biomarker data in ~ 120,000 UK Biobank participants. Sci Data 10:64 Additional Declarations There is NO Competing Interest. Supplementary Files ExtendedData17.06.2024.xlsx supplementaryfigurestables17.06.2024.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4642034","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Brief Communication","associatedPublications":[],"authors":[{"id":320262441,"identity":"a73ccae6-5b5f-48fb-bc05-adb191d1ee8b","order_by":0,"name":"Palle Villesen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYFAC5sYDDBUHeGA8AyK0MDYcYDhDshbGtgNwOwlrMWc/2HDg47w7MubtBxg/8zBYGxPUYtmT2HBw5rZnPDJnEpileRjSzQhqMTiQ2HCYd9thHgkJBgaglsM2hLWcf9hw+O8csBbm38RpuQG0hbEBrIUNZAthh1nOeNhwsOfYMx4JnsQ2yzkG6YS9b86ffPDBj5o79hLshw/feFNhbdhA0GEIJmMDCpcYLaNgFIyCUTAKcAAABH08Fow6OMsAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-7589-8781","institution":"Aarhus University","correspondingAuthor":true,"prefix":"","firstName":"Palle","middleName":"","lastName":"Villesen","suffix":""},{"id":320262442,"identity":"48178269-6f90-4334-a2dc-5f99c96e3dd6","order_by":1,"name":"Johan Lassen","email":"","orcid":"","institution":"Aarhus University","correspondingAuthor":false,"prefix":"","firstName":"Johan","middleName":"","lastName":"Lassen","suffix":""},{"id":320262443,"identity":"8c083fb5-cbc4-4adf-9a54-76444e04a775","order_by":2,"name":"Kirstine Nielsen","email":"","orcid":"","institution":"Aarhus University","correspondingAuthor":false,"prefix":"","firstName":"Kirstine","middleName":"","lastName":"Nielsen","suffix":""},{"id":320262444,"identity":"0c4d32f7-f6f6-40fc-a6db-9a40ae9c341d","order_by":3,"name":"Lotte Hindhede","email":"","orcid":"","institution":"Aarhus University","correspondingAuthor":false,"prefix":"","firstName":"Lotte","middleName":"","lastName":"Hindhede","suffix":""},{"id":320262445,"identity":"504e7e07-cee0-4282-9884-24d5cbab5e4e","order_by":4,"name":"Susan Mikkelsen","email":"","orcid":"","institution":"Department of Clinical Immunology, Aarhus University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Susan","middleName":"","lastName":"Mikkelsen","suffix":""},{"id":320262446,"identity":"1dac81a3-47c7-4bcb-b434-fc82c353e654","order_by":5,"name":"Bertram Kjerulff","email":"","orcid":"https://orcid.org/0000-0001-9953-9521","institution":"Aarhus University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bertram","middleName":"","lastName":"Kjerulff","suffix":""},{"id":320262447,"identity":"2d913c12-5784-4c15-b2eb-518e2a5c664a","order_by":6,"name":"Sisse Ostrowski","email":"","orcid":"https://orcid.org/0000-0001-5288-3851","institution":"Rigshospitalet, Copenhagen University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Sisse","middleName":"","lastName":"Ostrowski","suffix":""},{"id":320262448,"identity":"c570bb5c-c10e-4745-959e-7daba63b3d35","order_by":7,"name":"Erik Sørensen","email":"","orcid":"","institution":"Department of Clinical Immunology, Copenhagen University Hospital, Rigshospitalet","correspondingAuthor":false,"prefix":"","firstName":"Erik","middleName":"","lastName":"Sørensen","suffix":""},{"id":320262449,"identity":"a81a536b-53e6-491e-a9f3-18b2ef8dd9b7","order_by":8,"name":"Christina Mikkelsen","email":"","orcid":"https://orcid.org/0000-0002-2945-6197","institution":"Copenhagen University Hospital Rigshospitalet","correspondingAuthor":false,"prefix":"","firstName":"Christina","middleName":"","lastName":"Mikkelsen","suffix":""},{"id":320262450,"identity":"d7e0701d-c4dd-45e8-8baa-16f87aba7200","order_by":9,"name":"Ole Pedersen","email":"","orcid":"https://orcid.org/0000-0003-2312-5976","institution":"Aarhus University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ole","middleName":"","lastName":"Pedersen","suffix":""},{"id":320262451,"identity":"fa19408e-f820-46ee-82a7-7dbd4fe74841","order_by":10,"name":"Mie Topholm Bruun","email":"","orcid":"https://orcid.org/0000-0002-8819-5388","institution":"Odense University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Mie","middleName":"Topholm","lastName":"Bruun","suffix":""},{"id":320262452,"identity":"a9c735a3-8c19-4739-9133-5a8bcec944cb","order_by":11,"name":"Bitten Aagaard","email":"","orcid":"","institution":"Department of Clinical Immunology, Aalborg Univeristy Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bitten","middleName":"","lastName":"Aagaard","suffix":""},{"id":320262453,"identity":"5794bfd1-06aa-42bf-8633-d8cc94f79e18","order_by":12,"name":"Mogens Johannsen","email":"","orcid":"","institution":"Aarhus University","correspondingAuthor":false,"prefix":"","firstName":"Mogens","middleName":"","lastName":"Johannsen","suffix":""},{"id":320262454,"identity":"5bf87ee7-5c91-4c26-a7d0-a0c3432d61db","order_by":13,"name":"Christian Erikstrup","email":"","orcid":"https://orcid.org/0000-0001-6551-6647","institution":"Department of Clinical Immunology, Aarhus University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Christian","middleName":"","lastName":"Erikstrup","suffix":""}],"badges":[],"createdAt":"2024-06-26 10:35:41","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-4642034/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4642034/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":61346795,"identity":"882a752f-f9bf-4b34-aa36-9368a81a8071","added_by":"auto","created_at":"2024-07-29 18:11:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":82771,"visible":true,"origin":"","legend":"\u003cp\u003eSampling and major sources of variance. \u003cstrong\u003ea\u003c/strong\u003e. Going back in time and selecting samples from participants later diagnosed with osteoporosis and matched controls. Adjacent lines represent matched pairs of cases and controls and their donation history used in this study. \u003cstrong\u003eb.\u003c/strong\u003e PCA of the negative ionization data and the outcome illustrates that osteoporosis is not a major contributor to variance, see Fig. S3 for PC1-6 in both ionization modes. \u003cstrong\u003ec.\u003c/strong\u003e PCA of the negative ionization data and three randomly selected participants illustrate that individual variance is on the same scale as global variance. The lines depict the sequence of donations.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4642034/v1/5616c1ab2d7c4a7ebc0798bb.png"},{"id":61346796,"identity":"69872181-27a5-42b8-a082-8295092dd2a3","added_by":"auto","created_at":"2024-07-29 18:11:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":95899,"visible":true,"origin":"","legend":"\u003cp\u003eModeling diagnosis within one year of the most recent sample \u003cstrong\u003ea\u003c/strong\u003e. Leave-one-participant-out cross-validation yields insignificant performance in single-sample elastic net modeling. The gray area depicts the null distribution of the best hyperparameter-tuned models based on permuted data. \u003cstrong\u003eb\u003c/strong\u003e. The longitudinal model detects a signal in the molecule slopes and predicts osteoporosis one year in advance of diagnosis. \u003cstrong\u003ec-d\u003c/strong\u003e. Top 5 important features in each ionization dataset of the longitudinal models in panel b.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4642034/v1/8a11e106478956b33ea62dcb.png"},{"id":101239187,"identity":"8c06a34e-40d9-4296-b1d0-e8ff2317c8ae","added_by":"auto","created_at":"2026-01-27 15:12:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":625876,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4642034/v1/b2c8dc63-092b-4e39-82c2-7c675e1636ff.pdf"},{"id":61347975,"identity":"e6c9aa62-4907-479b-8823-07d2a66f42d5","added_by":"auto","created_at":"2024-07-29 18:19:20","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":565437,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"ExtendedData17.06.2024.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4642034/v1/24620c2615c389549c3bfc42.xlsx"},{"id":61346798,"identity":"6e281cc2-4eb7-4ea3-a388-394773d520eb","added_by":"auto","created_at":"2024-07-29 18:11:20","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":3338797,"visible":true,"origin":"","legend":"","description":"","filename":"supplementaryfigurestables17.06.2024.docx","url":"https://assets-eu.researchsquare.com/files/rs-4642034/v1/63fb59e23e7ccd53999eff5e.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Longitudinal pre-diagnostic samples allow early osteoporosis diagnosis","fulltext":[{"header":"Main","content":"\u003cp\u003eEarly diagnosis of degenerative diseases can improve treatment outcomes and reduce healthcare costs. However, finding reliable biomarkers is difficult due to several factors, such as low statistical power, selection bias, and biological variability\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Studies seeking to identify disease biomarkers often use cross-sectional designs, comparing samples from patients and healthy controls at a single time point\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. This approach has several limitations, including selection bias, which may lead to non-validating studies and is caused by bias in age, lifestyle, medication, or comorbidities. Studies utilizing post-diagnosis samples often introduce biases due to treatment effects, lifestyle changes, or psychological stress, which complicate both the definition of the case group and the construction of suitable control groups. Biomarkers identified in symptomatic patients might reflect non-specific inflammation rather than distinct disease pathology, raising the question of whether control groups should include individuals with other conditions. Finally, day-to-day variability, influenced by factors such as diet, lifestyle, and circadian rhythm, contributes to inter-individual variance, further impeding biomarker discovery.\u003c/p\u003e \u003cp\u003eTo overcome these challenges, we introduce longitudinal and pre-diagnostic samples from the Danish Blood Donor Study (DBDS), focusing on participants later diagnosed with osteoporosis\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. The DBDS encompasses over 2.7\u0026nbsp;million plasma samples from more than 165,000 healthy donors included since 2010, and has previously been used for longitudinal studies\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. By linking the DBDS data with national health registries, we identified participants who were diagnosed with osteoporosis after their first donation. We then selected their pre-diagnostic samples, as well as samples from matched controls who did not receive an osteoporosis diagnosis within one year of the last donation. We matched cases and controls based on sex, age, sampling site, analytical batch, and plasma sample storage time, excluding participants with certain disease and medication histories before becoming donors. For each participant, we used up to six samples spaced 8 to 14 months apart (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The latest sample was collected within one year before diagnosis for the cases, and within the same period for the controls. The demographic characteristics of the participants are summarized in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eIn total, we included 78 participants (39 cases and 39 controls) with longitudinal data and 120 participants (60 cases and 60 controls) with single-sample data. Approximately 80% of the cases continued to donate blood after diagnosis, and around 96% of these were diagnosed without a fracture event. We performed untargeted metabolomics on the plasma samples using liquid chromatography-mass spectrometry (LC-MS) in both positive and negative ionization modes. This approach allowed us to capture the signatures of thousands of metabolites without prior knowledge of their identity or function, enabling a comprehensive and unbiased exploration of the metabolome.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe LC-MS metabolomics analysis yielded metabolomic profiles of 3221 and 1429 features in positive and negative ionization after data cleaning and normalization (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-\u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). Principal component analysis (PCA) of the metabolomics data revealed no strong patterns of osteoporosis in the first two principal components or the following four components (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb and Fig. S3). Consistent with previous findings, we did not expect osteoporosis to be a major source of variance in the metabolome\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Additionally, we found that within-participant variance was on the same scale as between-participant variance in the first two principal components, impeding identification of between-participant differences (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec). Notably, later components explained the between-individual variance in the negative ionization data (Fig. S4).\u003c/p\u003e \u003cp\u003eWe then built predictive models based on two different strategies: a single-sample model and a longitudinal model. The single-sample model used the most recent sample from each participant (n\u0026thinsp;=\u0026thinsp;198 participants) and compared the metabolite levels between cases and controls. The longitudinal model used all the available samples from each participant (n\u0026thinsp;=\u0026thinsp;78 participants, 420 samples) and compared the metabolite changes over time (Fig. S5). The metabolite changes were represented as feature slopes for each individual rather than the relative feature abundance, thus normalizing inter-individual effects on abundance.\u003c/p\u003e \u003cp\u003eWe used elastic net regression and random forest to evaluate the effects of non-linearity and interactions. The elastic net model performed the best in all settings, but its performance was comparable to random in the single-sample model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea, Fig. S6). This indicated insignificant metabolite differences between cases and controls at the latest time point. However, the longitudinal elastic net model yielded significant AUCs of 0.75 and 0.68 in negative and positive ionization mode (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). This suggests that the longitudinal setup identifies metabolites that change over time and captures the dynamic signal of osteoporosis progression (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec-d).\u003c/p\u003e \u003cp\u003eThe longitudinal elastic net model selected 52 features in positive ionization mode and 42 features in negative ionization mode, having non-zero coefficients (Extended Data 1). We identified and annotated 24 of the selected features across both datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec-d and Extended Data 1). Here, several compounds including amino acids (derivates), bisphenol A, cortisol, and hippuric acid have previously been found to associate with bone metabolism and osteoporosis\u003csup\u003e\u003cspan additionalcitationids=\"CR8 CR9 CR10 CR11\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFor hippuric acid, we observed a time dependent increase in the heatlhy controls, and a decrease in the cases. This corresponds with the litterature, which states that hippuric acid increases with age but remains low for individuals with degenerative diseases or frailty\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Also, hippuric acid has been shown to inhibit osteclast differentiation, resulting in decreased bone resorption and higher bone mineral density (BMD)\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Bisphenol A binds to the estrogen receptor and is an exogenous driver of osteoporosis. In the longitudinal setting, the contribition of bisphenol A might be explained by its deposition in adipose fat tissue\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Hence, depositioned exogenous compounds might serve as biomarkers, if they prove stable to day-to-day variance. Finally, amino acid metabolites have previously been identified in other studies as associated with osteoporosis in agreement with our identified amino acid derivates\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThese findings suggest that the longitudinal model uses biologically relevant features for predicting osteoporosis. Some of the metabolites remained unknown or unconfirmed, and further efforts should be made to elucidate their identity and function.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eComprising 78 participants with longitudinal samples and several thousand untargeted molecules, our pilot study faces challenges due to its sample size. Although the vast number of molecules leads to overfitting, we found that the method validates across independent batches in hold-one-batch-out cross-validation with an AUC of 0.68 in the negative ionization dataset (Fig. S7). We attribute the small performance loss to reduced training data when more data is held out. The validation indicates that the method is stable and suggests that more data might improve performance. However, we cannot conclude whether the predictors and models will pass biological validation on new cohorts.\u003c/p\u003e \u003cp\u003eThe method validation is partially explained by the longitudinal setup, which improves the chances of selecting robustly measured features across individuals. Untargeted metabolomics is susceptible to technical variance (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), meaning that any method improving stability also enhances the likelihood of identifying novel biomarkers. We found that batch effects and storage time accounted for the most variance in the data, but we regressed out these effects (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). Other methods, such as proteomics, where platforms like OLINK have matured, might also provide results robust to technical variation\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe difficulty in distinguishing between cases and controls could be explained by similar pathogensis progression among controls. A majority of participants in the case group (80%) remained healthy enough to continue to give blood after receiving a diagnosis of osteoporosis. As age increases, BMD decreases, and with a lifetime risk of osteoporosis exceeding 10% in the Danish population, it is likely that some controls also experience osteoporosis progression without being diagnosed\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. This suggests that BMD might not differ substantially between cases and controls. Unfortunately, BMD measurements were not available.\u003c/p\u003e \u003cp\u003eGiven the progressive nature of osteoporosis, classification models cannot represent the true progression state. We suggest interpreting the model probabilities as osteoporosis acceleration scores rather than binary outcomes. Acceleration scores offer a means for individual feedback on healthy aging and reduce the model performance demands. Established markers such as cholesterol and blood pressure guide lifestyle modifications to prevent cardiovascular diseases\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Similarly, osteoporosis progression scores could promote physical activity to preserve BMD in the aging population\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite our thorough study design, the findings may not generalize to high-risk patients assessed in an outpatient clinic. Instead, the identified biomarkers in the current study, which used a design with asymptomatic cases and controls, propose a potential use in screening an asymptomatic population. In such a setting, diseases covarying with osteoporosis might bias the model, meaning disease specific acceleration scores should also rely on appropiate controls with e.g. orthopedic or degenerative diseases.\u003c/p\u003e \u003cp\u003eTo our knowledge, only one other study has performed deep longitudinal sampling using untargeted metabolomics, aiming to describe aging profiles\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Other longitudinal studies have used few timepoints or short sampling periods to understand e.g., diabetes induced polyneuropathy\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, Alzheimer\u0026rsquo;s\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, and general health profiles\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e \u0026ndash; albeit with great success, we believe more samples and follow-up years may provide deeper insights into the disease progression. We conclude that this study warrants more research in the Danish Blood Donor Study, which has thousands of participants, some with longitudinal data for over ten years.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants and data collection\u003c/h2\u003e \u003cp\u003eCases included Danish Blood Donor Study participants aged 18\u0026ndash;75 years, diagnosed with osteoporosis (ICD10-DM80 and ICD10-DM81), who had given consent for research participation. Eligible participants had at least one sample collected a year prior to their first diagnosis and up to six years of donation history. All samples, both for cases and controls, were collected between 2010 and 2022 and had not undergone any freeze-thaw cycles. See C. Erikstrup et al. for technical details about the cohort, storage, and sample collection\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eExclusion criteria included missing data on BMI and smoking status, diagnoses (ICD and ATC codes in Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e) of eating disorders, alcoholism, or gonad dysfunctions, and absence of osteoporosis medication during the donation period. Medications considered included Bisphosphonates (M05BA01, M05BA04, M05BA06, M05BA07), bisphosphonates combinations (M05BB01, M05BB03), strontiumrenalate (M05BX03), denosumab (M05BX04), raloxifene (G03XC01), teriparatide (H05AA02), and parathyroid hormone (Preotact, H05AA01). This resulted in 70 cases, 56 of which continued donation post-diagnosis, and 54 were diagnosed without a pathologic fracture (DM81).\u003c/p\u003e \u003cp\u003eDue to resource constraints, 39 longitudinal and 60 single time point cases were selected and paired with matched controls. Controls were chosen based on the absence of osteoporosis diagnosis and matched based on birth year, region of inclusion, sex, and donation date (+/- 60 days) to ensure similar storage times. The study comprised 540 samples and 198 participants in total. Matched pairs were analyzed within the same analytical batches.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eEthics\u003c/h2\u003e \u003cp\u003e DBDS was approved by the Scientific Ethical Committee of Central Denmark (M-20090237) and the Zealand Region (SJ-740) and by the Danish Data Protection Agency (P-2019\u0026ndash;99).\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003eChemicals\u003c/h2\u003e \u003cp\u003eAcetonitrile (LC-MS grade), methanol (LC-MS grade), formic acid (FA, LC-MS grade), amphetamine-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e, cocaine-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e3\u003c/em\u003e\u003c/sub\u003e, diazepam-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e, and phenopbarbital-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e were acquired from Merck (Darmstadt, Germany). MilliQ water was prepared using a Milli-Q IQ 7000.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003ePlasma sample extraction\u003c/h2\u003e \u003cp\u003ePlasma sample extraction was performed in 2mL 96 well plates using a Tecan FREEDOM EVO 200 (Tecan Trading AG, Switzerland). In brief, 200\u0026micro;L plasma were added 100\u0026micro;L 0.1M HCl followed by 134 \u0026micro;L of an internal standard mix containing amphetamine-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e (60ng/mL), cocaine-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e3\u003c/em\u003e\u003c/sub\u003e (60ng/mL), diazepam-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e (60ng/mL), and phenopbarbital-\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e (1.2\u0026micro;g/mL) in methanol. The mixture was shaken at 1650 rpm for 30 sec following each addition. Acetonitrile was added three times at a volume of 67 \u0026micro;L and shaken 30 sec in between. Finally, 200\u0026micro;L acetonitrile was added, and the mixture was left to stand for 15 min. The sample was centrifuged at 5000xg for 5 min and 500\u0026micro;L was transferred to an AcroPrep Advance 96 Well 1mL 30K Omega filter plate (Pall Corporation, NY, USA). The filter plate was centrifuged at 2000xg for 5 min and the filtrate was transferred to a 1mL glass insert in a 96 well plate that was evaporated under a stream of nitrogen. The pellet was redissolved in 100\u0026micro;L MilliQ water containing 5% acetonitrile and 0.1% FA.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eLC-HR-MS analysis\u003c/h2\u003e \u003cp\u003eThe samples were subject to untargeted metabolomics analysis using an ACQUITY I-Class UPLC system (Waters Corporation, Milford, MA, USA) coupled to a Bruker maXis Impact QTOF mass spectrometer (Bruker Daltonics, Bremen, Germany). The mass spectrometer was operated in both positive and negative electrospray ionization (ESI) mode. Chromatographic separation employed an ACQUITY UPLC HSS T3 C18 column (2.1mm x 100mm, 1.8\u0026micro;m, Waters) with gradient elution. Mobile phase A comprised MilliQ water with 0.1% FA, while mobile phase B consisted of acetonitrile with 0.1% FA. The column temperature was maintained at 50\u0026deg;C, with a flow rate of 0.4mL/min. The gradient elution commenced at 0% B (0-2min), followed by linear increases to 40% B (2-6min), 60% B (6-6.5min), 88% B (6.5-11min), 100% B (11-11.5min), held at 100% B (11.5-17min), linear decrease to 0% B (17-18min), and finally, equilibration at 0% B (18-21min). The injection volume was 5\u0026micro;L in both positive and negative ESI. The temperature in the autosampler was kept at 6\u0026deg;C.\u003c/p\u003e \u003cp\u003eMS scans were conducted in full scan mode at a sampling rate of 4Hz across a mass range of 50-1000 m/z. The nebulizing gas pressure was maintained at 1.2 bar, with a capillary voltage of 4.0 kV in positive ESI and 2.5 kV in negative ESI. The drying gas flow rate was 8.0L/min at a temperature of 220\u0026deg;C. MS/MS scans using data-dependent acquisition (auto-MS/MS) followed similar parameters, with collision energies set at 10, 20 or 30 eV and a sampling rate of 10Hz. Internal calibration was conducted at the conclusion of each run using sodium formate in both ionization modes.\u003c/p\u003e \u003cp\u003eThe samples were extracted and analyzed in seven batches, with all time-series samples from the same participant analyzed within the same batch. Matched cases and controls were additionally analyzed within the same batch. A control sample containing 37 standard compounds was analyzed in the beginning and the end of each batch for quality assessment of instrument performance. The system was equilibrated with four injections of a separate blood sample not part of the study (QC-sample). The QC-sample was additionally injected for every 10th sample during each batch to ensure instrument stability during the run. Samples within a batch were analyzed in randomized order. Eight samples within a batch were analyzed twice as short technical replicates. Additionally, four samples from a previous batch were analyzed within each new batch as batch replicates.\u003c/p\u003e \u003cp\u003eA sample containing 20 different pooled samples from the study were used for MS/MS fragmentation analysis for annotation/identification of metabolites.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003ePre-processing of LC-MS data\u003c/h2\u003e \u003cp\u003eThe mass spectrometry data from UPLC-HR-QTOF were converted to \u003cem\u003emzML\u003c/em\u003e. file format using msConvert from ProteoWizard (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://proteowizard.sourceforge.net\u003c/span\u003e\u003cspan address=\"http://proteowizard.sourceforge.net\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). XCMS was used to create the feature table\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Here, Centwave (ppm\u0026thinsp;=\u0026thinsp;12, snthresh\u0026thinsp;=\u0026thinsp;6, peakwidth\u0026thinsp;=\u0026thinsp;4\u0026ndash;25) was used for peak identification, peak density (min fraction\u0026thinsp;=\u0026thinsp;0.3) was used for peak grouping and CAMERA was used for isotope annotation. Metid (part of tidymass R package\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e) was used to automatically annotate the MS/MS scans, at level 2 identification according to the Metabolomics standard initiative\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. The following databases were used: Massbank, Mona, NIST, msdatabase (Snyder lab), HMDB, orbitrap, and the Fiehn Hilic database. Automatic matches were manually evaluated and compared against an in-house database containing approximately 600 compounds for level 1 identification when available.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eStatistical preprocessing\u003c/h2\u003e \u003cp\u003eThe data was log-transformed, and features with more than 10% missing values were removed. Next, features were processed with robust row normalization as described in J. Lassen et al\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. The features were then normalized using the setup from Richie et al.\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e except all corrections were done simultaneously using the following model:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$In{t}_{conf}=Region+Batch+InjOrder+Storage+batch\\bullet InjOrder+batch\\bullet Storage$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$Intensit{y}_{corrected}=Intensity-In{t}_{conf}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(In{t}_{conf}\\)\u003c/span\u003e\u003c/span\u003e is the intensity explained by confounders, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(InjOrder\\)\u003c/span\u003e\u003c/span\u003e is the injection order during an LC-MS batch, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(Intensit{y}_{corrected}\\)\u003c/span\u003e\u003c/span\u003e is the corrected intensity after regressing out the confounding variables.\u003c/p\u003e \u003cp\u003eAfter normalization, features were removed based on technical replication correlations (remove if\u0026thinsp;\u0026lt;\u0026thinsp;0.8) and quality control relative standard deviation (RSD) (remove if\u0026thinsp;\u0026gt;\u0026thinsp;0.1). For positive ionization, 8851 features were removed, and for negative ionization, 3047 features were removed. The slopes of the longitudinal data were calculated for all selected features in each participant where technical replicates were included.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eModeling\u003c/h2\u003e \u003cp\u003eTwo models were trained: one based on the latest donation before diagnosis (or no diagnosis) and the other based on the feature slopes calculated from the longitudinal data. Caret was used for modeling. For the latest donation model custom cross-validation splits for leave-one-participant-out validation were used to avoid data leakage. An elastic net logistic regression was fitted using the glmnet model with a \u003cem\u003eTuneLength\u003c/em\u003e of 5 (hyperparameter tuning). Random forest was also evaluated using a \u003cem\u003eTuneLength\u003c/em\u003e of 5. The hyperparameter set returning the best predictions was used to assess the model performance and train a final model for feature importance estimation. The same setup was carried out for the slope model, again using elastic net and random forest at a \u003cem\u003eTuneLength\u003c/em\u003e of 5. The slope model was also evaluated with leave-one-participant-out cross-validation. Important features of the elastic net logistic regression model were extracted from Caret.\u003c/p\u003e \u003cp\u003eTo build null distributions of the model performance, each dataset was permuted, hyperparameter tuned (\u003cem\u003eTuneLength\u003c/em\u003e 5), and evaluated using elastic net logistic regression (the best performing model). The performance exceeds 0.5 (AUC) by random because the best performing model is always selected in hyperparameter tuning. In total the procedure was performed 100 times to calculate a mean and a 95% confidence interval of the null distribution in the ROC curve.\u003c/p\u003e \u003c/div\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMoqri M et al (2024) Validation of biomarkers of aging. Nat Med 30:360\u0026ndash;372\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMann CJ (2003) Observational research methods. Research design II: cohort, cross sectional, and case-control studies. Emerg Med J 20:54\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErikstrup C et al (2023) Cohort Profile: The Danish Blood Donor Study. Int J Epidemiol 52:e162\u0026ndash;e171\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJ\u0026oslash;rgensen S et al (2020) The value of circulating microRNAs for early diagnosis of B-cell lymphoma: A case-control study on historical samples. Sci Rep 10:9637\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang X et al (2021) Metabolomics Insights into Osteoporosis Through Association With Bone Mineral Density. J Bone Miner Res 36:729\u0026ndash;738\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J et al (2019) Discovery of potential biomarkers for osteoporosis using LC-MS/MS metabolomic methods. Osteoporos Int 30:1491\u0026ndash;1499\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang T-L et al (2020) A road map for understanding molecular and genetic determinants of osteoporosis. Nat Reviews Endocrinol 16:91\u0026ndash;103\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHwang JK et al (2013) Bisphenol A reduces differentiation and stimulates apoptosis of osteoclasts and osteoblasts. Life Sci 93:367\u0026ndash;372\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaduranga Karunarathne WAH, Choi YH, Park SR, Lee C-M, Kim G-Y (2022) Bisphenol A inhibits osteogenic activity and causes bone resorption via the activation of retinoic acid-related orphan receptor α. J Hazard Mater 438:129458\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOsella G et al (2012) Cortisol secretion, bone health, and bone loss: a cross-sectional and prospective study in normal nonosteoporotic women in the early postmenopausal period. Eur J Endocrinol 166:855\u0026ndash;860\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Simone G, Balducci C, Forloni G, Pastorelli R, Brunelli L (2021) Hippuric acid: Could became a barometer for frailty and geriatric syndromes? Ageing Res Rev 72:101466\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen J-R et al (2021) GPR109A mediates the effects of hippuric acid on regulating osteoclastogenesis and bone resorption in mice. Commun Biology 4:53\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang L, Asimakopoulos AG, Kannan K (2015) Accumulation of 19 environmental phenolic and xenobiotic heterocyclic aromatic compounds in human adipose tissue. Environ Int 78:45\u0026ndash;50\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanahi N et al (2022) Association of amino acid metabolites with osteoporosis, a metabolomic approach: Bushehr elderly health program. Metabolomics 18:63\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEldjarn GH et al (2023) Large-scale plasma proteomics comparisons through genetics and disease associations. Nature 622:348\u0026ndash;358\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarming L, Hassager C, Christiansen C (2002) Changes in Bone Mineral Density with Age in Men and Women: A Longitudinal Study. Osteoporos Int 13:105\u0026ndash;112\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWillers C et al (2022) Osteoporosis in Europe: a compendium of country-specific reports. Archives Osteoporos 17:23\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD\u0026rsquo;Agostino RB et al (2008) General Cardiovascular Risk Profile for Use in Primary Care. Circulation 117:743\u0026ndash;753\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKai MC, Anderson M, Lau EM (2003) Exercise interventions: defusing the world's osteoporosis time bomb. Bull World Health Organ 81:827\u0026ndash;830\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhadi S et al (2020) Personal aging markers and ageotypes revealed by deep longitudinal profiling. Nat Med 26:83\u0026ndash;90\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eM\u0026auml;\u0026auml;tt\u0026auml; LL et al (2024) Longitudinal Change in Serum Neurofilament Light Chain in Type 2 Diabetes and Early Diabetic Polyneuropathy: ADDITION-Denmark. Diabetes Care 47:986\u0026ndash;994\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLibiger O et al (2021) Longitudinal CSF proteomics identifies NPTX2 as a prognostic biomarker of Alzheimer's disease. Alzheimer's Dement 17:1976\u0026ndash;1987\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDodig-Crnković T et al (2020) Facets of individual-specific health signatures determined from longitudinal plasma proteome profiling. eBioMedicine 57\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak GXCMS (2006) Processing Mass Spectrometry Data for Metabolite Profiling Using Nonlinear Peak Alignment, Matching, and Identification. Anal Chem 78:779\u0026ndash;787\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen X et al (2022) TidyMass an object-oriented reproducible analysis framework for LC\u0026ndash;MS data. Nat Commun 13:4365\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSumner LW et al (2007) Proposed minimum reporting standards for chemical analysis. Metabolomics 3:211\u0026ndash;221\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLassen J, Nielsen KL, Johannsen M, Villesen P (2021) Assessment of XCMS Optimization Methods with Machine-Learning Performance. Anal Chem 93:13459\u0026ndash;13466\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitchie SC et al (2023) Quality control and removal of technical variation of NMR metabolic biomarker data in ~\u0026thinsp;120,000 UK Biobank participants. Sci Data 10:64\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-4642034/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4642034/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBiomarker discovery for degenerative diseases is challenging due to low statistical power, selection bias, and biological variability. To address these problems, we introduced pre-diagnostic longitudinal sampling using samples from the Danish Blood Donor Study. We obtained up to six longitudinal metabolomics profiles using one-year intervals with the latest profile within one year before osteoporosis diagnosis, including 99 cases and 99 controls. We matched the patients with controls based on sex, age, sampling site, disease history, body mass index, analytical batch, and sample storage time. Our longitudinal model of molecular changes improved the signal from non-significant in single-sample modeling between patient cases and controls to an area under the curve (AUC) of 0.75. This pilot study demonstrates the advantages of longitudinal data in biomarker research, including robustness to day-to-day biological variance, inter-individual variance, and post-diagnostic biases.\u003c/p\u003e","manuscriptTitle":"Longitudinal pre-diagnostic samples allow early osteoporosis diagnosis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-29 18:11:15","doi":"10.21203/rs.3.rs-4642034/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3b092b61-09d9-4b45-ab04-e286752c6de5","owner":[],"postedDate":"July 29th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":33865451,"name":"Biological sciences/Systems biology/Time series"},{"id":33865452,"name":"Health sciences/Biomarkers/Diagnostic markers"},{"id":33865453,"name":"Health sciences/Health care/Quality of life"},{"id":33865454,"name":"Health sciences/Diseases"},{"id":33865455,"name":"Biological sciences/Biochemistry/Metabolomics"}],"tags":[],"updatedAt":"2026-01-30T11:51:12+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-29 18:11:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4642034","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4642034","identity":"rs-4642034","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}