Identification of potential NAD-related biomarkers of recurrent miscarriage risk

Abstract

STUDY QUESTION

Do women with a history of recurrent miscarriage have altered nicotinamide adenine dinucleotide (NAD)-related metabolites that can be detected in blood, plasma, or urine samples?

SUMMARY ANSWER

Women with a history of recurrent miscarriage have higher blood, plasma, and urine concentrations of NAD Salvage Pathway excretion products, and urinary excretion of nicotinamide (NAM) is also elevated, compared to control women.

WHAT IS KNOWN ALREADY

Recurrent miscarriage risk is associated with advancing age, high and low BMI, dietary factors, various medical conditions including inflammation, as well as environmental exposures, e.g. to chemicals and pollution. Perturbation of NAD synthesis due to genetic and/or environmental factors causes NAD deficiency, which is implicated in Congenital NAD Deficiency Disorder (CNDD) characterized by recurrent pregnancy loss and congenital anomalies. In CNDD mouse models, foetal anomalies and embryo loss are prevented if NAD levels are raised by supplementing the mother’s diet with an NAD precursor, such as vitamin B3, during pregnancy.

STUDY DESIGN, SIZE, DURATION

This prospective pilot cohort study included 88 non-pregnant women between 20 and 40 years of age, 37 with and 51 without a history of recurrent miscarriage. Recurrent miscarriage was defined as a history of two or more consecutive spontaneous miscarriages <20 weeks’ gestation, with the last miscarriage between 6 weeks and 2 years prior to recruitment. The study was conducted at the Royal Hospital for Women, Sydney, Australia, between March 2022 and December 2023.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Participants completed a questionnaire about their socio-demographic characteristics, health, lifestyle, diet, medication, and vitamin use; and provided morning fasting blood and urine samples. Levels of NAD and 25 related metabolites were measured in whole blood, plasma, and urine using a validated ultra-high performance liquid chromatography–tandem mass spectrometry method. Differences in NAD metabolism between the groups were assessed by volcano plots and partial least-squares discriminant analysis. Characteristics of women between the two groups were compared using chi-squared statistics. Multivariable generalized additive models were used to assess the association between NAD metabolites and miscarriage. Predictive accuracy of metabolites alone and with age was examined using three machine learning models, including Logistic Regression, Random Forest, and Gradient Boosting Classifier and assessed using the area under the receiver operating characteristic curve (AUROC).

MAIN RESULTS AND THE ROLE OF CHANCE

Elevated levels of the metabolites 1-methylnicotinamide (1MNA), N-methyl-2-pyridone-5-carboxamide (2PY), and N-methyl-4-pyridone-3-carboxamide (4PY), representing excretion metabolites of the NAD Salvage Pathway, were associated with a higher risk of recurrent miscarriage. These metabolites showed a strong positive correlation among the three tested biological matrices, confirming the suitability of all three matrices to quantify these markers. Whole blood anthranilic acid and urine NAM levels were also elevated in women with recurrent miscarriage. 1MNA in plasma was associated with recurrent miscarriage on univariate analysis, with the effect sustained after taking into account maternal age (adjusted odds ratio 1.02; 95% CI 1.01, 1.03) with every one-unit increase in 1MNA, the odds of miscarriage increasing by 2%. The predictive accuracy using machine learning approaches was highest when all three metabolites 1MNA, 2PY, and 4PY, and age were included in the model (AUROC 0.89; 95% CI 0.83, 0.95).

LIMITATIONS, REASONS FOR CAUTION

This study was conducted in a single hospital, which assures high internal validity but may limit external validity. Sample size was small, and findings should be replicated in larger studies. The measured levels of NAD-related metabolites represent a snapshot at the time of sample collection but might not be reflective of when the women had been pregnant.

WIDER IMPLICATIONS OF THE FINDINGS

Our novel finding of elevated excretion of NAM and its derived products in women experiencing recurrent miscarriage suggests differences in the NAD synthesis pathway are linked to adverse pregnancy outcomes. The significant differences in 2PY and 4PY levels in the circulation and urine between the study groups indicate that these metabolites are potential biomarkers associated with recurrent miscarriage. Several conditions which are associated with adverse pregnancy outcomes also affect NAD metabolism and cause elevated levels of these metabolites. Thus, these metabolites may not simply be biomarkers but also be an indicator of the underlying mechanisms in many cases of recurrent miscarriage. Further evaluation of NAD metabolism in women with recurrent miscarriage in other populations and of other associations including diet is required.

STUDY FUNDING/COMPETING INTEREST(S)

This research was supported by funds to S.L.D. from: the National Health and Medical Research Council (NHMRC), Principal Research Fellowship (ID1135886), Leadership Level 3 Fellowship (ID2007896), and Project Grant (ID1162878); a New South Wales (NSW) Health Cardiovascular Research Capacity Program Senior Researcher Grant; philanthropic support from The Key Foundation, The Ross Trust, and Steven and Linda Harker. N.N. was supported by NHMRC Leadership Level 1 Fellowship (ID1197940) and Financial Markets Foundation for Children. We gratefully acknowledge the Victor Chang Cardiac Research Institute Innovation Centre, funded by the NSW Government, as well as funding from the Freedman Foundation for the Metabolomics Facility. The authors declare no conflicts of interest.

TRIAL REGISTRATION NUMBER

N/A.

Introduction

Miscarriage affects around 15% of recognized pregnancies, and has physical, psychological, and economic effects (Quenby et al., 2021). Recurrent miscarriage is defined as the spontaneous loss of two or more pregnancies before viability (Practice Committee of the American Society for Reproductive Medicine; ESHRE Guideline Group on RPL et al., 2023) that may or may not be consecutive (Quenby et al., 2021). Recurrent miscarriage affects 1–2% of couples trying to conceive (Quenby et al., 2021). Causes are varied and complex, and a significant number remain unexplained despite full investigation (Regan et al., 2023). Recurrent miscarriage risk is associated with advancing maternal and paternal age, high and low BMI, excess alcohol and caffeine consumption, various medical conditions such as poorly controlled diabetes, antiphospholipid antibodies, thyroid disease, inflammation, as well as environmental exposure from smoking, air pollution, pesticides, and dietary intake (Alijotas-Reig and Garrido-Gimenez, 2013; Dimitriadis et al., 2020; Quenby et al., 2021; Regan et al., 2023). Women with a history of miscarriage, and particularly recurrent miscarriage, have an increased risk of future pregnancy complications, including preterm birth, low birth weight, placenta praevia, placental abruption, and stillbirth (Quenby et al., 2021). Women are also at-risk of longer-term health issues, such as mental health conditions, cardiovascular disease, and venous thromboembolism (Sugiura-Ogasawara et al., 2019; Quenby et al., 2021; Linehan et al., 2022). Micronutrient deficiencies are common in reproductive-aged women, especially in low and middle-income countries, due to limited intake of fruit, vegetables, and animal products, and evidence suggests nutrient deficiencies may influence risk of miscarriage (Keats et al., 2019). Specifically, vitamin B3 has recently been implicated (Shi et al., 2017).

Vitamin B3 is the main circulatory precursor of nicotinamide adenine dinucleotide (NAD; collectively referring to its oxidized and reduced form NAD⁺ and NADH, respectively) (Fig. 1). NAD is a coenzyme involved in numerous metabolic processes and is essential for mitochondrial function, cellular energy, redox balance, and DNA repair, all of which are also vital during pregnancy (Dunwoodie et al., 2023). Studies in humans and mice show that reduced NAD levels due to impairment of NAD synthesis during embryo development cause structural congenital anomalies in the foetus and miscarriage in these families, a condition termed Congenital NAD Deficiency Disorder (CNDD) (Shi et al., 2017; Dunwoodie et al., 2023). Importantly, in mice, CNDD-like structural foetal anomalies and embryo loss are prevented when NAD levels are increased with dietary vitamin B3 supplementation of the mother during pregnancy, implicating NAD and vitamin B3 in adverse pregnancy outcomes.

Figure 1.

Overview of the study and NAD synthesis pathways. NAD, representing the redox couple NAD⁺ and NADH, is central to cellular function as it is required for energy metabolism, redox homeostasis, and as a cofactor for NAD-consuming enzymes such as sirtuins, poly(ADP-ribose) polymerases, and CD38 (purple). These enzymes convert NAD to nicotinamide and are essential for cellular processes including gene regulation, cell division and DNA damage repair. NAD is replenished from nicotinamide via the NAD Salvage Pathway (orange). Nicotinamide as well as nicotinic acid, nicotinamide riboside, and nicotinamide mononucleotide (green) are NAD precursors taken up by the diet, with the former three being collectively termed vitamin B3. Nicotinic acid is the precursor for the Preiss-Handler Pathway (grey). Apart from vitamin B3, NAD is also synthesized from the essential amino acid tryptophan via the NAD de novo Synthesis Pathway (blue). When NAD levels are in excess, nicotinamide is either recycled back to NAD or converted to the excretion products 1-methylnicotinamide (1MNA), N-methyl-2-pyridone-5-carboxamide (2PY), and N-methyl-4-pyridone-3-carboxamide (4PY) and excreted in the urine. In addition, nicotinamide is converted to nicotinamide N-oxide by cytochrome P450 under conditions of excess NAD precursor availability (not shown). Enzymes are indicated in red. All metabolites shown in the diagram (black and green) were measured in the present study. Additional metabolites measured but not shown in the diagram are NADP⁺, creatinine, and the B6 vitamers 4-pyridoxic acid, pyridoxal, pyridoxamine, pyridoxine, and pyridoxal 5’-phosphate (26 metabolites measured in total). Created in BioRender. Cuny, H. (2025) https://BioRender.com/wsgjfl6.

But apart from the CNDD patients described so far with biallelic loss-of-function variants in genes of the NAD de novo Synthesis Pathway (Fig. 1), there is limited to no data about NAD-related metabolites in women and their association with adverse pregnancy outcomes. Furthermore, it is unknown whether NAD itself or other NAD-related metabolites are associated with risk of adverse pregnancy outcomes.

Therefore, we determined and compared the levels of NAD and related metabolites among women with and without a history of recurrent miscarriage and whether these are associated with recurrent miscarriage risk. Identifying women at-risk of miscarriage may provide opportunities to optimize maternal health and improve future pregnancy outcomes.

Materials and methods

Study population

Women aged between 20 and 40 years with and without a history of recurrent miscarriage were recruited at the Royal Hospital for Women, Randwick, New South Wales, Australia between March 2022 and December 2023 (Fig. 1). The non-pregnant group (controls) included women who had not been pregnant in the 12 months before recruitment and did not have a history of two or more spontaneous consecutive miscarriages <20 weeks’ gestation. The recurrent miscarriage group included women who had experienced two or more consecutive spontaneous miscarriages <20 weeks’ gestation, with the last miscarriage occurring within the last 2 years. Sample collection was conducted ≥6 weeks after the last miscarriage. Women with a known chromosomal cause for miscarriage were excluded and terminations of pregnancy or ectopic pregnancies were not included in the recurrent miscarriage count.

Data collection

Questionnaire

After written informed consent and enrolment in the study, participants completed a self-administered questionnaire via a web link to the electronic data capture tool Research Electronic Data Capture (REDCap), with the collected data hosted on secure servers at University of Sydney. REDCap is a secure web-based application for building and managing online surveys and databases (Harris et al., 2009). The questionnaire involved a survey of a woman’s socio-demographic characteristics, weight, height, health, lifestyle, diet, medication, and supplement use. This included questions on age, ethnicity, country of birth, marital/partner status, height and weight, occurrence and type of diabetes (if applicable), smoking and alcohol use (including amount and frequency), and medical history. In addition, data on physical activity, diet, medication, contraception, current supplement use, and family history of congenital anomalies and other medical conditions as well as pregnancy history and prior pregnancy outcome(s) (if applicable, including miscarriages, births, or terminations), were also collected. Participants’ cumulative daily vitamin B3 supplementation dosages from multivitamins, B vitamin complex, pure vitamin B3, and/or other vitamin B3-containing products were calculated, and levels were categorized as above or below the daily recommended intake for pregnancy of 18 mg (National Health and Medical Research Council; Australian Government Department of Health and Ageing; New Zealand Ministry of Health, 2006). Women with dietary restriction were defined as those who self-reported as being vegetarian (not eating red or white meat or fish), vegan, or having a diet with no gluten, no milk, no white or red meat, no fish, no eggs, no nuts, no wheat, no soy, no fruits, or no lactose. Women with potential inflammatory conditions were defined as those with autoimmune or mental health conditions, asthma, smoking, hypertension, obesity, or diabetes (Nielsen et al., 2024).

Quantification of NAD⁺ and related metabolites in plasma, whole blood, and urine samples

A blood and a spot urine sample were collected in the morning from each participant after overnight fasting. Participants in the non-pregnant control group provided an additional 24-hour urine sample, as soon as feasible, in the days after the fasting blood and urine test. The blood was collected into a heparin-coated tube (BD Vacutainer^®, Becton Dickinson, NJ, USA), and half of this sample was aliquoted and snap frozen on dry ice (= whole blood for analysis). The other half was centrifuged for 15 min at 2000 g and 4 °C, and the supernatant aliquoted and snap frozen (= plasma for analysis). Urine (20 ml of spot urine, and 24-hour urine, where applicable) was centrifuged for 10 min at 2000 g, filtered with a 0.2 μM pore size syringe filter, aliquoted, and snap frozen (= urine for analysis). All sample aliquots were stored at −70 to −80°C until analysis.

Metabolites were extracted and quantified using ultra-high performance liquid chromatography–tandem mass spectrometry (UHPLC-MS/MS) as previously described (Cuny et al., 2021; Szot et al., 2024). The method included 21 metabolites of the NAD synthesis pathways (Fig. 1), four vitamin B6 vitamers, and creatinine. Creatinine was only used for normalization of urine samples but not included in analyses comparing metabolic profiles between groups. If the measured concentrations of a metabolite were below the limit of detection in >15% of participants’ samples, this metabolite was entirely excluded from analysis in the respective biological matrix. Similarly, metabolites that did not meet the assay quality control parameters were excluded. Any values below the limit of detection among the remaining metabolites were set to half the limit of detection value to allow statistical analyses.

Three biological matrices (plasma, whole blood, and urine) were included to compare their utility for NAD-related biomarker discovery because each has its inherent advantages. Plasma metabolites are reflective of circulatory levels but NAD⁺ and several related metabolites are in the low nanomolar range. Therefore, haemolysis due to rupture of blood cells during blood sampling, will significantly affect measured concentrations in plasma. Whole blood concentrations of NAD⁺ and some related metabolites are orders of magnitude higher than in plasma and therefore easier to quantify, but delays between drawing and freezing blood may affect measurements due to continuing metabolic activity in blood cells. Urine samples are easiest to collect and process but require normalization to 24-hour urine volume, osmolality, creatinine, or other parameters to account for differences in hydration status. Here, urinary metabolite levels were normalized to creatinine concentration, measured by UHPLC-MS/MS together with the NAD-related metabolites. Spot and 24-hour urine samples of the non-pregnant control group were compared to each other to determine suitability of spot urine sampling. All other statistical analyses were conducted with spot urine data only.

Statistical analyses

Descriptive statistics and contingency tests were calculated to examine questionnaire data and compare the frequency and percentage of socio-demographic characteristics, health and medical conditions, lifestyle, diet and supplement use between the two groups. Median and interquartile ranges (IQR) of quantified levels of NAD⁺ and related metabolites measured in each media of blood, plasma, and urine samples were compared between groups using Mann–Whitney test and box plots. Metabolite concentrations were then log-transformed and Pareto scaled, and partial least squares-discriminant analysis (PLS-DA) applied to compare overall metabolite profiles and differentiate between the two study groups. The most discriminating metabolites were presented in descending order of their importance in projection plots.

Volcano plots were then used to define NAD metabolites of interest in each biological matrix and those meeting the P-value and fold-change cutoffs were defined as significant. Correlation between blood and urine levels was examined with scatter plot visualization and Spearman’s rank correlation, while median level values in spot versus 24-hour urine were compared using Mann–Whitney test. PLS-DA and volcano plots, as well as Pearson’s correlation matrix to compare metabolite levels across all three matrices were generated using MetaboAnalyst 6.0 (Pang et al., 2024). Scatter plots and statistical analyses were performed using Prism (version 10; GraphPad Software) with P-values <0.05 considered to be statistically significant.

The association between levels of significant NAD metabolites of interest in each of the matrices and recurrent miscarriage group was assessed using univariable and multivariable generalized additive models to allow for nonlinear association and adjust for potential confounding by maternal age and/or vitamin B3 intake. Odds ratios (OR) and 95% CI were estimated using the data analysis software SAS (version 9.4; Sydney, Australia) and statistical significance confirmed where OR CI did not include one.

Further analyses to examine the accuracy of metabolites alone and with age in predicting miscarriage was applied using three machine learning models, including Logistic Regression, Random Forest, and Gradient Boosting Classifier. Metabolite levels were scaled to normalize their distributions, ensuring compatibility with algorithms. Hyperparameter optimization was conducted for all models using grid search with 3-fold cross-validation to identify the best model configurations. A 5-fold stratified cross-validation scheme was then used to evaluate model performance while preserving class distribution in each fold. Predictive accuracy was assessed using the area under the receiver operating characteristic curve (AUROC) as well as other measures of Classification Accuracy (CA), F1 Score, Precision and Recall, and Matthews Correlation Coefficient. These metrics were computed for each fold and averaged across folds and reported as mean values with 95% confidence intervals. Values for the AUROC were interpreted as 90–100=excellent; 80–90=good; 70–80=fair; 60–70=poor; 50–60=fail. Feature importance was assessed and plotted for each model, derived from the Gini importance (Random Forest and Gradient Boosting), and using the absolute magnitude of model coefficients (Logistic Regression). Sensitivity analyses to assess robustness of findings were conducted by restricting to women with vitamin B3 supplement intake on top of diet below recommended daily intake of 18 mg, and age 32–40 years and differences between groups assessed using Mann–Whitney test. Finally, additional analyses were conducted among women with recurrent miscarriage to assess the association between number and time since previous miscarriage on significant NAD metabolite levels as measured in plasma.

Ethical approval

Ethics approval was obtained from the South Eastern Sydney Local Health District Human Research Ethics Committee (Approval number 2021_ETH11117).

Results

Study group characteristics

Table 1 presents the socio-demographic, diet, and supplement use characteristics of the two study groups consisting of 51 non-pregnant control women and 37 women with recurrent miscarriages. Samples from the recurrent miscarriage group were collected between 6 and 104 weeks after the end of the last pregnancy and miscarriage occurrence. Women in the recurrent miscarriage group had a higher median age (36 vs 29 years), were more likely to be taking any vitamin supplements (78% vs 35%) and were less likely to be born in Australia (41% vs 73%). No women had diabetes mellitus and there was no difference between the two groups in exposure to potential inflammatory conditions (Table 1). More women in the miscarriage group were taking supplements containing vitamin B3 (40% vs 6%) and at higher doses (median 18 mg; range 12–100) compared to the non-pregnant group (9 mg; range 7.5–20).

Table 1.

Comparison of participant characteristics between women with and without recurrent miscarriage.

Participant characteristics	Non-pregnant control (n = 51)		Recurrent miscarriage (n = 37)
	N	%	N	%	P-value
Age (years), median and rangea	29	21.2–40.5	36	23.3–40.9	<0.0001
BMI (kg/m2), median and range	23	17–41.5	23	19.1–44.6	0.47
Highest education level achieved					0.69
Less than or up to Year 12 or equivalent	3	5.9	2	5.4
Trade / Certificate / Diploma	4	7.8	5	13.5
University	44	86.3	30	81.1
Smoking					0.92
No	48	94.1	35	94.6
Yes	3	5.9	2	5.4
Alcohol drinking					0.06
No	12	23.5	16	43.2
Yes	39	76.5	21	56.8
Ethnicity					0.47
Caucasian / White	42	82.4	28	75.7
Other	6	11.8	4	10.8
Asian / South East Asian	3	5.9	5	13.5
Country of birth					0.01
Australia	37	72.5	15	40.5
UK, Ireland, EU, USA, Canadab	8	15.7	12	32.4
Other	6	11.8	10	27.0
Any potential inflammatory conditionc					0.65
No	36	70.6	24	64.9
Yes	15	29.4	13	35.1
Immediate family with congenital condition					0.72
Yes	4	7.8	4	10.8
No/unsure	47	92.2	33	89.2
Any dietary restrictiond					0.25
No	40	78.4	33	89.2
Yes	11	21.6	4	10.8
Vegetarian or vegan	2	4.0	0	0.0
Red meat consumption					0.07
No	8	15.7	1	2.7
Some	43	84.3	36	97.3
Supplement combination					<0.0001
None	33	64.7	8	21.6
Multivitamin only	3	5.9	8	21.6
Multivitamin plus other vitamins and/or supplements	1	2.0	15	40.5
Vitamins and/or supplements only, no multivitamin	14	27.5	6	16.2
Taking vitamin B3 containing supplementse					<0.0001
No	48	94.1	22	59.5
Yes	3	5.9	15	40.5
Vitamin B3 (mg), median and rangee	8.5	7.5–20	18	12–100
Vigorous exercise per week					0.01
None	6	11.8	10	27.0
≤1 h/week	14	27.5	16	43.2
≥1 h/week	31	60.8	11	29.7
Trying to conceive
No or not known	49	96.1	6	16.2
Yes	2	3.9	31	83.8
Gravidity
None	44	86.3	–	–
One	3	5.9	–	–
Two	3	5.9	13	35.1
Three or more	1	2.0	24	64.9
Previous birth ≥20 weeks gestation
Any livebirth or stillbirth	7	13.7	17	45.9

^aAt time of sample collection.

^bUK, United Kingdom; EU, European Union; USA, United States of America.

^cMedical conditions: autoimmune or mental health conditions, asthma, smoking, hypertension, obesity, or diabetes.

^dIncludes no gluten, milk, white or red meat, fish, eggs, nuts, wheat, soy, fruits, or lactose.

^eVitamin B3 from self-reported supplements may be nicotinic acid, nicotinamide, nicotinamide mononucleotide, or nicotinamide riboside.

Concentrations of NAD⁺ and related metabolites in whole blood, plasma, and urine

We determined the levels of NAD⁺ and related metabolites in three biological matrices. Plasma represents the circulation which distributes NAD precursors to cells throughout the body (Liu et al., 2018) and is ideally suited to identify perturbations of NAD metabolism due to health conditions (Matsuoka et al., 2017; Wang et al., 2022; Li et al., 2025). Whole blood NAD metabolite levels are a good indicator for tissue NAD metabolite levels and NAD precursor availability (Conze et al., 2019; Elhassan et al., 2019; Pirinen et al., 2020). Urine levels show excreted NAD-related metabolites, which are another good marker of whole-body NAD precursor availability (Shibata and Matsuo, 1989; Trammell et al., 2016; Kremer et al., 2018). Comparison of metabolite levels between the two groups revealed significantly different levels of nine metabolites in whole blood (Supplementary Table S1 and Fig. S1), six metabolites in plasma (Table 2, Fig. 2), and five in urine (Supplementary Table S2 and Fig. S2). Of the NAD-related metabolite differences, 1-methylnicotinamide (1MNA), N-methyl-2-pyridone-5-carboxamide (2PY), and N-methyl-4-pyridone-3-carboxamide (4PY) levels in all three matrices, as well as anthranilic acid (AA) in whole blood and nicotinamide (NAM) in urine were the most elevated in the recurrent miscarriage group using discriminant analyses and confirmed using volcano plots (Fig. 3). While the overall concentration of 1MNA, 2PY, and 4PY was higher in women with recurrent miscarriage, the relative proportion of these metabolites to each other between groups were not different, indicating that this collective increase in NAD-related excretion metabolites arose from elevated NAM levels (Supplementary Fig. S3, see Fig. 1 for metabolic pathways). Correlation of metabolite concentrations from the entire study cohort across all three biological matrices showed the strongest positive correlation between the concentrations of 1MNA, 2PY, and 4PY, and these also correlated with levels of NAM in urine. This indicates that all three biological matrices are suitable to identify changes in these four metabolites. In addition, levels of vitamin B6, specifically the vitamers 4-pyridoxic acid (4PA) in all matrices and pyridoxal (PL) in whole blood, also correlated to each other (Supplementary Fig. S4).

Table 2.

Plasma metabolite concentrations.

	Non-pregnant control (n = 51)	Recurrent miscarriage (n = 37)
NAD-related metabolites	Median (interquartile range)	Median (interquartile range)	P-valuea
TRP (µM)	55.7 [48.2–62.5)	47.8 [41.9–52.9]	<0.001*
KYN (µM)	1.76 [1.55–2.09]	1.70 [1.32–2.37]	0.73
3HK (nM)	32.3 [26.6–38.0]	31.6 [26.2–41.2]	0.93
3HAA (nM)	53.2 [41.0–67.9]	45.7 [34.6–62.9]	0.16
KA (nM)	56.2 [43.9–71.2]	66.4 (44.1–75.4]	0.27
AA (nM)	14.4 [10.9–21.3]	12.9 [6.25–17.2]	0.10
XA (nM)	24.3 [21.2–29.6]	19.0 [13.0–26.2]	0.002*
NAD+ (nM)	5.72 [4.34–7.96]	7.04 [4.38–13.6]	0.08
NAM (nM)	93.5 [67.6–130]	120 [103–167]	0.01*
NMN (nM)	2.32 [1.65–3.33]	1.56 [1.54–3.35]	0.09
1MNA (nM)	89.2 [59.3–134]	151 [106–215]	<0.001*
2PY (µM)	1.05 [0.769–1.40]	1.69 [1.13–2.79]	<0.001*
4PY (nM)	206 [150–316]	325 [213–649]	0.001*
1MNA, 2PY, 4PY combined (µM)	1.36 [1.02–1.77]	2.22 [1.54–3.63]	<0.001*
Other metabolites
4PA (nM)	21.4 [16.0–46.1]	27.4 [20.2–70.7]	0.11

^a P values were determined by Mann–Whitney test and those <0.05 highlighted with an asterisk (*).

Metabolites that were measured but did not meet quality control parameters (nicotinic acid adenine dinucleotide, NADH, NADP⁺) and those with >15% of values across the study cohort being below the detection limit, which were quinolinic acid (QA), nicotinic acid (NA), nicotinic acid mononucleotide (NAMN), nicotinic acid riboside (NAR), nicotinamide riboside (NR), pyridoxal (PL), pyridoxal 5’-phosphate (PLP), pyridoxamine (PM), and pyridoxine (PN) were excluded. The limits of detection were 12.5 nM for QA, 3.13 nM for NA, 0.313 nM for NAMN, 0.313 nM for NAR, 0.0125 nM for NR, 3.13 nM for PL, 3.13 nM for PLP, 3.13 nM for PM, and 0.313 nM for PN.

TRP, tryptophan; KYN, kynurenine; 3HK, 3-hydroxykynurenine; 3HAA, 3-hydroxyanthranilic acid; KA, kynurenic acid; AA, anthranilic acid; XA, xanthurenic acid; NAM, nicotinamide; NMN, nicotinamide mononucleotide, 1MNA, 1-methylnicotinamide; 2PY, N-methyl-2-pyridone-5-carboxamide; 4PY, N-methyl-4-pyridone-3-carboxamide; 4PA, 4-pyridoxic acid.

Figure 2.

Concentrations of NAD-related metabolites in plasma samples. Horizontal line indicates the median and bars indicate the interquartile range. N, non-pregnant controls (n = 51); M, women with recurrent miscarriage (n = 37). ****P < 0.0001; ***P < 0.001; **P < 0.01; ns, not significant (Mann–Whitney test). TRP, tryptophan; KYN, kynurenine; 3HK, 3-hydroxykynurenine; 3HAA, 3-hydroxyanthranilic acid; KA, kynurenic acid; AA, anthranilic acid; XA, xanthurenic acid; NAM, nicotinamide; NMN, nicotinamide mononucleotide, 1MNA, 1-methylnicotinamide; 2PY, N-methyl-2-pyridone-5-carboxamide; 4PY, N-methyl-4-pyridone-3-carboxamide; 4PA, 4-pyridoxic acid.

Figure 3.

Statistical assessment of plasma, blood, and urine NAD metabolite profiles in women with and without history of recurrent miscarriages. (A, D, G) Partial least squares-discriminant analysis (PLS-DA) 2-dimensional score plots of 14 quantifiable metabolites in plasma (A), 14 in blood (D), and 11 in spot urine (G) of 51 non-pregnant control women and 37 women with recurrent miscarriage. Each dot represents a participant. (B, E, H) Respective variable importance in projection (VIP) plots. The most discriminating metabolites are shown in descending order of their VIP scores. Red and blue squares indicate whether metabolite levels are higher or lower in the respective study group. (C, F, I) Volcano plots of metabolite levels in plasma (C), blood (F), and urine (I). Each dot represents a metabolite, the x-axis displays the log2-fold change (FC), and the y-axis displays the negative logarithm (-log10) of the P-value. Positive FC values indicate metabolite levels are higher in women with recurrent miscarriage, negative values indicate they are lower. Dashed lines show the applied thresholds for significance (P = 0.05) and for FC (1.5-fold). Large dot sizes and red colour indicate metabolites that are significantly elevated in women with recurrent miscarriage, and their names are shown in the plots. TRP, tryptophan; KYN, kynurenine; 3HK, 3-hydroxykynurenine; 3HAA, 3-hydroxyanthranilic acid; KA, kynurenic acid; AA, anthranilic acid; XA, xanthurenic acid; NAM, nicotinamide; NMN, nicotinamide mononucleotide, 1MNA, 1-methylnicotinamide; 2PY, N-methyl-2-pyridone-5-carboxamide; 4PY, N-methyl-4-pyridone-3-carboxamide; 4PA, 4-pyridoxic acid; PL, pyridoxal.

All NAD Salvage Pathway metabolites (NAD⁺, NAM, NMN) were significantly elevated in blood of the recurrent miscarriage group and this was also seen for NAM in plasma, and NAM and NMN in urine, when analysed in pairwise comparisons (Fig. 2, Supplementary Figs S1 and S2). This suggests that the recurrent miscarriage group had an overall higher NAD precursor availability (Elhassan et al., 2019) and Salvage Pathway activity, which was further investigated by considering dietary supplementation with vitamin B3 (see below). Tryptophan (TRP) levels were significantly lower in whole blood of women with recurrent miscarriage but not significantly different between the groups in plasma. Conversely, levels of AA were only different in whole blood (Fig. 2, Supplementary Fig. S1), highlighting that not all metabolic trends are consistent between matrices and thus measuring metabolites in multiple matrices provides a more comprehensive overview.

Spot urine was compared to 24-hour urine, showing that most metabolites had comparable levels between the sample types when normalized to creatinine concentration. However, median levels of kynurenine (KYN), 3-hydroxyanthranilic acid (3HAA), xanthurenic acid (XA), and quinolinic acid (QA), representing metabolites of the NAD de novo Synthesis Pathway, were significantly higher in the 24-hour urine than in the spot urine (Supplementary Table S3).

Association and predictive accuracy of NAD-related metabolites for recurrent miscarriage

Of the significant NAD metabolites, 1MNA, 2PY and 4PY, only 1MNA in plasma was associated with recurrent miscarriage on univariate analysis, with the effect sustained after taking into account maternal age and vitamin B3 supplement intake (Table 3). For every one-unit increase in 1MNA, the odds of miscarriage increased by 2% (adjusted OR 1.02; 95% CI 1.01, 1.03). It should be noted that maternal age had an independent and strong effect on miscarriage, increasing odds by 34% (OR 1.34; 95% CI 1.19, 1.52). When assessing the predictive accuracy of 1MNA, 2PY, and 4PY using machine learning approaches, the inclusion of all three together in the model revealed the highest predictive accuracy in the gradient boosting model (AUC 0.81; 95% CI 0.78, 0.84), and when age was included in the models this increased accuracy by a further 10% to 0.889 (95% CI 0.83, 0.95) sensitivity in the Logistic Regression model (Table 3). Feature importance of 2PY and 4PY in predicting miscarriage was higher than age in Logistic Regression but importance of age was superior in Random Forest and Gradient Boosting algorithms (Supplementary Fig. S5).

Table 3.

Association and predictive accuracy of plasma concentrations of significant NAD-related metabolites and recurrent miscarriage.

Factors of interest	Univariate OR (95% CI)	Model 1—adj age OR (95% CI)	Model 2—adj B3 Supp OR (95% CI)	Model 3—adj Age+B3 Supp OR (95% CI)	Model 4—adj 1MNA, 2PY, 4PY OR (95% CI)	Model 5—adj 1MNA, 2PY, 4PY, age OR (95% CI)
NAD metabolites
1MNA	1.020 (1.010, 1.030)	1.020 (1.010, 1.030)	1.020 (1.010, 1.030)	1.018 (1.005, 1.03)	1.012 (1.001, 1.024)	1.014 (0.999, 1.028)
2PY	1.001 (1.001, 1.002)	1.001 (1.000, 1.002)	1.001 (1.000, 1.002)	1.001 (1.000, 1.002)	1.005 (1.001, 1.009)	1.005 (1.001, 1.009)
4PY	1.004 (1.002, 1.007)	1.003 (1.000, 1.006)	1.004 (1.001, 1.007)	1.003 (1.000, 1.006)	0.985 (0.972, 0.997)	0.983 (0.969, 0.998)
Participant factors	Univariate  OR (95% CI)	Model 1—adj 1MNA  OR (95% CI)	Model 2—adj 2PY  OR (95% CI)	Model 3—adj 4PY  OR (95% CI)
Age	1.344 (1.186, 1.524)	1.319 (1.138, 1.530)	1.293 (1.124, 1.487)	1.294 (1.128, 1.484)
B3 Suppa	10.908 (2.861, 41.589)	2.923 (0.636, 13.426)	3.068 (0.667, 14.126)	3.549 (0.794, 15.864)
Machine Learning results	Logistic Regression—  AUROC (95% CI)	Random Forest—  AUROC (95% CI)	Gradient Boosting—  AUROC (95% CI)
1MNA	0.730 (0.675, 0.786)	0.655 (0.597, 0.713)	0.676 (0.631, 0.721)
2PY	0.735 (0.622, 0.848)	0.660 (0.489, 0.831)	0.656 (0.487, 0.825)
4PY	0.714 (0.597, 0.831)	0.703 (0.598, 0.809)	0.681 (0.581, 0.780)
1MNA, 2PY, 4PY	0.781 (0.669, 0.893)	0.807 (0.755, 0.859)	0.811 (0.783, 0.839)
1MNA, 2PY, 4PY, age	0.889 (0.828, 0.950)	0.844 (0.775, 0.913)	0.836 (0.751, 0.921)

^aB3 Supp refers to women who supplemented their diet with 7.5–100 mg/day of vitamin B3 on top of the B3 intake from the diet.

OR, odds ratio; AUROC, area under the receiver operating characteristic curve; adj, adjusted for; 1MNA, 1-methylnicotinamide; 2PY, N-methyl-2-pyridone-5-carboxamide; 4PY, N-methyl-4-pyridone-3-carboxamide.

Effect of vitamin B3 supplementation and demographic characteristics on NAD-related metabolites

In the sensitivity analysis of participants who did not supplement with vitamin B3 (n = 48 control group, n = 22 recurrent miscarriage group), the trend in median IQR concentrations and pairwise comparisons (Supplementary Tables S4, S5 and S6 and Supplementary Figs S6, S7, S8 and S9) were similar to those seen for the entire cohort. Specifically, women with recurrent miscarriages who did not take vitamin B3 supplements had significantly elevated levels of 1MNA, 2PY, and 4PY in all three matrices, and whole blood AA and urine NAM were elevated (Supplementary Figs S6, S7, S8 and S9). In pairwise comparisons, whole blood NAD⁺ levels were still significantly elevated in the recurrent miscarriage subgroup that did not supplement with vitamin B3 (Supplementary Fig. S7). Conversely, TRP levels were lower in plasma, as were plasma and blood XA levels (Supplementary Figs S6 and S7).

Similarly, when restricting analysis to women aged 32–40 years, results were comparable to the whole cohort with levels of 1MNA, 2PY, and 4PY significantly higher in participants with recurrent miscarriage independent of age (Supplementary Tables S7, S8 and S9 and Supplementary Figures S10, S11, S12 and S13). Whole blood NAD⁺ levels were still significantly elevated in the recurrent miscarriage group when restricted to 32–40 years (Supplementary Fig. S11), but plasma TRP levels were no longer different between the groups (Supplementary Fig. S10). Finally, we observed no correlation between the time since the last miscarriage and metabolite levels of 1MNA, 2PY, and 4PY (Supplementary Fig. S14).

Discussion

Principal findings

Our study found that women with previous recurrent miscarriage had higher concentrations of NAD Salvage Pathway excretion products 1MNA, 2PY, and 4PY in whole blood, plasma, and urine than women who had not had recurrent miscarriages. Furthermore, women with recurrent miscarriage excreted more NAM, the precursor of 1MNA, 2PY, and 4PY in urine. The association between the altered NAD-related metabolite levels and recurrent miscarriage was independent of age and whether women were taking vitamin B3-containing supplements or not. The predictive performance of the metabolites was good, with 88.9% accuracy in identifying women with recurrent miscarriage.

Interpretation of findings—potential mechanisms

The metabolites 1MNA, 2PY, and 4PY are generated from NAM (Fig. 1) and all four of these metabolites have been shown to be elevated in circulation and urine when the dietary supply of NAD precursors is in excess (Shibata and Matsuo, 1989; Trammell et al., 2016; Kremer et al., 2018). All four (NAM, 1MNA, 2PY, 4PY) were elevated in the urine of the recurrent miscarriage group. We also observed elevated concentrations of NAD⁺ and the Salvage Pathway metabolites NAM and NMN in the blood of women with recurrent miscarriage, and this was independent of vitamin B3 supplementation. Together, this suggests the recurrent miscarriage study group was not NAD deficient and instead NAM was in excess and excreted. Besides excretion of NAM itself, levels of the NAM-derived metabolite nicotinamide N-oxide could be measured to confirm this, as it is only detectable under excess NAM availability (Stratford and Dennis, 1992; Shibata et al., 2014; Giner et al., 2021).

In addition to elevated NAD status, high levels of NAM, 1MNA, 2PY, and 4PY can be a sign of increased activity of NAD-consuming enzymes which convert NAD to NAM (Fig. 1) (Canto et al., 2015; Katsyuba et al., 2020). Chronic inflammation is associated with increased NAD-consuming enzyme activity and a subsequent decreased NAD content, and both effects have been observed in placentas of women with inflammation-driven preeclampsia (Jahan et al., 2024). Furthermore, changes in NNMT activity, the enzyme converting NAM to 1MNA (Fig. 1), are linked to chronic inflammation, inflammatory bowel disease, chronic kidney disease, and type 2 diabetes (Lenglet et al., 2016; Pissios, 2017; Xue et al., 2023; Pugel et al., 2024), all health conditions associated with elevated miscarriage risk (Khashan et al., 2012; Chinnappa et al., 2013; Yong et al., 2023; Miao and Yang, 2024). A study using different mouse models of pre-eclampsia, found that NAM prolonged pregnancy, reduced miscarriage and prematurity, and corrected foetal growth restriction (Takahashi et al., 2018). Over-activation of NAD-consuming enzymes generally results in NAD⁺ depletion (Chini et al., 2021; Chu and Raju, 2022). As we observed higher, and not depleted, blood NAD levels in the recurrent miscarriage group, over-activation of NAD-consuming enzymes may not be the main mechanism driving recurrent miscarriage risk in these women or it was compensated by higher NAD precursor availability.

Recent findings suggest that elevated levels of 1MNA, 2PY, and 4PY might be exacerbating conditions of increased miscarriage risk and inflammation. Ferrell et al. reported that 2PY and 4PY levels are positively correlated with an increased risk of major adverse cardiovascular events and additionally, 4PY was positively correlated with C-reactive protein, a marker of inflammation (Ferrell et al., 2024). The proposed mechanism involves induction of vascular cell adhesion molecule 1, which promotes vascular inflammation, by 4PY (Ferrell et al., 2024). Also, 2PY was shown to inhibit the activity of poly(ADP-ribose) polymerase-1, an NAD-consuming enzyme involved in DNA damage repair, transcription, replication, and other roles (Lenglet et al., 2016). While we explored self-reported autoimmune and other potential inflammatory conditions at the time of sampling, occurrence of chronic or acute inflammation cannot be ruled out.

Oxidative stress has been implicated in recurrent miscarriage (Gupta et al., 2007; Khadzhieva et al., 2014; Duhig et al., 2016). Therefore, other potential markers of recurrent miscarriage may include over-oxidized NAD and its derivatives, the ribosylated forms of N-methylated pyridones (pyridone ribosides). These occur under conditions of oxidative stress and their circulatory and urinary levels are elevated with the progression of metabolic diseases, age, and cancer (Hayat et al., 2024). Thus, future studies could test if women with recurrent miscarriage have elevated levels of pyridone ribosides.

The enzyme NNMT methylates NAM to generate 1MNA (Fig. 1), and elevated excretion of 1MNA and its derived metabolites 2PY and 4PY indicate that more methyl groups have been consumed irreversibly. These methyl groups are provided by the methionine cycle involving S-adenosyl-methionine and homocysteine (Hong et al., 2018; Roberti et al., 2021). Therefore, women with recurrent miscarriage and elevated levels of NAM-derived excretion products might have reduced availability of methyl groups for other cellular processes. For example, the methionine cycle is critical for epigenetic regulation of gene expression (Hong et al., 2018; Roberti et al., 2021). Therefore, disruption of the methionine cycle and altered gene expression is another possible mechanism of adverse pregnancy outcomes (Dai et al., 2021). Moreover, investigation of phosphoribosyl diphosphate may be warranted given it is a cofactor for the conversion from NAM to NMN in the Salvage Pathway (Takahashi et al., 2010).

Pregnant women excrete elevated levels of 1MNA and its derivatives 2PY and 4PY compared to non-pregnant women (Fukuwatari et al., 2004). Animal model data suggest that this elevation originates from increased release of TRP from protein breakdown and usage in NAD de novo synthesis (Sano et al., 2016). A study of women administered NAM found no difference in plasma NAM in pregnant women compared to non-pregnant women, however, plasma 2PY was lower and 2PY metabolites in urine were higher in pregnant women due to pregnancy-related increase in renal elimination (Fashe et al., 2024). In our study, samples in the recurrent miscarriage group were collected between 6 and 104 weeks after the end of pregnancy, therefore, the excess levels of these metabolites were unlikely to be from ongoing metabolic changes from pregnancy. To confirm, we assessed whether the time since the last miscarriage occurrence affected the 1MNA, 2PY, and 4PY levels but no correlation was evident.

In a study of pregnant women in China, where vitamin B levels were quantified using isotope-dilution mass spectrometry in early pregnancy, vitamin B3 deficiency rates were found to be higher in women with two or more spontaneous miscarriages, than in women with ongoing pregnancies and women with one miscarriage (Wu et al., 2024). They also found that low levels of vitamin B9 (folate), B6 (pyridoxine), and B5 (pantothenic acid) were independent factors for recurrent miscarriage.

The likelihood of experiencing a miscarriage gradually increases with higher maternal age, from 9% in women 20–24 years of age to 51% at 40–44 years, and 75% at 45 years or older (Rai and Regan, 2006). While there are numerous factors contributing to this miscarriage rate, it has previously been shown that NAD levels decline with age (Verdin, 2015; Chu and Raju, 2022). Lower NAD levels are linked to a decrease in oocyte quality (Liang et al., 2023), suggesting that a woman’s NAD status influences fertility, especially in the context of increasing age. However, our findings did not show differences in metabolic results when restricting to older maternal age >32 years.

Findings in relation to other studies

Other studies assessing associations between NAD-related metabolites and adverse pregnancy outcomes (including miscarriage) have mostly focused on the NAD de novo Synthesis Pathway. A systematic review concluded that metabolites of this pathway are altered in a variety of maternal pregnancy and foetal conditions, indicating that pregnancy requires a tight balance of these metabolites (van Zundert et al., 2022). This is supported by the findings that metabolites of this pathway are required in the placenta for immunoregulation and to prevent foetal rejection (Broekhuizen et al., 2021) and that elevated first-trimester KYN levels are associated with impaired development of the utero-placental vasculature (van Zundert et al., 2024).

Studies exploring the association between miscarriage and the NAD Salvage Pathway or the metabolites 1MNA, 2PY, and 4PY, are very limited. A small prospective study of 24 first-trimester pregnant women found that eight women who miscarried had a lower dietary intake of vitamin B3 and lower urinary levels of 2PY and 2PY/1MNA ratio during pregnancy compared to the 16 women with a successful pregnancy (Yakob et al., 2021), but given that 8 of the 16 control women had prior miscarriages, it is unclear whether the metabolic findings are representative for women at risk of miscarriage.

Strengths and limitations

To the best of our knowledge, this is the first prospective cohort study that explored whether levels of NAD-related metabolites are different in women with a history of recurrent miscarriages. We determined levels of multiple NAD-related metabolites simultaneously, including precursors, intermediates, and products, in three biological matrices: plasma, whole blood, and urine. The identified potential biomarkers are correlated across all matrices, indicating they are equally suitable, and that spot urine results were overall comparable to 24-hour urine.

While the effect of vitamin B3 supplementation on metabolite levels was assessed, the exact total dietary vitamin B3 status of the study cohort from food and supplements was unknown, and it cannot be ruled out that differences in lifestyle including the known differences in physical activity and country of origin, dietary habits, and nutritional status between the groups account as known and unknown confounding factors. Exercise, coffee, and foods high in vitamin B3 may all influence NAD metabolism (Kremer et al., 2018; Walzik et al., 2023).

It is not known whether the elevated excretion of NAM, the main circulatory NAD precursor, and of the methylated excretion products is a significant risk factor of miscarriage or whether it is only a common feature of different health conditions that increase the risk of miscarriage via other mechanisms. Additional studies are needed to investigate how recurrent miscarriage is linked to environmental and/or genetic changes of NAD metabolism, while addressing other risk factors for miscarriage such as inflammation, oxidative stress, altered homocysteine levels, dietary intake, age, and other health conditions. Specifically, studies among women with miscarriages, across a wider socioeconomic and demographic spectrum are required to confirm NAD-related biomarkers of elevated risk of miscarriage.

Conclusions and future research

There is an unmet need for biomarkers that allow identification of women at risk of miscarriage to enable early intervention prior to conception and in early pregnancy. While a larger cohort study is required to establish 1MNA, 2PY, and 4PY as biomarkers for recurrent miscarriage risk into clinical practice, this study is a promising first step towards that goal. The causes of recurrent miscarriage are varied and multifactorial. But our findings, in the context of previous studies, suggest that there may be common underlying molecular mechanisms that unify subgroups of recurrent miscarriage cases with respect to their metabolic perturbations. Given the numerous essential roles of NAD for embryonic development, further research into the role of NAD and related metabolites in miscarriage causation will improve our understanding towards preventative intervention strategies.

Data availability

The clinical data underpinning this study are not publicly available due to data confidentiality and privacy policies. Data may be shared following reasonable request to the corresponding author, submission of protocol and after relevant study team, ethics, and governance approvals have been obtained.

Authors’ roles

Conceptualized and designed the study: H.C., A.W.S., N.N., S.L.D. Recruited participants and collected samples: J.G., T.T., O.A. Processed participants’ samples: H.C., D.Z.S., E.M.M.A.M., A.S. Acquired metabolic data: H.C., D.Z.S. Acquired clinical data: A.W.S. Analysed and interpreted data: H.C., A.W.S., F.S., N.N., S.L.D. Supervised the study: N.N., S.L.D. Provided funding for the study: N.N., S.L.D. Drafted the article: H.C., A.W.S., F.S., N.N., S.L.D. All authors revised the article for important intellectual content and approved the final version of the article.

Funding

This research was supported by funds to S.L.D. from: the National Health and Medical Research Council (NHMRC), Principal Research Fellowship (ID1135886), Leadership Level 3 Fellowship (ID2007896), and Project Grant (ID1162878); a New South Wales (NSW) Health Cardiovascular Research Capacity Program Senior Researcher Grant; philanthropic support from the Key Foundation, the Ross Trust, and Steven and Linda Harker. N.N. was supported by NHMRC Leadership Level 1 Fellowship (ID1197940) and Financial Markets Foundation for Children. We gratefully acknowledge the Victor Chang Cardiac Research Institute Innovation Centre, funded by the NSW Government, as well as funding from the Freedman Foundation for the Metabolomics Facility.

Conflict of interest

The authors declare no conflict of interests.