Exercise MR-proANP unmasks latent right heart failure in CTEPH

Background

Chronic right heart failure is the major determinant of outcome in chronic thromboembolic pulmonary hypertension (CTEPH). Thus, its early detection is crucial for optimal patient management. Hemodynamic assessment with invasive right heart catheterization and the measurement of natriuretic peptides at rest are established diagnostic tools in this context. An elevated right atrial pressure (RAP) is a sensitive parameter of right right heart failure. Mid-regional pro-atrial natriuretic peptide (MR-proANP) is a natriuretic peptide that reflects right atrial stress levels.

There is a growing recognition of exercise diagnostics in the assessment of right heart failure, particularly in patients with normal or borderline findings at rest. There are no data on the dynamics of MR-proANP in correlation to RAP during physical exercise.

Purpose

The present study was designed to investigate the dynamics of RAP and MR-proANP during physical exercise in patients with CTEPH and to determine whether these parameters might serve as a tool to measure exercise-dependent atrial stress as an indicator of right heart failure.

Methods

This observational cohort study included 100 CTEPH patients who underwent right heart catheterization during physical exercise (eRHC). Blood samples for MR-proANP measurement were taken prior, during, and after eRHC. MR-proANP levels were correlated to RAP levels at rest, at peak exercise (eRAP), and during recovery. RAP at rest ≤7 mmHg was defined as normal and eRAP >15 mmHg as suggestive of right heart failure.

Results

During eRHC mean RAP increased from 6±4 mmHg to 16±7 mmHg (p<0.001). MR-proANP levels and dynamics correlated with RAP at rest (rs=0.61; p<0.001; figure 1a) and at peak exercise (rs=0.66; p<0.001; Figure 1b). Furthermore, the relative percent increase in MR-proANP correlated with the relative percent increase in eRAP (rs=0.52; p<0.001; Figure 1c) Logistic regression analysis revealed the peak MR-proANP level (B=0.058; p=0.004) and the right atrial area (B=0.389; p<0.001) to be associated with eRAP dynamics. A peak MR-proANP level ≥139 pmol/L (AUC=0.81) and recovery level ≥159 pmol/L (AUC=0.82) predicted an eRAP >15 mmHg. Physical exercise unmasked RH failure in 39% of patients with normal RAP at rest; they were also characterized by a more distinct increase in MR-proANP levels (p=0.005) and higher peak (p<0.001) and recovery levels (p<0.001).

Conclusions

RAP and MR-proANP dynamics unmask manifest and latent right heart failure in CTEPH patients, which may be useful in estimating prognosis.

Funding Acknowledgement

Type of funding sources: Foundation. Main funding source(s): SFB 1213 area CP-01 projectWilliam G. Kerckhoff-FoundationKerckhoff Heart Research Institute (KHFI)German Center for Cardiovascular Research (DZHK)

A data-pipeline processing electrocardiogram recordings for use in artificial intelligence algorithms

Introduction

Artificial intelligence (AI) can be used for various tasks in medicine and specifically in cardiology. Medical data such as electrocardiogram recordings (ECGs) are widely used and universally accepted as diagnostic and prognostic tools. It has been shown that deep learning methods using ECGs yield excellent results detecting cardiac pathologies. A significant amount of reliable data is required for supervised learning algorithms such as deep learning models. However, only a small fraction of ECG data generated in daily practice is available in a fully digital and machine-readable format, such as XML. Frequently, used ECG devices produce PDF files or even paper-based print outs, which need to be digitised later for inclusion in clinical information systems. Such ECGs cannot be used without further effort for training or application of deep learning models. Therefore, aim of the present project was to develop a data-pipeline that generates machine-readable ECG data for AI use data irrespective of the initial ECG format.

Methods

We propose an end-to-end pipeline that can not only process data from modern digital ECG devices but is also capable of extracting all necessary information from PDF files (both scanned hard copies and digitally generated PDFs) (see Figure 1). By using different techniques including adaption of open source libraries for vectorisation of image data, and modern computer vision technologies, such as optical character recognition (OCR), our pipeline is able to flexibly process data from different recording devices and read both data in PDF format and data from native digital devices delivered in XML. The processed files from various sources are either saved as a common and easily accessible CSV file format, or are processed directly with deep learning models (see Figure 2).

Results

The developed data-pipeline was validated using data from a set of 113 12-lead ECGs for which data was available in multiple formats. Each format dataset was separately processed by our pipeline and then used for training and validation of a deep learning architecture for myocardial scar detection based on raw ECG signals. The quality of the extraction process by our pipeline was assessed by the respective deep learning models with their prediction capability depicted by receiver operator characteristic analyses (ROC). Comparing the benchmark model that was generated from XML data against a model that was purely trained on PDF data processed by the pipeline shows that both models produced comparable results, reaching area under the curve (AUC) values of 0:79±0:10 (XML) and 0:83±0:07 (PDF).

Conclusion

The data pipeline facilitates acceleration of ECG-based AI research and application of AI algorithms by providing access to ECG data irrespective of the format of the stored ECG. Future work will focus on independent validation as well as expanding this pipeline to include additional ECG types.

Funding Acknowledgement

Type of funding sources: Public Institution(s). Main funding source(s): Flexi Funds by Forschungscampus Mittelhessen

On the importance of representative datasets in ECG-based artificial intelligence

Background/Introduction

ECG-based artificial intelligence (AI) is an emerging field in digital cardiology. Training on diseased records vs. healthy controls is common practice. We aimed to evaluate if such an approach can lead to unwanted behaviour in real-world settings and thus unnecessarily reduce diagnostic precision of the developed AI model.

Purpose

Several studies have shown that deep neural networks are able to exceed performance of medical experts. However, when these models are applied to different cohorts, results vary strongly. We hypothesise that this is because the datasets used for training were not representative for the target population.

Methods

Based on the public ECG database PTB-XL we sampled three distinct subsets of n=150 records representing ECG groups labelled for diagnoses 'old myocardial infarction' (M), 'normal ECG' (N), or 'other cardiac abnormality' (O). These groups were combined to three datasets ([M, N] (n=300), [M, O] (n=300), [M, N, O] (n=450)), representing different approaches to data sampling. On each dataset, we trained a separate but equally structured deep neural network using 100-fold bootstrapping. The diagnostic performance of each model was validated on unseen data from all datasets with sensitivity, specificity and area under the receiver operator characteristic curve.

Results

Evaluation of the three differently trained models shows best diagnostic performance on the M vs. N records and worst on the M vs. O records. However, in the out-of-dataset setting, the best-performing model (trained on [M, N]) shows weaker performance on the [M, N, O] and [M, O] datasets. Sensitivity for the same model remained equal, as identical M records were used throughout corresponding bootstrapping folds. Detailed results are presented in Table 1.

Conclusions

Our results suggest that the model trained on a dataset including only diseased records vs. healthy controls [M, N] learned to recognise healthy (N) instead of diseased (M) records, which explains why it performed poorly on datasets including records showing other cardiac abnormalities (O). Such behaviour is a common problem in AI and requires special attention in dataset sampling. For small cohorts, it is tempting to increase the amount of training data by using healthy controls. However, we have shown that this can be a poor option, since classifiers can more easily rely on features that are not actually related to the target disease. Training and validation of classifiers should therefore be performed on representative datasets that are as close as possible to the target population.

Funding Acknowledgement

Type of funding sources: Public Institution(s). Main funding source(s): Forschungscampus Mittelhessen, Flexi Funds Table 1

Predicting mortality in cardiovascular patients using electrocardiogram data and artificial intelligence

Background/Introduction

The electrocardiogram (ECG) is an ubiquitously used non-invasive tool for diagnosis and risk prediction in cardiology, granting deep extensive insights into the heart. Artificial intelligence (AI) is a modern resource allowing the processing of vast complex datasets in a way that is comparable to humans. Risk stratification in cardiovascular patients is mainly based on scoring systems, such as the ESC-SCORE, relying on traditional risk variables like cholesterol levels or arterial hypertension, rather than actual cardiac structure and function. Goal of this project was to predict mortality using AI in patients with cardiovascular risk based on the current cardiac situation represented by a standard 12-lead ECG recording.

Methods

The study population is based on an ongoing registry that started in 2010 and enrolled patients scheduled for an invasive coronary angiography due to suspected chronic coronary syndrome. Data of the following study patients were analysed: enrolment within the first two study years with available long-term follow-up data on the outcome measure overall mortality, availability of an ECG at admission without pacemaker stimulation and availability of all variables needed to calculate the ESC-SCORE (in the version weighed for a German population) as comparison. This led to a cohort of 720 patients, of whom 70 died within the follow-up period. Information on presence of a relevant coronary artery disease (CAD) was available for all patients, to differentiate between primary and secondary prevention. A deep learning architecture that was previously developed to detect myocardial scar in raw ECG time-series data was used. This model was trained with 1400 ECG recordings, from the publicly available PTB-XL dataset with 700 of those ECGs labelled for acute, recent or old myocardial infarction while 700 were labelled as healthy. This pre-trained model was then applied to our study cohort to predict long-term mortality based on a single 12-lead ECG obtained at admission.

Results

For mortality prediction in patients without CAD (primary prevention) the AI model compares to the ESC-SCORE with an AUROC of 0.606 vs 0.584. For CAD patients (secondary prevention) the AI model compares with an AUROC of 0.612 vs 0.658. Detailed results are presented in Table 1.

Conclusion(s)

Our data underlines the potential of an AI based approach, predicting mortality in cardiovascular patients using only single 12-lead ECG recordings. Additionally, our model achieved similar predictive information to established risk classification systems, such as the ESC-SCORE. Since data acquisition is still ongoing, we will continue to improve our model. In future work training AI to specifically predict mortality while also exploring explainable AI could lead to breakthrough findings in ECG interpretation.

Funding Acknowledgement

Type of funding sources: Public Institution(s). Main funding source(s): FlexiFunds by Forschungscampus Mittelhessen

Detecting a broader spectrum of cardiac pathologies in electrocardiogram data by applying a deep neural network designed to detect a specific cardiac disease

Background/Introduction

The electrocardiogram (ECG) is a widely used and inexpensive tool that provides extensive insights into the cardiac structure and function. Artificial intelligence (AI) algorithms, especially deep learning (DL) models, are efficient computer based instruments with which large and complex datasets can be processed for identification of e.g. specific diseases. PhysioNet is a NIH research resource for complex signals including a large amount of labelled ECG time-series data. Our aim was to evaluate the diagnostic performance of an AI architecture developed to detect a specific cardiac pathology in a large ECG data set including a broad range of cardiac abnormalities.

Methods

The PhysioNet ECG dataset provided as part of the PhysioNet Challenge 2020 consists of five distinct databases with a total of 43100 12-Lead ECG recordings of varying length stemming from patients from China, Russia, Europe and the United States. Each ECG recording is annotated with diagnoses based on a set of 111 possible labels, which express either a cardiac pathology, e.g. atrial flutter or anterior wall ischemia, or unspecific changes in the ECG, e.g. a prolonged qt interval or low qrs voltages. Based on these labels we defined 10 groups merging PhysioNet labels describing related cardiac abnormalities (see Table 1). We adapted a recently published DL model which used raw ECG time-series data of all 12-leads rather than extracted features as model input. This DL model was adapted to the larger number of output variables and then trained on 80% (n=34480 ECGs) of the PhysioNet dataset. The remaining 20% (n=8620 ECGs) of the PhysioNet dataset were used to evaluate the diagnostic performance of the AI model. Sensitivities, specificities and the areas under the receiver operator characteristic curves (AUROC) were used as performance metrices.

Results

The AI model, that was initially designed to detect a specific cardiac pathology, performed well in the large PhysioNet dataset providing AUROCs ranging from 0.78 to 0.95 to detect the defined 10 cardiac abnormality groups. Interestingly, the AI model was able to detect disease groups with changes in the chronological sequence of the ECG, e.g. arrhythmia, with comparable precision as disease groups associated primarily with changes in the ECG amplitude like e.g. ischemia. Detailed results are presented in Table 2.

Conclusion(s)

Our evaluation shows that an AI model that uses raw ECG time-series data rather than extracted features as model input can be easily transferred to other large datasets with different prediction variables. This might also serve as a proof of concept that raw data instead of pre-selected features should be used as model input if developing AI applications for medical use cases.

Funding Acknowledgement

Type of funding sources: Public Institution(s). Main funding source(s): FlexiFunds by Forschungscampus Mittelhessen

Dark Energy Survey Year 1 Results: Cosmological Constraints from Cluster Abundances, Weak Lensing, and Galaxy Correlations

We present the first joint analysis of cluster abundances and auto or cross-correlations of three cosmic tracer fields: galaxy density, weak gravitational lensing shear, and cluster density split by optical richness. From a joint analysis (4×2pt+N) of cluster abundances, three cluster cross-correlations, and the auto correlations of the galaxy density measured from the first year data of the Dark Energy Survey, we obtain Ω_{m}=0.305_{-0.038}^{+0.055} and σ_{8}=0.783_{-0.054}^{+0.064}. This result is consistent with constraints from the DES-Y1 galaxy clustering and weak lensing two-point correlation functions for the flat νΛCDM model. Consequently, we combine cluster abundances and all two-point correlations from across all three cosmic tracer fields (6×2pt+N) and find improved constraints on cosmological parameters as well as on the cluster observable-mass scaling relation. This analysis is an important advance in both optical cluster cosmology and multiprobe analyses of upcoming wide imaging surveys.

Constraints on Dark Matter Properties from Observations of Milky Way Satellite Galaxies

We perform a comprehensive study of Milky Way (MW) satellite galaxies to constrain the fundamental properties of dark matter (DM). This analysis fully incorporates inhomogeneities in the spatial distribution and detectability of MW satellites and marginalizes over uncertainties in the mapping between galaxies and DM halos, the properties of the MW system, and the disruption of subhalos by the MW disk. Our results are consistent with the cold, collisionless DM paradigm and yield the strongest cosmological constraints to date on particle models of warm, interacting, and fuzzy dark matter. At 95% confidence, we report limits on (i) the mass of thermal relic warm DM, m_{WDM}>6.5  keV (free-streaming length, λ_{fs}≲10h^{-1}  kpc), (ii) the velocity-independent DM-proton scattering cross section, σ_{0}<8.8×10^{-29}  cm^{2} for a 100 MeV DM particle mass [DM-proton coupling, c_{p}≲(0.3  GeV)^{-2}], and (iii) the mass of fuzzy DM, m_{ϕ}>2.9×10^{-21}  eV (de Broglie wavelength, λ_{dB}≲0.5  kpc). These constraints are complementary to other observational and laboratory constraints on DM properties.

Shift in transcriptional landscape of human right ventricle in chronic thromboembolic pulmonary arterial hypertension

Background

Chronic thromboembolic pulmonary hypertension (CTEPH) is a sub group of pulmonary hypertension (PH). CTEPH is characterized by the existence of thromboemboli and vascular remodeling in pulmonary vessels. The effect of increase in pulmonary artery pressures causes right ventricle (RV) hypertrophy and dilatation and finally leads to right heart failure and death. Surgical intervention in operable patients makes the CTEPH as an only curable and unique form of ph. Pulmonary endarterectomy (PEA) is the surgical procedure to remove the thromboembolic clots from the pulmonary vasculature, which restores RV function back to normal with significant improvements in cardiovascular magnetic resonance.

Purpose

The aim of this study is to use transcriptomic profiling to identify signaling pathways, master regulators, and potentially new biomarkers that specifically indicate the effect of PEA on the RV of patients with chronic thromboembolic pulmonary hypertension.

Results

RNA -sequencing (RNA-seq) was performed on RV biopsies obtained from CTEPH patients at PEA baseline (before PEA surgery) and the results were compared with those from RV biopsies obtained during follow-up evaluation. Bioinformatic analysis of RNA-seq data identified 2799 genes (n=14, −0.585 ≤ Log2 fold change ≥0.585, FDR ≤0.05) differentially regulated between the PEA baseline and follow-up sample groups. The great number of genes (2799) differentially expressed after PEA surgery in CTEPH patients confirms a major shift in the transcriptional landscape of RV in these patients. To further identify potential biomarker candidates from the large pool of 2799 differentially expressed genes (DEGs), extensive bioinformatic analysis of different data sets shortlisted 250 DEGs that were functionally associated with cardiovascular development or disease. The findings of this study reveal prominent transcriptional changes that occur in response to PEA. Gene ontology enrichment and pathway analysis confirmed altered regulation of hypoxia-inducible factor 1 (HIF-1) signaling, advanced glycation end products and their receptors (AGE-RAGE), mitogen-activated protein kinase (MAPK) signaling, hippo signaling, the Janus kinase/ signal transducers and activators of transcription (Jak-STAT) signaling pathway, and proteoglycans after PEA compared with before PEA.

Conclusion

Comparison of the results of RNA-seq analysis of RV biopsies of CTEPH patients, pre and post PEA, revealed a major shift in the transcriptional landscape of these patients after reducing the pressure overload of the RV by PEA.

Funding Acknowledgement

Type of funding source: Foundation. Main funding source(s): German Research Foundation (DFG)

Detection of Cross-Correlation between Gravitational Lensing and γ Rays

In recent years, many γ-ray sources have been identified, yet the unresolved component hosts valuable information on the faintest emission. In order to extract it, a cross-correlation with gravitational tracers of matter in the Universe has been shown to be a promising tool. We report here the first identification of a cross-correlation signal between γ rays and the distribution of mass in the Universe probed by weak gravitational lensing. We use data from the Dark Energy Survey Y1 weak lensing data and the Fermi Large Area Telescope 9-yr γ-ray data, obtaining a signal-to-noise ratio of 5.3. The signal is mostly localized at small angular scales and high γ-ray energies, with a hint of correlation at extended separation. Blazar emission is likely the origin of the small-scale effect. We investigate implications of the large-scale component in terms of astrophysical sources and particle dark matter emission.

P5478Information on weather conditions improves the prognostic ability of 25 OH-vitamin D in stable coronary artery disease

Introduction and aim

Vitamin D deficiency is associated with an adverse prognosis in patients with coronary artery disease (CAD). Decreased levels of vitamin D are associated with low sunshine exposure, resulting in seasonal variations of vitamin D. The aims of this study were to investigate the influence of different specific weather conditions on vitamin D levels and to explore a possible improvement of risk stratification by vitamin D levels in stable patients with CAD using meteorological data.

Methods

The study population consists of two independent cohorts of stable patients undergoing coronary angiography with suspected or known CAD: as derivation cohort, the ongoing biomarker registry BioPROSPECTIVE with n=1,766 enrolled patients between 2010 and 2013 (median age 70.1 yrs; 30.8% females); and as validation cohort, the Ludwigshafen Risk and Cardiovascular Health (LURIC) Study with n=3,299 patients (median age 63.5 yrs; 30.3% females). In the derivation cohort 235 (13.3%) patients were known to be deceased by 08/2018. In the validation cohort 760 (23.0%) patients died within a median follow-up time of 7.75 years. 25-OH vitamin D levels were measured by commercial assays. Vitamin D deficiency was defined as 25-OH vitamin D levels ≤20 ng/mL. Daily averaged data on six weather conditions of the 180 days prior to enrolment were collected for each patient from the weather station located closest to the respective study centre. Using air pressure, precipitation height, sunshine duration, temperature, relative humidity, and vapour pressure a weather model was constructed that significantly correlated with vitamin D levels (r=0.37; p<0.001).

Results

In the derivation cohort, median vitamin D levels were lower in non-survivors (13.3 [9.65–19.65] ng/mL) than in survivors (15.70 [10.7–22.65] ng/mL; p<0.001). Vitamin D predicted all-cause mortality with an area under the receiver operator characteristic curve (AUROC) of 0.576 (CI: 0.54–0.62). Adding the weather model to vitamin D significantly improved the AUROC to 0.601 (CI: 0.56–0.64; p=0.031). The vitamin D/weather model combination enhanced the prognostic value of the ESC SCORE to predict mortality (AUROC=0.571 [CI: 0.53–0.61] vs. 0.628 [CI: 0.59–0.67]; p=0.004). Comparable results were observed in the validation cohort. Here, vitamin D deficiency predicted mortality with a hazard ratio (HR) of 1.89 (CI: 1.59–2.26) after adjustment for ESC SCORE. Adding the weather model improved this HR to 1.92 (1.62–2.32). Reclassification analyses support the additive prognostic information of weather conditions with a continuous net reclassification improvement of 0.114 ([0.033–0.194]; p=0.006) if adding the weather model to vitamin D as base model for predicting mortality.

Conclusions

Different weather conditions show a significant impact on vitamin D levels in stable patients. Adding data on weather conditions improve the risk stratification by vitamin D for predicting mortality in stable CAD patients.

Acknowledgement/Funding

The study is financially supported by the Kerckhoff Heart Research Institute (KHFI) and the German Center for Cardiovascular Research (DZHK).

Cosmological Constraints from Multiple Probes in the Dark Energy Survey

The combination of multiple observational probes has long been advocated as a powerful technique to constrain cosmological parameters, in particular dark energy. The Dark Energy Survey has measured 207 spectroscopically confirmed type Ia supernova light curves, the baryon acoustic oscillation feature, weak gravitational lensing, and galaxy clustering. Here we present combined results from these probes, deriving constraints on the equation of state, w, of dark energy and its energy density in the Universe. Independently of other experiments, such as those that measure the cosmic microwave background, the probes from this single photometric survey rule out a Universe with no dark energy, finding w=-0.80_{-0.11}^{+0.09}. The geometry is shown to be consistent with a spatially flat Universe, and we obtain a constraint on the baryon density of Ω_{b}=0.069_{-0.012}^{+0.009} that is independent of early Universe measurements. These results demonstrate the potential power of large multiprobe photometric surveys and pave the way for order of magnitude advances in our constraints on properties of dark energy and cosmology over the next decade.

Effect of 3-nitrooxypropanol on methane and hydrogen emissions, methane isotopic signature, and ruminal fermentation in dairy cows

The objective of this crossover experiment was to investigate the effect of a methane inhibitor, 3-nitrooxypropanol (3NOP), on enteric methane emission, methane isotopic composition, and rumen fermentation and microbial profile in lactating dairy cows. The experiment involved 6 ruminally cannulated late-lactation Holstein cows assigned to 2 treatments: control and 3NOP (60 mg/kg of feed dry matter). Compared with the control, 3NOP decreased methane emission by 31% and increased hydrogen emission from undetectable to 1.33 g/d. Methane emissions per kilogram of dry matter intake and milk yield were also decreased 34% by 3NOP. Milk production and composition were not affected by 3NOP, except milk fat concentration was increased compared with the control. Concentrations of total VFA and propionate in ruminal fluid were not affected by treatment, but acetate concentration tended to be lower and acetate-to-propionate ratio was lower for 3NOP compared with the control. The 3NOP decreased the molar proportion of acetate and increase those of propionate, butyrate, valerate, and isovalerate. Deuterium-to-hydrogen ratios of methane and the abundance of (13)CH3D were similar between treatments. Compared with the control, minor (4‰) depletion in the (13)C/(12)C ratio was observed for 3NOP. Genus composition of methanogenic archaea (Methanobrevibacter, Methanosphaera, and Methanomicrobium) was not affected by 3NOP, but the proportion of methanogens in the total cell counts tended to be decreased by 3NOP. Prevotella spp., the predominant bacterial genus in ruminal contents in this experiment, was also not affected by 3NOP. Compared with the control, Ruminococcus and Clostridium spp. were decreased and Butyrivibrio spp. was increased by 3NOP. This experiment demonstrated that a substantial inhibition of enteric methane emission by 3NOP in dairy cows was accompanied with increased hydrogen emission and decreased acetate-to-propionate ratio; however, neither an effect on rumen archaeal community composition nor a significant change in the isotope composition of methane was observed.

Wide-Field Lensing Mass Maps from Dark Energy Survey Science Verification Data

We present a mass map reconstructed from weak gravitational lensing shear measurements over 139  deg2 from the Dark Energy Survey science verification data. The mass map probes both luminous and dark matter, thus providing a tool for studying cosmology. We find good agreement between the mass map and the distribution of massive galaxy clusters identified using a red-sequence cluster finder. Potential candidates for superclusters and voids are identified using these maps. We measure the cross-correlation between the mass map and a magnitude-limited foreground galaxy sample and find a detection at the 6.8σ level with 20 arc min smoothing. These measurements are consistent with simulated galaxy catalogs based on N-body simulations from a cold dark matter model with a cosmological constant. This suggests low systematics uncertainties in the map. We summarize our key findings in this Letter; the detailed methodology and tests for systematics are presented in a companion paper.

Prevalence of nonclassical steroid 21-hydroxylase deficiency based on a morning salivary 17-hydroxyprogesterone screening test: a small sample study

Early morning salivary 17 alpha-hydroxyprogesterone (17-OHP) determination differentiates patients with non-classical 21-hydroxylase deficiency (NC21OHD) from those who are not affected. Using this test, we have conducted a trial screening study for NC21OHD and have compared the study results with previously reported figures for the frequency of this disorder. Testing was performed on 258 subjects recruited from among the medical students and employees of the New York Hospital-Cornell Medical Center. In 2 of the 249 admissible subjects, the 0700-0900 h salivary 17-OHP level was within the range for NC21OHD patients (0.72-6.7 nmol/L; n = 8). These 2 individuals were subsequently confirmed to be affected by ACTH testing. Of the subjects with morning salivary 17-OHP levels below the cut-off point of 0.72 nmol/L, 29 were recalled for ACTH testing and were confirmed to be unaffected. Prevalence of NC21OHD in the test population was determined according to ethnic group. Our study gives a prevalence by screening of 1.14% among caucasians, which agrees with values of 0.81% and 1.06% obtained by different analytical methods. Further, both affected subjects were Ashkenazi Jews, and the prevalence of 3.23% among study members from this group concurs with increased rates of 3.64% and 4.97% already reported. On the basis of a small population sample, screening so far confirms the claim that NC21OHD is the most common autosomal recessive human disorder. Using values from ACTH-proven unaffected subjects (n = 47) and NC21OHD patients (n = 10), we establish preliminary normative data for morning salivary 17-OHP levels of 0.172 nmol/L for unaffected subjects (95% confidence interval, 0.05-0.54 nmol/L) and 1.76 nmol/L for NC21OHD-affected subjects (95% confidence interval, 0.42-7.32 nmol/L).