Age- and Sex-specific Reference Ranges for Cardiac Function and Structure in Germany: Cardiovascular Magnetic Resonance Imaging (CMR) in the German National Cohort (NAKO)

This study establishes age- and sex-specific reference ranges for cardiac structure and function using cardiovascular magnetic resonance imaging data from a large, population-based German cohort to provide a normative framework for clinical interpretation and research on healthy aging.

The Gist

Imagine you are trying to tune a car engine. To know if the engine is running perfectly, you need a "standard" manual that tells you exactly how much fuel it should use, how fast the pistons should move, and how big the engine block should be for a car of a specific size and age. If you don't have that manual, you might think a slightly older, slightly smaller engine is broken, when it's actually just doing exactly what it should be doing.

This paper is essentially building that "Owner's Manual" for the human heart, but specifically for people living in Germany.

Here is the story of how they did it, broken down into simple concepts:

1. The Problem: We Were Guessing

For a long time, doctors had to guess what a "normal" heart looked like. They used small studies or data from very specific groups of people. It was like trying to guess the average height of all humans by only measuring basketball players. This made it hard to tell if a patient's heart was truly sick or just different because of their age or gender.

2. The Solution: A Massive "Heart Census"

The researchers used data from the NAKO, a huge German project that is like a giant census of health. They didn't just look at a few hundred people; they looked at nearly 25,000 healthy adults (ages 20 to 72).

• The Tool: They used CMR (Cardiovascular Magnetic Resonance). Think of this as a super-powered, 3D camera that can see inside the heart without using radiation. It's the "gold standard" because it's incredibly precise.
• The Robot Helpers: Instead of humans manually measuring every single heart (which would take forever and be prone to errors), they used AI (Artificial Intelligence). They trained a smart computer program to look at the heart images and measure the chambers, muscle mass, and pumping power automatically.
• The Quality Check: Just like a teacher grading a test, they had a strict system to throw out any blurry or bad images. Only the "A+" quality scans made it into the final book.

3. The Two "Normal" Groups

The researchers created two different versions of the "Normal Heart Manual":

• The "Real World" Group: This includes people who don't have heart disease, but they might have high blood pressure, high cholesterol, or be a bit overweight. This is the "average healthy German."
• The "Super Healthy" Group: This is a smaller group of people who have no heart disease AND no risk factors (no smoking, no high blood pressure, etc.). This is the "ideal" heart.

4. The Big Discoveries (The "Rules" of the Heart)

By analyzing these thousands of hearts, they found some fascinating patterns that change as we get older:

• The Shrinking Engine: As people get older, their heart chambers (the rooms inside the heart) actually get slightly smaller. It's like a house that gets a little more compact over time.
• The Stronger Pump: Even though the rooms get smaller, the heart's ability to pump blood (the Ejection Fraction) stays remarkably steady. It's like a small engine that is tuned to run just as efficiently as a big one.
• Men vs. Women (The Size Difference):
  • Men generally have bigger hearts and larger chambers, just like men generally have bigger bodies.
  • Women have slightly smaller hearts but pump blood with slightly higher efficiency.
  • The Convergence: The biggest difference between men and women is when they are young. As they get older (past age 50), their hearts start to look more similar. The researchers suspect this is because of hormonal changes (like menopause) and lifestyle factors that make older hearts look more alike.
• The Wall Thickening: As we age, the muscle walls of the heart get a tiny bit thicker. This is the heart's way of adapting to the years of work it has done.

5. Why This Matters

This paper is a game-changer for doctors.

• Before: A doctor might see a 65-year-old man with a slightly smaller heart and think, "Oh no, his heart is shrinking because of disease!"
• Now: The doctor can open this new "Manual," look up "65-year-old male," and see, "Ah, a heart that size is actually perfectly normal for his age."

The Bottom Line

This study provides a customized, age-specific, and gender-specific rulebook for what a healthy heart looks like in Germany. It helps doctors distinguish between a heart that is simply aging gracefully and a heart that is actually sick. It's like finally getting the exact factory specifications for your car, so you know exactly when to worry and when to just keep driving.

Technical Summary

1. Problem Statement

Cardiovascular Magnetic Resonance (CMR) is the clinical reference standard for quantifying cardiac structure and function. However, widely applicable, population-based reference values are limited. Existing reference ranges often rely on:

• Small, specialized cohorts of healthy volunteers.
• Meta-analyses of heterogeneous studies with varying acquisition protocols.
• Large cohorts (e.g., UK Biobank) that are skewed toward older populations (45–74 years) and lack sufficient representation of younger adults (<40 years).
• A lack of standardized, age- and sex-specific modeling that accounts for non-linear aging trajectories and interactions between sex and age.

There is a critical need for a large-scale, uniformly imaged dataset to define "normal" cardiac parameters across the entire adult lifespan, distinguishing between a broad "CVD-free" population and a strictly "healthy" (risk-factor-free) subcohort.

2. Methodology

Study Design and Cohort:

• Source: The German National Cohort (NAKO), a prospective population-based study involving 205,415 participants.
• Imaging Substudy: A nested CMR substudy conducted at five specialized centers using identical 3T Siemens MAGNETOM Skyra scanners with uniform software versions.
• Participants:
  • Initial: 30,868 participants underwent whole-body MRI.
  • Main Reference Cohort: 24,371 participants (mean age 47.6 years; 44.7% women) free of manifest Cardiovascular Disease (CVD).
  • Healthy Subcohort: 5,550 participants (23% of the main cohort) free of both CVD and traditional cardiovascular risk factors (hypertension, diabetes, hypercholesterolemia, obesity, smoking).
• Exclusions: Participants with insufficient image quality, no CMR data, or prevalent/unclear CVD history were excluded.

Image Acquisition and Analysis:

• Protocol: Short-axis (SAX) balanced steady-state free precession (SSFP) cine images covering the heart from base to apex (12 slices, 25 phases).
• Segmentation: A validated deep learning pipeline (nnU-Net ensemble) was used for automated segmentation of LV/RV endocardium and epicardium.
  • Includes 2D and 3D U-Net models to leverage slice context.
  • Papillary muscles were included in the blood pool and excluded from mass calculations.
• Quality Control (QC): A rigorous, structured visual review was performed on outliers. Images were rated on a 5-point scale for image quality and segmentation accuracy. Only participants with "complete" (5) or "high" (4) confidence ratings in both domains were included.

Statistical Analysis:

• Modeling: Generalized Additive Models (GAMs) with cubic spline bases (up to 20 knots) were used to model non-linear age effects.
• Reference Intervals: Additive quantile regression was employed to derive the 5th and 95th percentiles (reference ranges) for all metrics.
• Stratification: Analyses were stratified by sex. Formal testing (F-tests) was conducted to evaluate age-sex interactions.
• Indexing: Metrics were indexed to Body Surface Area (BSA), height (m$^{1.7}$), and raw values.

3. Key Contributions

1. Scale and Uniformity: The study provides reference ranges for the largest uniformly imaged national cohort to date (N=24,371), eliminating inter-scanner and inter-protocol variability common in multi-center meta-analyses.
2. Age Range: It significantly extends the age range of reference data, with robust representation of young adults (20–40 years), a demographic often underrepresented in previous large-scale CMR studies.
3. Two-Tier Definition of Health: It offers reference ranges for both a "real-world" CVD-free population and a strictly "healthy" subcohort, allowing clinicians to choose the appropriate benchmark based on patient risk profiles.
4. Advanced Modeling: The use of quantile regression and GAMs allows for the capture of non-linear age trends and heteroscedasticity, providing smooth percentile curves rather than static cut-offs.
5. Sex-Age Interactions: The study explicitly models and quantifies how the relationship between age and cardiac metrics differs by sex, revealing that sex differences are most pronounced in younger ages and attenuate in later life.

4. Key Results

Demographics and Cohort Characteristics:

• The main cohort had a median 10-year Framingham risk of 8.0% and a prevalence of hypertension (36.7%) and hypercholesterolemia (31.8%).
• The healthy subcohort was younger (mean 40.7 vs. 47.6 years) and had significantly lower risk scores.

Physiological Trends:

• Volumes: Left Ventricular (LV) and Right Ventricular (RV) end-diastolic and end-systolic volumes declined with age in both sexes. The decline was steeper in men in absolute terms.
• Ejection Fraction (EF): Both LV and RV EF remained largely stable across the age range.
  • Sex Differences: Women consistently exhibited higher EF (LVEF ~64–65% vs. men ~62%; RVEF ~55–59% vs. men ~53–56%).
• Mass and Wall Thickness:
  • LV Mass showed stable to slightly decreasing levels with age.
  • Wall thickness (including mid-septal) increased with age, indicating concentric remodeling.
• Sex Differences in Size: Men had significantly larger absolute volumes (+30–40% for LVEDV) compared to women. After BSA indexing, differences persisted but were attenuated.

Age-Sex Interactions:

• Significant interactions were found for all major metrics (p<0.05).
• Convergence: Sex differences in cardiac size and function were most pronounced at younger ages (20–40) and converged (became less distinct) in older age groups (>60).
• Risk Factor Impact: Adjusting for CVD risk factors in the main cohort explained very little additional variance in most metrics (e.g., <1% for LVEF), suggesting that the "CVD-free" reference ranges are robust even in the presence of risk factors. However, mid-septal wall thickness showed a larger contribution from risk factors.

Comparison of Cohorts:

• Reference ranges derived from the "healthy" subcohort were similar to the main reference cohort, with no clinically relevant deviations, though the healthy cohort showed slightly lower mass and higher stroke volumes.

5. Significance

• Clinical Utility: These data provide a nuanced, percentile-based framework for interpreting CMR studies. Clinicians can distinguish between "borderline" findings in older adults (which may be normal aging) and pathological values in younger adults.
• Standardization: By utilizing a single national cohort with harmonized 3T scanners and automated deep learning analysis, these ranges reduce technical heterogeneity, offering a more reliable benchmark than pooled international data.
• Research Foundation: The dataset serves as a baseline for future research into "biological heart age," sex-specific cardiac remodeling, and the impact of specific risk factors on cardiac structure.
• Policy and Guidelines: The findings support the development of updated clinical guidelines that account for age- and sex-specific trajectories, moving away from static "normal" cut-offs toward dynamic reference intervals.

In conclusion, this study establishes a new, high-resolution normative framework for cardiac CMR in Germany, addressing critical gaps in age representation and standardization, and highlighting the complex interplay between aging, sex, and cardiac structure.