Identifying Left Ventricular Hypertrophy on an ECG

The standard ECG can suggest left ventricular hypertrophy (LVH) through increased QRS voltage in specific leads, but the tracing misses the majority of confirmed cases. Across multiple studies, the most commonly used voltage criteria detect only about one in five to one in three people whose hearts are genuinely thickened when checked by imaging. That gap between what the ECG promises and what it delivers makes recognizing both the classic patterns and their limitations essential for anyone trying to interpret one of these tracings.

What the ECG Is Actually Measuring

The traditional reasoning is straightforward: a thicker left ventricle has more muscle, which generates a stronger electrical signal, which shows up as taller waves on the ECG. That logic has been the backbone of voltage-based LVH criteria for decades. But the relationship is not that clean. The voltage recorded on the surface depends not just on how much muscle is firing but on the electrical properties of that muscle and the tissue surrounding it. Hypertrophied heart cells develop structural changes, including fibrosis in the tissue between cells, that alter how electricity moves through the wall. These changes can either amplify or dampen the signal that reaches the skin electrodes, which is one reason the same degree of thickening can look different on two people’s ECGs.

The Major Voltage Criteria and How They Work

Several scoring systems have been developed to flag LVH on a 12-lead ECG. They all share the same basic idea: measure the height of certain waves and compare them to a threshold. What differs is which leads they look at and where they draw the line.

The Sokolow-Lyon index adds the depth of the S wave in lead V1 to the height of the R wave in V5 or V6 (whichever is taller). If the sum exceeds 35 mm, LVH is suggested. It has been in use since 1949 and remains one of the most widely taught criteria. The Cornell voltage criterion takes a different approach, adding the R wave in lead aVL to the S wave in V3. The threshold is 28 mm for men and 20 mm for women. The Cornell product multiplies that voltage sum by the QRS duration, with a cutoff around 2440 mm·ms.

The Peguero-Lo Presti criterion, a newer addition, sums the deepest S wave found in any lead with the S wave in V4. Its threshold is 28 mm for men and 23 mm for women. Point-based systems like the Romhilt-Estes score take a broader view, awarding points for voltage, strain-pattern ST changes, left atrial abnormality, left axis deviation, and prolonged QRS duration. A score of 5 or more is considered definite LVH; 4 is probable.

The Sensitivity Problem

The uncomfortable truth about all these criteria is that they are very good at ruling LVH in when they are positive, but they miss most cases. Specificity tends to sit above 90%, meaning a positive result is usually real. Sensitivity, though, is a different story. In one head-to-head comparison using echocardiography as the reference, the Sokolow-Lyon index correctly identified LVH in only about 2% of confirmed cases, Cornell voltage in about 4%, the Romhilt-Estes score in about 10%, and the Peguero-Lo Presti criterion in about 19%.

Larger studies with broader populations show somewhat better numbers but the same overall pattern. In one analysis from the Groningen cohort, the highest sensitivity achieved in men was 44% using a 12-lead voltage sum, while in women the Peguero-Lo Presti criteria topped out at 31%. Specificity remained above 90% for every criterion tested.

The practical implication is that a normal ECG does not rule out LVH. If there is clinical suspicion based on longstanding high blood pressure, symptoms, or risk factors, imaging with echocardiography or cardiac MRI is needed to confirm or exclude it. The ECG is a screening tool, not a definitive one.

How the Cause of Hypertrophy Affects Detection

Not all LVH looks the same on an ECG, partly because the underlying cause shapes the geometry of the thickening. In hypertensive heart disease and aortic stenosis, the ventricle typically thickens concentrically, with the walls growing inward. In conditions causing volume overload, the chamber dilates and the walls stretch in what is called eccentric hypertrophy. Computational modeling shows that eccentric hypertrophy tends to increase voltage in the precordial (chest) leads, which is what most voltage criteria are built to detect, while concentric hypertrophy produces more variable changes across all 12 leads without a clear pattern of increase.

⁴

In aortic valve disease studied with cardiac MRI as the reference, researchers found no significant difference in standard voltage criteria between concentric and eccentric LVH. However, the timing of the intrinsicoid deflection in V6 was shorter in concentric hypertrophy, and the ST-segment depression and T-wave inversion in the anterolateral leads were deeper.

⁵

When the cause is something other than high blood pressure or valve disease, the ECG picture shifts further. One study comparing Fabry disease, cardiac amyloidosis, and hypertrophic cardiomyopathy against hypertensive heart disease and aortic stenosis found that a combined analysis of the PQ interval, corrected QT duration, and the Sokolow-Lyon index could help distinguish between these conditions with high accuracy.

⁶

The diagnostic difficulty is even worse in specific populations. Among anabolic-androgenic steroid users with echocardiographic LVH, roughly two-thirds to seven in ten went undetected by even the better-performing ECG criteria.

⁷

Why Obesity Makes ECG Detection Harder

Excess body weight is one of the biggest confounders in ECG-based LVH detection, and it works against you in two directions at once. Extra tissue between the heart and the chest electrodes insulates the signal, dampening the voltages that criteria like Sokolow-Lyon depend on. Data from the large LIFE study showed that obese and overweight patients had dramatically lower rates of ECG-detected LVH by Sokolow-Lyon criteria compared to normal-weight patients (about 11% versus 31%), even though obesity itself is a risk factor for actual ventricular thickening.

⁸

The Cornell product, by contrast, actually performed better in heavier patients in that same study, with a higher prevalence of ECG-detected LVH in obese participants. This divergence means that the choice of criterion matters more in overweight individuals than in lean ones.

Research using cardiac MRI as the gold standard found that obesity reduced sensitivity through voltage attenuation across multiple criteria, while also lowering specificity because obese people without true hypertrophy often have left ventricular remodeling that mimics some LVH features on the tracing.

⁹

One group tried to fix the Sokolow-Lyon problem by adjusting for the leftward shift in heart position that occurs with increasing body size. After correction, sensitivity in obese patients climbed from about 3% to 27%, approaching the 38% seen in normal-weight individuals.

¹⁰

Sex and Race Differences

Women tend to have lower baseline QRS voltages than men, and the standard voltage criteria consistently perform worse in women even after the thresholds are adjusted. In the Groningen data, sensitivities for most criteria were lower in women than in men, with the exception of the Cornell criteria and the Sokolow-Lyon product, which actually performed slightly better in women.

¹¹ An earlier study using receiver operating characteristic analysis concluded that the reduced accuracy in women is not entirely explained by their smaller body size and lower heart mass, meaning there are intrinsic sex-related differences in cardiac electrical properties that the current criteria do not fully account for.¹²

Recalibration against cardiac MRI found that the best-performing criterion differed by sex: the Sokolow-Lyon product was most sensitive in women, while the Cornell and Cornell product criteria were strongest in men. Even with optimized cutoffs, though, sensitivity at 95% specificity topped out around 26%.

¹³

Race introduces another layer of complexity. In older populations, ECG amplitudes used in LVH criteria were substantially higher in African Americans, producing an apparent LVH prevalence two to three times greater than in white men and women, despite no significant racial difference in actual left ventricular mass measured by echocardiography. Part of this gap in men was attributable to smaller chest diameter, but in women the reasons remained largely unexplained.

¹⁴

The Strain Pattern and What It Means for Prognosis

Beyond voltage, the ECG can show a pattern of ST-segment depression and asymmetric T-wave inversion in the lateral leads (V5, V6, and sometimes I and aVL) that is referred to as the “strain” pattern. The ST segment slopes downward with a convex shape, and the T wave flips in the direction opposite to the main QRS deflection. This pattern indicates that the thickened muscle is under hemodynamic stress, and it carries clinical weight beyond what the voltage criteria alone tell you.

In the LIFE hypertension study, strain was present in about 11% of patients at baseline. Those with strain had roughly double the risk of cardiovascular death and heart attack compared to those without it, even after accounting for blood pressure, standard risk factors, and the severity of LVH on the ECG. After full adjustment, strain remained a significant independent predictor of cardiovascular mortality, with a hazard ratio around 1.5.

¹⁵

A broader review confirmed that the strain pattern is associated with increased all-cause and cardiovascular morbidity and mortality, and is the strongest ECG marker for risk stratification among hypertensive patients when LVH criteria are used.

¹⁶

Why Conduction Blocks and Electrode Placement Complicate Things

Left bundle branch block (LBBB) completely changes the way the left ventricle activates electrically, widening the QRS complex and making standard voltage criteria unreliable. In a classic study, echocardiography showed LVH in 89% of patients with LBBB, but the standard ECG voltage criteria could not reliably identify it. The best predictors in that setting turned out to be left atrial abnormality on the ECG and an enlarged heart shadow on chest X-ray, not the usual voltage measurements.

¹⁷

Even in patients with normal conduction, the reproducibility of voltage criteria is worse than most people assume. Random small shifts in electrode position from one recording to the next substantially impair day-to-day consistency of the Cornell voltage, Sokolow-Lyon, and Romhilt-Estes criteria. The strain pattern and the Cornell strain criterion showed less variability, making them somewhat more reliable for serial comparisons.

¹⁸

Tracking Regression Over Time

One of the most clinically useful things ECG-LVH criteria can do is track whether ventricular thickening is getting better or worse with treatment. In the LIFE trial, lower in-treatment Cornell product values were associated with roughly a 15% decrease in the combined cardiovascular endpoint and a 22% lower risk of cardiovascular death per standard-deviation decrease. Lower Sokolow-Lyon voltage similarly tracked with a 17% drop in the composite endpoint and about a 20% reduction in cardiovascular death.

¹⁹

The link to sudden cardiac death is particularly striking. Absence of LVH by both Sokolow-Lyon and Cornell product criteria during treatment was associated with a 30% lower risk of sudden cardiac death, even after adjusting for blood pressure, heart rate, diabetes, and a long list of other variables.

²⁰

Conversely, persistence or new development of LVH by both criteria during antihypertensive therapy signals markedly higher risk for cardiovascular events and death.

²¹

Athletes and the Problem of Benign Thickening

Regular intense exercise causes the heart to enlarge and the walls to thicken as a normal adaptation, sometimes called athlete’s heart. This can produce increased QRS voltages on an ECG that meet standard LVH criteria in perfectly healthy people. Distinguishing this benign remodeling from hypertrophic cardiomyopathy (HCM), a condition that can cause sudden cardiac death in young athletes, is one of the most consequential diagnostic challenges in sports medicine.

A meta-analysis in the pediatric and adolescent population found that ECG features can help tell the two apart. Prolonged QTc intervals, T-wave abnormalities, ST-segment changes, and pathological Q waves were significantly more common in HCM than in athlete’s heart. When additional features like axis deviation, atrial enlargement, and bundle branch blocks were considered, the distinction became even clearer.

²²

The key takeaway is that isolated voltage criteria meeting LVH thresholds in a young, fit person are not alarming on their own. It is the accompanying features, including repolarization abnormalities and conduction disturbances, that raise the red flag for pathological hypertrophy.

Children and Age-Dependent Normals

Pediatric ECG interpretation for LVH uses different reference values because normal QRS voltages change dramatically with age. In infants, right ventricular dominance is normal, and the transition to adult left-dominant patterns happens over the first few years of life. A study of over 1,900 children derived age-dependent percentile curves and found that at 95% specificity, standard ECG criteria had sensitivities below 25% when an elevated left ventricular mass index was the reference. Sensitivity improved to under 43% when clinical evidence of LVH was also factored in. Combining ECG parameters boosted performance further, and a modified Sokolow-Lyon criterion (using V3R and V7 rather than the standard leads) performed best across age groups.

²³

Where AI Fits In

Machine learning models trained on large ECG databases paired with imaging data represent the most promising avenue for improving detection. A systematic review and meta-analysis of AI-based approaches found a summary area under the curve of 0.87 for AI models, compared to 0.68 for Sokolow-Lyon and 0.60 for Cornell criteria. The AI models achieved a sensitivity of 69% with a specificity of 87%, a major jump over the roughly 19-25% sensitivity typical of traditional criteria at similar specificity.

²⁴

One deep learning model trained to infer cardiac MRI-derived left ventricular mass from standard 12-lead ECGs showed that its mass estimates correlated with actual MRI measurements and were associated with incident cardiovascular disease. In one validation cohort, the AI model had statistically superior discrimination compared to any traditional ECG rule; in another large cohort it performed comparably.

²⁵

These tools are not yet standard in everyday clinical practice, but they are being built into commercial ECG interpretation software at some institutions. Their main advantage is extracting information from the full waveform morphology rather than relying on a few hand-picked measurements, which lets them pick up on subtler electrical changes that voltage criteria ignore. The sensitivity gap between traditional criteria and imaging may never fully close with a surface tracing alone, but AI is narrowing it faster than any recalibration of cutoffs has managed in the past half century.