Myocardial Strain Measured by Cardiac Magnetic Resonance Predicts Cardiovascular Morbidity and Death

Background Myocardial strain using cardiac magnetic resonance (CMR) is a sensitive marker for predicting adverse outcomes in many cardiac disease states, but the prognostic value in the general population has not been studied conclusively. Objectives The goal of this study was to assess the independent prognostic value of CMR feature tracking (FT)—derived LV global longitudinal (GLS), circumferential (GCS), and radial strain (GRS) metrics in predicting adverse outcomes (heart failure, myocardial infarction, stroke, and death). Methods Participants from the UK Biobank population imaging study were included. Univariable and multivariable Cox models were used for each outcome and each strain marker (GLS, GCS, GRS) separately. The multivariable models were tested with adjustment for prognostically important clinical features and conventional global LV imaging markers relevant for each outcome. Results Overall, 45,700 participants were included in the study (average age 65 ± 8 years), with a median follow-up period of 3 years. All univariable and multivariable models demonstrated that lower absolute GLS, GCS, and GRS were associated with increased incidence of heart failure, myocardial infarction, stroke, and death. All strain markers were independent predictors (incrementally above some respective conventional LV imaging markers) for the morbidity outcomes, but only GLS predicted death independently: (HR: 1.18; 95% CI: 1.07-1.30). Conclusions In the general population, LV strain metrics derived using CMR-FT in radial, circumferential, and longitudinal directions are strongly and independently predictive of heart failure, myocardial infarction, and stroke, but only GLS is independently predictive of death in an adult population cohort.

V entricular myocardial strain is presented as a percentage change in length when tracking the deformation of myocardium from the original length, which is commonly defined as the end-diastolic phase. 1 It is traditionally measured with 3 different markers that reflect the contraction of the left ventricle (LV) along the longitudinal, radial, and circumferential axes. 2 There is considerable interest in these markers, inasmuch as they have been shown to be more sensitive markers of cardiac dysfunction than LV ejection fraction (LVEF), which is widely used in the clinical setting, [3][4][5] and multiple studies have demonstrated incremental diagnostic and prognostic value of strain over LVEF alone. 6,7e use of echocardiography-derived myocardial strain is well established in the clinical setting for detecting, monitoring, and planning interventions in various cardiac disease states. 8,90][11][12][13][14] However, very few studies using echocardiography or CMR have demonstrated the prognostic value of myocardial strain in the general population.A systematic review has identified 8 echocardiography studies that have demonstrated prediction of adverse outcomes in the general population using speckle-tracking LV strain; most focus on global longitudinal strain (GLS), with only 1 investigating global circumferential strain (GCS) and global radial strain (GRS). 15,16R feature tracking (CMR-FT), also known as tissue tracking, is now widely used in clinical and research settings.FT uses balanced steady-state free precession short-axis (SAX) and long-axis (LAX) cine images and applies postprocessing "trackers" to features identified within the defined myocardium (within the epicardial and endocardial contours).9,17 Various software algorithms are available that then track the features through the cardiac phases, usually starting with LV end-diastole as the reference phase.
Studies in cohorts with specific cardiac diseases have shown the prognostic value of CMR-FT-derived myocardial strain.One study has demonstrated that CMR-FT adds incremental prognostic value above the use of clinical features, LVEF, and presence of late gadolinium enhancement (LGE) in patients with myocarditis. 7Another study conducted in >3,000 participants in the MESA (Multi-Ethnic Study of Atherosclerosis) has shown the LV GCS is a sensitive marker to differentiate between individuals with cardiovascular risk factors (obesity, smoking, hypertension, diabetes, and hypercholesterolemia) and those without. 18other MESA study demonstrated the incremental prognostic value of CMR-derived GCS in predicting incident HF and cardiovascular events. 19Only 1 study using CMR tagging (another method of calculating strain) in an unselected cohort demonstrated the incremental prognostic value (in addition to LVEF and LGE) of LV GCS in predicting a composite cardiac endpoint. 6In summary, no prior studies have investigated the prognostic value of all 3 CMR-derived myocardial strain metrics such as GLS, GCS, and GRS (by any method) in a general population for individual cardiovascular endpoints and death.
This study aimed to investigate the prognostic value of CMR-FT-derived LV GLS, LV GCS, and LV GRS in the general population.
We addressed the existing knowledge gap in the literature by assessing the predictive value of these strain metrics for heart failure (HF), myocardial infarction (MI), stroke, and death.Finally, we evaluated the incremental prognostic value of the strain markers over and above existing imaging markers known to be highly predictive of the respective outcomes being assessed.Image analysis to derive volumetric data was performed using convolutional networks with a quality control process that has been described elsewhere. 26e left atrial volumes were derived using cvi42
The quality control of the atrial volumes included removing statistical outliers (3 Â IQR).
Given the variety of automated image analysis pipelines and subsequent quality control used to generate the imaging markers of interest, there is variability in the availability of the necessary data.
STATISTICAL ANALYSIS.Analysis was systematically completed for each outcome (HF, MI, stroke, and death) and for each strain marker (LV GLS, LV GCS,

LV GRS).
A univariable analysis was first carried out to demonstrate whether there was a significant association between each outcome and each strain marker, which were divided by quartiles.The association between each strain marker and each of the endpoints was then assessed using multivariable Cox regression models.The first models tested for all the strain markers and for all the endpoints included clinical covariates of interest, which were age, sex, ethnicity, body mass index (BMI), smoking and alcohol use, diabetes mellitus, prevalent hypertension, hypercholesterolemia, and prevalent coronary disease (the latter was excluded in models for MI as the endpoint).
The subsequent models included the clinical covariates, the strain marker of interest, and additional imaging marker(s) that have been shown to provide significant prognostic or predictive value for the respective endpoint being tested.Left ventricular global function index (LVGFI), LVEDV, and LVEF were chosen as the imaging markers against which to test the incremental value of strain metrics for HF outcome.They had all been shown to have either significant predictive value (LVGFI) 27,28 or wide clinical use (LVEDV and LVEF) 3,29 for HF.Left ventricular end-systolic volume (LVESV), which had been shown to have significant prognostic value, 30 LVEDV, and LVEF were chosen for the MI outcome.LVEDV and LVEF, despite their limitations, are widely used for predicting cardiovascular outcomes and are readily available markers. 3,292][33] Left atrial (LA) maximal volume, shown to have significant predictive value for stroke, 34,35 was chosen for the stroke outcome.Finally, LV mass, which had been shown to be significantly predictive of allcause death, 36,37 was chosen as the convention imaging marker to test against the death outcome.
Incremental value of strain markers was tested using concordance statistics (C-statistics).All models used a complete-case only approach (see Supplemental Table 2 for sample size and number of values excluded for each model) and were tested for violation of the proportional hazards assumption.Multiple testing correction was applied to the multivariable results using the Bonferroni method (0.05/30 ¼ 0.002).

RESULTS
BASELINE CHARACTERISTICS.The baseline characteristics of the study population are presented in Table 1.The UK Biobank field IDs used to define these characteristics can be found in Supplemental Table 3.
The average age of the population was 55 years, with an equal distribution of men and women.The majority of our study cohort were White and nonsmokers.The average values of CMR parameters were within normal limits for the study population. 39,40e average LV GLS and GCS were lower (less negative) than the published normal average values for CMR-FT studies but not below the published cutoff value of À14%.

No. of Events P Value
Clinical features included in all models were age, ethnicity, sex, smoking and alcohol status, body mass index, diabetes status, prevalent coronary disease, hypertension, and hypercholesterolemia.There was no violation of the proportional hazard assumption in all models.*Indicates model significant after multiple testing correction using the Bonferroni method.LV GCS ¼ left ventricular global circumferential strain (%); LV GLS ¼ left ventricular global longitudinal strain (%); LV GRS ¼ left ventricular global radial strain (%); LVEDV ¼ left ventricular end-diastolic volume (mL), LVEF ¼ left ventricular ejection fraction (%); LVGFI ¼ left ventricular global function index.

Prognostic Value of CMR-Derived Myocardial Strain
A U G U S T 1 3 , 2 0 2 4 : 6 4 8 -6 5 9 model with LV GLS and clinical features was significant after applying multiple testing correction, and the model C-statistic analysis did not demonstrate incremental value (Supplemental Table 7).
Only GLS models remained significant after applying multiple testing correction, and none provided evidence of incremental value (Supplemental Table 8).

DISCUSSION
This unique study investigated the predictive and prognostic value of CMR-FT-generated strain measurements in a large general population.The prevalence of diabetes, coronary artery disease, and other cardiovascular risk factors in the study population was comparable with that in the general population (Supplemental Table 9). 42,43However, the proportion of smokers 43 and participants from other ethnic minority backgrounds was lower in the study population than in the general population, 44 with the average BMI lower than the average for this age group in the general population. 45

No. of Events P Value
Clinical features included in all models were age, ethnicity, sex, smoking and alcohol status, body mass index, diabetes status, hypertension, and hypercholesterolemia.
There was no violation of the proportional hazard assumption in all models.*Indicates model significant after multiple testing correction using the Bonferroni method.
LVESV ¼ left ventricular end-systolic volume; other abbreviations as in Figure 1.Furthermore, we showed that the strain markers provided independent predictive value for adverse outcomes even after accounting for demographics, cardiovascular risk factors, and other established prognostic imaging measurements.Uniquely, our study demonstrated these results in 3 separate components of myocardial strain, not just LV GLS, which is the most studied strain parameter in the literature.
Most of the association results remained significant (especially for HF and MI) even after applying the strictest multiple testing correction using the Bonferroni method.
Our findings demonstrated that CMR myocardial strain using feature tracking is highly sensitive to cardiac dysfunction and superior to current commonly used imaging markers in a communitybased low-risk population.LV GLS and LV GRS predicted all-cause death in a multivariable model adjusted for clinical covariates.LV GLS independently predicted death after adjusting for LV mass.
It also showed an incremental value over clinical covariates that are established predictors of death.
The LV GCS and GRS results were not as promising, with either borderline or nonsignificant results for all-cause mortality.This finding is expected, because LV GCS and GRS are perhaps more specific to cardiac-specific adverse outcomes reflecting dysfunction in the mid and epicardial layers. 46In studies using speckle tracking echocardiography, LV GLS has been shown to predict death in the general population. 15,47One study has demonstrated the incremental benefit of CMR-FT-generated LV GLS over LVEF and LGE in predicting death in those with dilatated cardiomyopathy. 48Our study showed an added prognostic value of CMR-generated LV GLS independently of LV mass for all-cause mortality in a general population.
In clinical practice and in research, echocardiography techniques using Doppler imaging or speckle tracking are leading the way in evaluating myocardial strain.These are associated with lower costs and greater accessibility compared with CMR imaging.
However, a significant limitation is that the reliability of the data is heavily influenced by the acquisition of good-quality images, which can be patient-and operator dependent.CMR images are not impeded by the limitation of good acoustic windows and have been shown to have superior Clinical features included in all models were age, ethnicity, sex, smoking and alcohol status, body mass index, diabetes status, prevalent coronary disease, hypertension, and hypercholesterolemia.There was no violation of the proportional hazard assumption in all models.*Indicates model significant after multiple testing correction using the Bonferroni method.Max LA ¼ maximum left atrial volume (mL); other abbreviations as in Figure 1.
reproducibility with less operator dependence compared with echocardiography. 49UDY STRENGTHS AND LIMITATIONS.This study addresses a gap in current knowledge by investigating and demonstrating the prognostic value of CMR-FT myocardial strain measurements in a large general population.One key advantage of the feature tracking method is its applicability in a standard CMR study without the need for additional dedicated image sequences.Given that feature tracking can be applied in noncontrast medium cine images, important and prognostic information can still be inferred in those patients with contraindications to the use of contrast material.
Specific benefits are offered by the software used in this study, which was developed by Circle Cardiovascular Imaging, Inc (cvi42).It uses automated algorithms to define end systole and end diastole, and also to define the endocardial and epicardial borders, which reduces any interoperator variability, which was previously quoted to be a limitation of feature tracking software. 50Given that cvi42 is widely used in clinical settings, integration of fully automated strain module can be relatively straightforward.
STUDY POPULATION AND DESIGN.The UK Biobank recruited participants aged 40 to 69 years between 2006 and 2010, with data collected on approximately 500,000 volunteers.Data collected include detailed sociodemographic and lifestyle information, physical measures, and blood samples.20 Longitudinal data on health outcomes was recorded using linkage to Hospital Episode Statistics and death registers.Further details on the UK Biobank protocol are publicly available.21Additionally, approximately 50,000 CMR scans were performed (2015-2020) as part of the UK Biobank imaging study, which aims to perform imaging on #100,000 of the recruited participants.22There were 45,893 scans available at the time of image analysis; however, 193 scans did not have any corresponding data output because of missing image sequences or corrupted files.Therefore, the study population consisted of the 45,700 participants whose scans were analyzable to derive strain data.The endpoints of incident HF, incident MI, incident stroke, and death were defined using the Hospital Episodes Statistics data as detailed in SEE PAGE 660 A B B R E V I A T I O N S A N D A C R O N Y M S CMR = cardiac magnetic resonance FT = feature tracking HF = heart failure LA max = left atrium maximum volume LAX = long axis LV GCS = left ventricular global circumferential strain LV GLS = left ventricular global longitudinal strain LV GRS = left ventricular global radial strain LVEDV = left ventricular end-diastolic volume LVEF = left ventricular ejection fraction LVESV = left ventricular end-systolic volume LVGFI = left ventricular global function index LVM = left ventricular mass MI = myocardial infarction SAX = short axis J A C C V O L .8 4 , N O .7 , 2 0 2 4 All 3 CMR myocardial strain markers (LV GLS, LV GCS, LV GRS), demonstrated predictive value for HF, MI, stroke, and death in univariable analyses, as summarized in the Central Illustration.These strain markers also provided independent predictive value when adjusted for clinical covariates for heart failure, MI, and stroke.These results are comparable with those of other population studies that have assessed the prognostic value of myocardial strain.As an important novel finding, our study also demonstrated incremental prognostic values of GLS and GCS over well-established imaging markers for HF and MI.LVGFI, LVEDV, and LVEF

FIGURE 2
FIGURE 2 Global Myocardial Strain Predicts Myocardial Infarction

Prognostic
Value of CMR-Derived Myocardial Strain have been shown to be very good indicators of cardiac dysfunction resulting in HF, but the results have demonstrated that GLS and GCS can provide additional prognostic information.Similarly, LVESV, LVEDV, and LVEF have been shown to be very good indicators of poor prognosis associated with MI, but the results demonstrate the GLS and GCS can provide additional prognostic information.

Prognostic
Value of CMR-Derived Myocardial Strain

TABLE 1
Participant Characteristics (N ¼ 45,700) Values are mean AE SD, n (%), or median (Q1-Q3), unless otherwise indicated.BMI ¼ body mass index; CMR ¼ cardiac magnetic resonance.PROGNOSTIC VALUE OF LV STRAIN MARKERS INPREDICTING MI.All 3 strain markers were predictive of MI in the univariable analysis (Supplemental Figures4 to 6).In fully adjusted multivariable Cox regression models, all 3 strain markers provided independent prognostic value (HR: 1.30; 95% CI: 1.17-1.45;HR:1.25;95%CI:1.11-1.40;andHR:0.8095%CI:0.71-0.91 for LV GLS, GCS, and GRS, respectively), as well as in models with additional imaging markers (Figure2).Additionally, GLS provided incremental prognostic value in a model with LVESV, and both 1.09-1.41;HR:1.2205%CI: 1.06-1.41;HR:0.81;95%CI: 0.70-0.94forLVGLS,GCS, and GRS, respectively).The global strain markers also provided independent value even after the inclusion of LA maximum volume in the model (HR: 1.23; 95% CI: 1.06-1.43;HR:1.21;95% CI: 1.04-1.42;HR:0.85;95% CI: 0.72-0.99 for LV GLS, GCS, and GRS respectively) (Figure3).Only 1 Clinical features included in all models were age, ethnicity, sex, smoking and alcohol status, body mass index, diabetes status, prevalent coronary disease, hypertension, and hypercholesterolemia.There was no violation of the proportional hazard assumption in all models.*Indicatesmodelsignificant after multiple testing correction using the Bonferroni method.LV mass ¼ left ventricular mass (g); other abbreviations as in Figure1.Queen Mary University of London, St George's University Hospitals NHS Foundation Trust and St George's University of London.Barts Charity (G-002346) contributed to fees required to access UK Biobank data [access application #2964].This paper is supported by the London Medical Imaging and Artificial Intelligence Centre for Value Based Healthcare (AI4VBH), which is funded from the Data to Early Diagnosis and Precision Medicine strand of the government's Industrial Strategy Challenge Fund, managed and delivered by Innovate UK on behalf of UK Research and Innovation (UKRI).Views expressed are those of the authors and not necessarily those of the AI4VBH Consortium members, the NHS, Innovate UK, or UKRI.The funders did not have any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.Dr Petersen acknowledges the British Heart Foundation for funding the manual analysis to create a cardiovascular magnetic resonance imaging reference standard for the UK Biobank imaging resource in 5,000 CMR scans (PG/14/89/31194).Dr Aung acknowledges the Medical Research Council for supporting his Clinician Scientist Fellowship (MR/ X020924/1).Dr Chadalavada was funded by European Union's Horizon 2020 research and innovation program under grant agreement no.825903 (euCanSHare project).Dr Rauseo is supported by the mini-Centre for Doctoral Training (CDT) award through the Faculty of Science and Engineering, Queen Mary University of London, United Kingdom.Dr Naderi was supported by the British Heart Foundation Pat Merriman Clinical Research Training Fellowship (FS/20/22/ 34640).Dr Petersen and Dr Lee acknowledge support from the SmartHeart EPSRC program grant (EP/P001009/1) and the European Union's Horizon 2020 research and innovation program under grant agreement No 825903 (euCanSHare project).Dr Petersen has served as a consultant for Cardiovascular Imaging Inc, Calgary, Alberta, Canada.All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
This work uses data provided by patients and collected by the NHS as part of their care and support.This research used data assets made available by National Safe Haven as part of the Data and Connectivity National Core Study, led by Health Data Research UK in partnership with the Office for National Statistics and funded by UK Research and Innovation (research which commenced between TRANSLATIONAL OUTLOOK: Further research is needed to define and standardize normal ranges and disease-specific and risk factor-specific thresholds for CMR-derived myocardial train to facilitate broad clinical use.