Association of Titin-Truncating Genetic Variants With Life-threatening Cardiac Arrhythmias in Patients With Dilated Cardiomyopathy and Implanted Defibrillators

This cohort study investigates the association of titin-truncating genetic variants with life-threatening ventricular arrhythmias in adult patients with dilated cardiomyopathy who have implanted cardiac defibrillators or cardiac resynchronization therapy defibrillators.


Biobank recruitment
There were three routes to recruitment into the NIHR Royal Brompton Cardiovascular Biobank -patients were referred to the Royal Brompton Hospital or Harefield Hospital either for device (ICD/CRT-D) implantation, for cardiovascular magnetic resonance (CMR) imaging, or for review at the dedicated cardiomyopathy or heart failure clinics. The Biobank database includes demographic and clinical information collected via patient interview and clinical records, including the results of diagnostic tests. For the current study, subsequent clinical and device follow-up was then at Royal Brompton Hospital, Harefield Hospital or one of the following 12 centres: Basildon University Hospital, Bristol Heart Institute, Ealing Hospital, Kings College Hospital, Maidstone Hospital, Northwick Park Hospital, Royal Berkshire Hospital, Royal United Hospital Bath, St Helier University Hospital, Tunbridge Wells Hospital, Watford General Hospital and Wexham Park Hospital.

Control group selection
Before commencing data collection, we calculated that a population of 30 TTNtv +ve patients and 90 TTNtv-ve patients would give 80% power to detect a hazard ratio of ≥ 3.7 at the 5% level, assuming a probability of ICD therapy of 20% over the study period 1 . Power calculations showed that increasing the number of control subjects beyond ~3x the number of TTNtv+ve subjects led to only very minimal increases in power. For example, increasing the number of controls to 210 (7 x the number of TTNtv +ve patients) would increase our power to detect a hazard ratio ≥ 3.7 by only 5% (to 85%) 1 , or reduce the hazard ratio detectable with 80% power from 3.7 to 3.4. Therefore, to balance statistical power with cost and time of collection and interpretation of electrogram data, we selected 111 control subjects (3 x the number of TTNtv +ve patients) from the Biobank, selected at random using a random number generator applied to a list of anonymised patient identifiers.

Formula for power calculations 1
Power (1-Type II error) was determined using: where = (ln( ) − ln( 0 ))√ and  is the hazard ratio pE is the overall probability of the event occurring within the study period (taken as 20% for our calculations, based on previous studies [2][3][4] pA and pB are the proportions of the sample size allocated to each group (e.g. in our example with 30 TTNtv+ve and 90 TTNtv-ve patients the proportions would be 0.25 and 0.75) n = the total sample size  is the standard Normal distribution function  -1 is the standard Normal quantile function  is Type I error

Genetic analysis
All potential participants underwent targeted next generation sequencing using the Illumina TruSight Cardio Sequencing kit or a custom Agilent SureSelect XT target capture with equivalent content, followed by sequencing on Illumina platforms or Life Technologies 5500XL. DNA libraries were prepared and sequenced according to manufacturers' protocols. Sequencing reads were aligned with BWA, and variants identified with GATK and annotated using the Ensembl Variant Effect Predictor as previously described 5 . TTNtv adjudicated likely pathogenic (as previously described 5 -minor allele frequency in the Exome Aggregation Consortium (ExAC) dataset <0.001 and impacting all principal cardiac transcripts) were confirmed by Sanger sequencing, or by review of mapped reads in Integrative Genomics Viewer.
In addition, variants in 39 other genes associated with inherited cardiac conditions 6

CMR -Late gadolinium enhancement
For patients who underwent CMR, LGE images were acquired using a breath-hold inversion recovery sequence following administration of gadolinium contrast agent with inversion times optimised to null normal myocardium. Mid-wall myocardial fibrosis was recorded as present if detected in the primary and phase swapped image with cross cuts taken as appropriate.
LGE was assessed by a CMR cardiologist, blinded to genotype.

Censoring
If a patient did not experience an event (n=90 for the primary outcome measure), data were right-censored on the day of their most recent device interrogation. In 91% of censored cases (13/15 TTNtv +ve, 69/75 TTNtvve), censoring was due to patients coming to the natural end of the study. As patients were followed-up from the time of their device implantation until their most recent device interrogation (up to 9 years post-implant), patients contributed differing amounts of time to the study. E.g. a patient having an ICD fitted in Jan 2016, whose most recent device interrogation was in Jan 2017 would be censored at 1 year, whereas a patient whose device was inserted in Jan 2012 whose most recent interrogation was on the same day would be censored at 5 years. All such patients were still alive and being actively followed-up clinically and by the Biobank research teami.e .had not been lost to follow-up.

Genetic Analysis
Details of the pathogenic/likely pathogenic TTNtv in the cohort are given in eTable 3.
Thirty-nine other core interpretable inherited cardiac disease genes were also asessed 16 , and two likelypathogenic/pathogenic variants were identifiedan inframe deletion in TNNT2 (found in a patient also carrying a TTNtv) and a predicted truncating variant in DSP (found in a patient with no other likely pathogenic variant) (eTable 4).

Analysis of recurrent events
The mean number of episodes of appropriate ICD therapy was 0.37 per 5 years follow-up for the TTNtv-ve group, versus 1.03 per 5 years follow-up in the TTNtv+ve group. The association of TTNtv with the total number of ICD events was assessed in a frailty model and was statistically significant (HR = 2.92, 95% CI 1.02-8.3, P = 0.04).

Patients with and without LGE on CMR
If the primary analysis is restricted to those patients with LGE on CMR, the HR for the association of TTNtv and ICD therapy is 9.0 (95% CI 2.3 -35), P = 0.002. For those with no LGE on CMR the HR is 6.3 but does not reach statistical significance (95% CI 0.9 -46, P = 0.07).

TTNtv variants in the A band
Of the TTNtv variants, 19/28 (68%) were in the A-band (see eTable 3). Patients with a TTNtv in the A-band were not more likely to receive an ICD therapy than those with variants in other regions (HR=1.3, 95% CI 0.23-2.6, P=0.67).

Tipping Point Analysisfindings are robust to extreme results in missing ICD data
There were 9 TTNtv +ve and 22 TTNtv -ve patients with missing ICD data due to missing or destroyed records. A tipping point analysis was conducted to assess how extreme the results would have to be amongst the missing data to tip the results into non-significance 9 . Online-only Figure 4 shows the results of this analysis. In summary, of the TTNtv +ve patients with ICD data, 13/28 (46%) experienced one or more appropriate ICD therapies, compared to 14/89 (16%) of TTNtv -ve patients (Fisher's exact test: OR = 4.6, P = 0.0016). If zero of the TTNtv +ve patients with missing ICD data had experienced an event (an extreme result given the fact that 46% of those with ICD data did have an event) then 6/22 (27%) of the TTNtv -ve patients with missing data would have to have experienced an event before the result tipped into non-significance -1.7x the number that would be predicted to occur based on the proportion of events observed in the non-missing data. If 46% of the TTNtv +ve patients with missing data experienced an event (the number predicted based on the observed data) then 16/22 (73%) of the TTNtv -ve patients with missing data would have to experience an event -4.6x the number expected.