3.5. Number of risk factors and the risk of PICM
Receiver operator characteristics (ROC) analysis was done for baseline LVEF and paced QRS duration to determine the cut-offs to predict PICM. Area under curve (AUC) for LVEF was 0.61 (95% CI 0.53 - 0.68, p-value 0.003) and paced QRS duration was 0.71 (95% CI 0.65 - 0.77, p-value < 0.001). The cut-off value with 90.0% specificity to predict PICM was 55.0 % for LVEF and 160msec for QRS duration. For RV pacing burden, a value of 34.0% (cut-off between the lowest tertile and the higher tertiles) was taken. Based on these cut-off values, the risk factors were: (i) baseline LVEF < 55.0 %, (ii) paced QRS duration > 160msec, and (iii) RV pacing burden > 34.0 %. Patients with two or more of the above risk factors were at highest risk (OR 11.62, 95% CI 4.62 - 29.21, p-value < 0.001) of developing PICM compared to those without any of the risk factors (reference group; figure.2). Those with one of the risk factors were at intermediate risk (OR 3.89, 95% CI 1.62 - 9.34, p-value 0.002).
Discussion
The main findings of the current study are : (i) PICM occurred in about 10% of this cohort of south Asian population, (ii) lower pre-implant LVEF, wider paced QRS duration, and higher RV pacing burden were independent predictors of PICM, and (iii) a combination of ≥ 2 of these risk factors increases the odds of developing PICM by 12 times compared to those who do not have any of the factors.
The observed incidence of PICM over long term follow up in our study was slightly lower compared to that from previous studies which ranged from 9% to 19.5% (5–8). This is probably due to heterogeneous definitions and methodological issues. Zhang et al reported a higher incidence of PICM of 26%, but they did not identify and exclude patients with alternative aetiologies of heart failure (13). Possible explanations for the lower incidence of PICM in our study include lesser overall RV pacing (>40% RV pacing was noted only in 65% of the cohort) and relatively short follow up (earlier reports show that risk of incident PICM persists years after exposure, sometimes as late as 15 years (6)). Other contributing factors include exclusion of patients having less than 6 months follow up (PICM can develop within 1 month of implantation (5)) and lesser use of apical pacing.
Kiehl et al’s multivariate analysis showed lower pre-PM LVEF ( HR: 1.047 per 1% LVEF decrease, p=0.42) as a statistically significant factor for the development of PICM (7). In our study too the results were similar (p=0.016). It is however still unclear as to whether each incremental percentage increase in LVEF above 50% is protective until it reaches a particular threshold.
The benefit of biventricular pacing resulting in a narrow QRSd and restoring cardiac synchrony in patients with LBBB and QRSd >150msec has been well established (14). Miyoshi et al in a prospective analysis of 92 patients with permanent RV apical pacing predicted the development of symptomatic heart failure (46.6% vs 11.6%, p<0.05) and worsening LV parameters using a pQRSd cut-off value of >190msec (15). PREDICT-HF was an observational cohort study which showed an incidence of heart failure of 56.8% with pQRSd >190msec compared to 9.4% in patients with pQRSd <160msec (p<0.001) (16). Our study too showed pQRSd to be an independent predictor and cut off values of >125msec had best sensitivity (90%) and >160msec had best specificity (90%) for development of PICM. These findings demonstrate that patients with a higher pQRSd need to undergo active and close follow up with more periodic echocardiography. As a corollary, patients with a low pQRSd can be followed up less rigorously which would be beneficial in a resource poor setting. Sharma et al demonstrated this in a prospective study where those with a narrow pQRSd had preserved LVEF at one year follow up using equilibrium radionucleotide angiography (17). This data further supports the concept that wide pQRSd is a marker of electrical dyssynchrony which leads to adverse remodelling causing PICM in patients with RV pacing. Hence striving to achieve a narrow QRS during the time of implantation could be the only modifiable risk factor that leads to better outcomes with respect to PICM. The advent of His and Left bundle pacing, result in less electromechanical dyssynchrony and are likely to replace the traditional pacing modalities in future (3,4).
RV pacing of more than 40% is the suggested threshold for the development of PICM by the MOST and DAVID trials (9). Though more than 65% of our study cohort had a RV pacing percentage of more than 40%, PICM occurred in some patients who were paced less than this threshold. Kiehl et al showed that RV pacing percentage independently predicted PICM both as a continuous variable (HR: 1.01 per 1% RV pacing, 95% CI: 1.002-1.02, p=0.02) and even more significantly as a categorical variable (<20% or ≥20% RV pacing) (HR: 6.76, 95%CI: 2.08-22.0, p=0.002) (7). Hence it should be noted that even in patients with low RV pacing percentages there is a need to consider the possibility of future occurrence of PICM as noted in our study. The odds of them developing PICM though low will still warrant periodic monitoring. This has significant cost implications for developing countries where the patients have to bear their own medical expenses.
A meta-analysis published in 2012 suggested that patients with RV non apical pacing may have higher LVEF on follow up compared to those with RV apical pacing (18). Randomised controlled trials comparing the two modes of RV pacing showed no difference in levels of brain natriuretic peptide levels, ejection fraction or exercise capacity over a follow up period of 18 months(19,20). The PROTECT-PACE trial which was planned towards identifying the optimal pacing site to reduce PICM showed no benefit of non-apical pacing on LV function over a 2 year follow up period (21). In our study too there was no significant difference between apical and non-apical positions for RV pacing towards the occurrence of PICM (p=0.9). Similar to our findings Chan et al reported that LV volume and systolic function are better predicted by pQRSd and not the pacing site, with pQRSd being a target for pacing site optimisation (22).
The native QRS duration (nQRSd) is predominantly influenced by the level of antegrade delay in the His Purkinje system during atrioventricular block. Khurshid et al in a single centre study in 2016 showed that nQRSd (HR:1.03 per msec: P<0.001) is an independent predictor for PICM (23). Pap et al showed that nQRSd is a good predictor of pQRSd in patients with AV block and normal LVEF undergoing PM implantation (24). Though our study did not find nQRSd as a significant risk factor for PICM; the p value suggests that with larger numbers this may not be true. A wider nQRSd prior to implantation of PM in patients with normal LVEF implies greater electrical dyssynchrony which over time translates to mechanical dyssynchrony. The latter perhaps predisposes them to PICM with the added insult of RV pacing.
Our study for the first time looked at whether the risk factors that predicted PICM added up and found that they did. Prescence of more than one risk factor was incrementally additive. Of the 3 risk factors: pre implant LVEF, RV pacing burden and pQRSd; only the latter was modifiable. Hence at implantation it would be worthwhile investing time in choosing a location where the pQRSd is the lowest possible.
Strengths and Weaknesses:
One of the main strengths of our study is a large sized contemporary cohort of patients from South Asia, who went through a thorough process of recruitment and have comprehensive data for analysis. This is the first study to not only identify the risk factors for PICM but also give an additive risk assessment. This would significantly help the identification of patients prone to developing this condition.
There were a few limitations noted in this study. First, though extensive data search was done to identify and exclude all other potential sources of cardiomyopathy, it is possible that some patients diagnosed with PICM may have had an alternate cause for decrease in LV function. However, this is what is commonly observed in clinical practice as PICM is a diagnosis of exclusion. The other limitation was significant long-term loss of follow up. This may have led to an inclusion bias with symptomatic patients in heart failure being more likely to follow up and hence being included in the study. This, however, is unlikely considering the incidence of PICM in our study is comparable if not lower than that in studies done elsewhere. We evaluated if the risk factors predisposing to PICM added up, but our numbers were not large enough to rule out interactions between them. Some individuals develop PICM despite lacking these suggesting that there are other predisposing factors that are hitherto unknown.
Conclusion:
In patients undergoing permanent PM, low-normal baseline LVEF, wider paced QRS duration and a higher RV pacing burden independently predicted the development of PICM in this cohort from South Asia. Presence of these risk factors was incrementally additive with a patient having 2 or more of these risk factors having twelve-fold increased odds of developing PICM. Striving to reduce the only modifiable risk factor the pQRSd at the time of implant, could help reduce the incidence of PICM.