Author + information
- Received March 23, 2015
- Revision received July 20, 2015
- Accepted August 13, 2015
- Published online February 1, 2016.
- James H.P. Gamble, BMBCh∗ (, )
- Neil Herring, DPhil,
- Matthew Ginks, MD,
- Kim Rajappan, MD,
- Yaver Bashir, DM and
- Timothy R. Betts, MD
- Oxford Heart Centre, John Radcliffe Hospital, Oxford University Hospitals NHS Trust, Oxford, United Kingdom
- ↵∗Reprint requests and correspondence:
Dr. James H.P. Gamble, Oxford Heart Centre, John Radcliffe Hospital, Oxford University Hospitals NHS Trust, Oxford OX3 9DU, United Kingdom.
Objectives The goal of this study was to assess the contemporary and historical success rates of transvenous left ventricular (LV) lead placement for cardiac resynchronization therapy (CRT), their change over time, and the reasons for failure.
Background In selected patients, CRT improves morbidity and mortality, but the placement of the LV lead can be technically challenging.
Methods A literature search was used to identify all studies reporting success rates of LV lead placement for CRT via the coronary sinus (CS) route. A total of 164 studies were identified, and a meta-analysis was performed.
Results The studies included 29,503 patients: 74% (95% confidence interval [CI]: 72% to 76%) were male; their mean age was 66 years (95% CI: 65 to 67); their mean New York Heart Association functional class was 2.8 (95% CI: 2.7 to 2.9); the mean LV ejection fraction was 26% (95% CI: 25% to 28%); and the mean QRS duration was 155 ms (95% CI: 150 to 160). The overall rate of failure of implantation of an LV lead was 3.6% (95% CI: 3.1 to 4.3). The rate of failure in studies commencing before 2005 was 5.4% (95% CI: 4.4% to 6.5%), and from 2005 onward it was 2.4% (95% CI: 1.9% to 3.1%; p < 0.001). Causes of failure (reported for 39% of failures) also changed over time. Failure to cannulate and navigate the CS decreased from 53% to 30% (p = 0.01), and the absence of any suitable, acceptable vein increased from 39% to 64% (p = 0.007). The proportion of leads in a lateral or posterolateral final position (reported for 26% of leads) increased from 66% to 82% (p = 0.004).
Conclusions The reported rate of failure to place an LV lead via the CS has decreased steadily over time. A greater proportion of failures in recent studies are due to coronary venous anatomy that is unsuitable for this technique.
Cardiac resynchronization therapy (CRT) has been shown by several large studies to improve symptoms and mortality in a suitably selected population of patients with systolic heart failure and prolonged QRS duration according to electrocardiogram data (1–3). CRT is most commonly achieved by transvenous placement of a left ventricular (LV) lead into a tributary vein of the coronary sinus (CS); this procedure can be technically demanding and may fail due to a variety of procedural or anatomic challenges (4,5).
The present meta-analysis of the published data was performed to assess the success rates, causes of failure, and procedural characteristics of CRT implantation both during the evolution of this procedure and in the modern era.
The meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (6). A systematic search of PubMed, Web of Science, and the Cochrane Library was performed to identify relevant articles published until the end of 2014. The key words “cardiac resynchronization therapy,” “cardiac device,” and “CRT” were used. Hand-searching of the reference lists of included publications allowed identification of articles not found in the primary search strategy. We also searched relevant U.S. Federal Drug Administration device pre-market approval summary reports.
All publications reporting the results of attempted implantation of LV leads for CRT via the transvenous CS route were included, whether within a formal trial design or as part of a case series or similar report. For those studies comparing conventional CRT with a different procedure or control, the conventional CRT arm was included. Publications were excluded that only reported outcomes after the implantation procedure; that did not make it clear if any procedures had failed; or that were entirely or largely duplicates of previously reported data. The study was deliberately inclusive to include the maximum number of patients to assess procedural success rates.
The titles of all publications identified were screened, and those that did not meet the study criteria were excluded. We subsequently reviewed the abstracts of remaining publications and, if necessary, the publication text. Methodological quality was not assessed, but publications were dichotomized into randomized controlled trials (RCT) and others.
The number of patients in whom the implantation procedure failed was recorded. Reasons for failure to place a LV lead were recorded as absolute numbers in one of a series of pre-defined categories. In addition, reported population parameters from the trials were recorded, as absolute numbers when provided or mean and SD. LV lead revision, when reported, was assessed as a secondary data point. The absolute numbers of revised LV leads and the study follow-up times were also recorded.
Given the large number of trials included over a long time period, significant heterogeneity was expected, and a random effects meta-analytical approach was therefore applied to all analyses. The logarithm ratio of the rate of failure was analyzed, and results are presented back-transformed as the median percentage of implantation failures. For the investigation of temporal trends, studies were considered according to the starting year of recruitment; if unreported, this was estimated from provided data. This approach was used to dichotomize included studies, using the midpoint of the total time period of all studies as the dividing point, and to assess temporal changes and trends by using meta-regression. Subgroups were compared by using mixed effects logistic meta-regression, allowing for differential heterogeneity between subgroups. Meta-analysis of event count data with relatively rare events, such as these data, is believed to be best performed with a binomial normal model; this method provides appropriate weight to studies with low event numbers (7).
A random effects meta-analysis was also used to estimate the mean of population parameters, using the logarithm transform to estimate population proportions and treating continuous variables in the standard meta-analytical manner; subgroups were similarly compared by using meta-regression. We analyzed lead revision as a mixed effects logistic regression meta-analysis of the incidence rate of revision. Several different LV lead classification schemes have been used to describe the anatomic location of the final LV lead position, including a broad classification as anterior, lateral, or posterior; according to the vein or vein territory used (8); or according to segment as anterior, lateral, posterolateral, or posterior. We recorded the absolute numbers of leads in the positions described, as well as the proportion of leads in the lateral-posterolateral region generally considered to be optimal (8), when this information could be estimated from the provided data. Proportions were estimated by using methods similar to the aforementioned meta-analytical techniques.
Statistical analysis was performed using R version 3.1 and the Metafor package, with use of the ggplot2 package for graphs (9,10). A significance level of 0.05 was used, and all testing was 2-tailed. All confidence intervals (CIs) quoted are at the 95% level.
Figure 1 summarizes the selection of the 164 included studies, which commenced recruitment between 1997 and 2012 and were published between 2000 and 2014. Seven additional studies were found for which the only available report was in U.S. Federal Drug Administration documents. A full list of included studies is detailed in the Online Appendix. Included studies involved 29,503 patients in whom implantation of a LV lead for CRT was attempted.
Characteristics of included patients
Summary characteristics of all included patients are detailed in Table 1. Characteristics are also further separated into early studies commencing recruitment before the midpoint of 2004 and later studies commencing recruitment from January 2005 onward. The start and end dates of recruitment were estimated for 20% of studies.
Reporting of baseline patient details was inconsistent between studies, although some data were reported for 99% of patients. Several studies did not include within their reported number those patients in whom implantation of a LV lead failed, and in other studies, data were reported for those who were excluded before implantation. Table 1 presents the percentage of patients included in the studies in which each data point was reported, which was >80% for all parameters. There was significant heterogeneity in all parameters, with I2 values (a measure of study heterogeneity) >95% for all parameters except the sex distribution (80%). There was a significant decrease in mean QRS duration between the early and late time periods, although if the 9 studies specifically including patients with a narrow QRS were excluded, this difference was reduced (after exclusion, early 163 ms [95% confidence interval [CI]: 159 to 167]; late 156 ms [95% CI: 152 to 160]; p = 0.003).
Occurrence of failure to place an LV lead
The proportion of patients in whom it was not possible to place a LV lead is shown in Table 2, and the data are presented as a forest plot in Figure 2. The proportion of failed LV lead placements decreased steadily over the time period by 0.48% per year (95% confidence interval [CI]: –0.28% to –0.69%; p < 0.001). Meta-regression of each half of the time period separately showed a trend toward a significant decrease over the early period (0.51% per year; 95% CI: –1.1% to 0.09%; p = 0.095) but no significant change over the late period (0.032% per year; 95% CI: –0.03% to 0.04%; p = 0.86). Figure 3 presents a graphical representation of failure rates over time.
As expected, and as is shown by the wide distribution of dots representing studies in Figure 3, there was significant heterogeneity between studies in this analysis. I2 values were high at 79% across the whole time period, although this value decreased from 89% in the early period to 69% in the later period. There was a strong trend toward a decrease in I2 across the time period.
Thirty-three percent of included studies were RCTs, which were larger than nonrandomized studies (median patients 175 vs. 89.5) and more likely to be multicenter (median centers 19 vs. 1) and included 46% of patients. Failure rates in RCTs are also shown in Figure 2. A more detailed comparison between RCTs and other studies is given in Online Table 1.
Funnel plots of the rate of failure to place a lead suggest significant underreporting of events in small studies (Online Figure 1). A modified Egger’s test also suggested significant underreporting (p < 0.001) (11). The trim-and-fill analysis, a method of estimating the likely result in the presence of additional hypothetical studies, suggests that 30 additional studies with more events would have been expected to be reported, which would have increased the overall failure rate estimate to 5% (95% CI: 4.3% to 5.9%).
Reasons for failure to place an LV lead
The reasons for the inability to place an LV lead were reported in 44 (27%) studies and for 613 patients (39% of LV lead placement failures) (Figure 4). Inability to place an LV lead due to difficulties with CS access and navigation decreased significantly from 53% (95% CI: 43% to 63%) to 30% (95% CI: 18% to 46%; p = 0.01) of failures, and failure due to a lack of an anatomically suitable site (no suitable vessel, no acceptable threshold, lead instability, or phrenic nerve stimulation) increased significantly from 39% (95% CI: 29% to 49%) to 64% (95% CI: 49% to 77%; p < 0.001).
Nineteen (12%) trials, including 6,455 (22%) patients, reported the failure rate at each attempt at LV lead implantation. The failure rate for the first attempt was 7.2% (95% CI: 5.8% to 8.9%). The results of a second attempt were recorded for 187 (34%) of these attempts, which failed in 41% of patients.
LV lead position
The final transverse plane location of the LV lead was reported for 29% of studies and for 7,695 leads (26% of successful procedures). Forty-six percent of lead positions were reported using the vein territory (summarized in Figure 5), 23% used the anterior/posterior/lateral classification, 13% used the segment, and 19% reported the position only for some leads.
Overall, 68% (95% CI: 56% to 79%) of leads were reported as being lateral or posterolateral. In the early period, 63% (95% CI: 48% to 76%) of leads were placed laterally or posterolaterally; in the late period, this increased to 81% (95% CI: 74% to 87%; p = 0.01). The proportion of lateral or posterolateral leads was not a significant predictor of success rates according to meta-regression (p = 0.1).
The combination of non-posterolateral leads and failure to place an LV lead also decreased over time, with a more prominent decrease in the second half of the time period, from 38% (95% CI: 28% to 48%) to 21% (95% CI: 15% to 28%; p = 0.004). This outcome is shown graphically in Online Figure 2.
Lead and device design
The only lead used frequently enough to make comparisons was the St. Jude Medical quadripolar lead (1458Q), which was used in 14 nonrandomized studies. These studies included 1,327 patients, and they commenced between 2009 and 2011. The rate of failure to place an LV lead in in the quadripolar lead arms of these studies was 3.8% (95% CI: 2.9% to 4.9%), which was not significantly different from other studies recruiting in the late time period (p = 0.19 by meta-regression).
LV lead revision
A total of 83 studies (including 15,222 patients [52% of the total]) reported LV lead revision rates over a mean follow-up of 8.0 months (range 1 to 24 months), including a total of 10,501 patient-years of follow-up. These were used to derive the rate of LV lead revision per 100 patient-years (Table 3). Figure 6 shows lead revision rates over time. LV lead revision did not change significantly over the whole time period (p = 0.18) or over either half of the time period (p = 0.16).
When lead revision was considered as a simple proportion of patients, studies following up patients for 1 to 3, 3 to 6, and 6 to 12 months reported similar LV lead revision proportions (4.7%, 4.7%, and 4.4%, respectively), but studies with >12 months’ follow-up reported a higher proportion of lead revision (6.3%; p < 0.001 vs. others). However, when considering lead revision rates as a function of study follow-up time, studies with shorter follow-up time had significantly higher rates of displacement per 100 patient-years (p < 0.001). The available data were insufficient to assess differential revision rates of LV lead designs. There was significant heterogeneity in study outcomes across the whole time period, with an I2 of 87%, which did not significantly change between the early and late periods.
We provide data on success rates of LV lead implantation in a very large population of patients, including the majority of patients involved in the landmark RCTs of CRT. The main finding of our analysis is that the rate of failure to place a LV lead via the CS route has fallen consistently since the introduction of the technique, with a reduction in the rate of change in the last few years. The predominant reasons for failure of an LV lead implantation have changed from those related to access and navigation of the CS to there being no suitable target vessel. The need for revision of the LV lead has not apparently changed significantly over the entire time period. These figures provide a robust basis for patient consent and benchmarking of CRT implantation programs.
Failure rates for implantation of an LV lead
The improvement seen in success rate over time would be expected for many procedures. The techniques and equipment available for CRT have improved markedly over the time period included, initially as equipment specifically designed for CRT was developed, and more recently as it has been refined (12–14). There is also undoubtedly an operator learning curve that applies to this procedure, observed both within studies and within centers (5,15–18). All of these effects would be expected to have been more prominent over the earlier years of the technique.
The apparent slowing of the improvement in success rates seen over the second half of the study period might be due to increasing numbers of cases in which the limiting factor becomes patient anatomy, rather than operator skill and equipment. This possibility is supported by the changes in reasons for implantation failure, in which failures caused by difficulty accessing and navigating the CS were a major issue in the early period but less so in the late period. This problem is less likely to be amenable to further technical developments that use a CS approach to implantation. Further lead design improvements, such as an increased variety of lead shapes, profiles, and multiple electrodes, might be expected to result in a further limited reduction in failed implantation procedures, although we did not find any such benefit from quadripolar leads in this study. The MORE-CRT (More Options Available With a Quadripolar LV Lead Provide In-clinic Solutions to CRT Challenges) study (19) did show a benefit of quadripolar leads, driven apparently by easier implants but with a rate of failure to place a LV lead similar to the contemporary studies in our analysis. A publication from our group has shown reduced phrenic nerve stimulation, LV lead revision, and displacement in patients with quadripolar leads, also with no apparent reduction in procedural success (20). A further factor in this lack of decline could be additional operators starting to place LV leads at a lower point in their individual learning curves.
It is notable how LV lead revision rates, after the initial years of CRT, appear to have changed very little, unlike the rate of failure to place a LV lead. The analysis of lead revision rates is limited because this aspect was a secondary outcome, but the results still represent a large dataset.
The infrequent occurrence in contemporary studies of failure to place a LV lead suggests that second-line procedures for surgical or endocardial lead placement will need to be refined and centralized into super-specialist centers for operators to have adequate experience and skills in these techniques (21,22). This suggestion is supported by the extremely high 41% failure rate of repeat attempts at placing LV leads in those patients in whom the first procedure had failed.
LV lead position
Our data suggest that a significant proportion of patients receive an LV lead that is positioned either anteriorly or apically, where it may be less likely to benefit them, and that some successful placements might be at the cost of a poor lead position (23,24). Reporting of lead positions and reasons for failed implantations were inconsistent across studies, and these data were therefore gathered from a subselection of included papers. Lead position in almost all studies was reported by the investigators (rather than centrally adjudicated), which is likely to reduce the accuracy of these data because it has been previously shown that investigator reporting of lead position is variable (25). It is possible that certain types of study are overrepresented in this group, which could bias the data. In particular, studies reporting base to apex lead position were frequently investigating quadripolar leads, which due to their design could be placed at the apex but set up to pace from a more proximal pole located basally. It is also impossible to definitively state that lead position was not optimal in those in whom leads were placed in a non-classical position, as at least some included studies targeted leads based on imaging data to positions optimized on a patient-specific basis (26,27).
Characteristics of patients undergoing CRT implantation
There were only very small shifts in patient characteristics over time, and these were largely of such small magnitude as to be clinically and statistically insignificant. A decrease in QRS duration was seen in the included studies, even after those studies specifically including patients with a narrow QRS duration were excluded, but no clinically relevant change in New York Heart Association functional class or left ventricular ejection fraction was reported.
Overall, it is notable how highly represented men in their sixth and seventh decade are in the CRT study population, and how the research population is younger than the heart failure population commonly seen clinically. Some reported real-world CRT implantation data (28) are similar to those seen in the current meta-analysis, but it is notable that recent U.S. data show that real-world patients are older and include more women (29). We did not attempt to make any links between study population parameters and success rates; these factors were not consistently reported for all patients in whom implantation was attempted, and those in whom placements failed were often excluded from study reporting. Hence, these population parameters can only provide a snapshot of CRT practice within trials.
The conclusions of our analysis are limited by its meta-analytical nature and by the variable completeness of data reporting in the included studies. The significantly lower failure rate in nonrandomized studies could reflect underreporting of failed placements, due to either omission of reporting these or to publication bias. This possibility is supported by our statistical assessment of a likely bias toward underreporting, although an alternative explanation could be that these studies were largely single-center series performed by experienced operators. It is also possible that reporting of the reasons for failed implantation was inconsistent, with operators reluctant to report certain complications such as CS dissection or perforation. These limitations may make the failure rate estimate from the RCT subgroup more reliable. It is notable in the graph of LV lead failure rates over time (Figure 3) that there was a significantly lower failure rate in 2006, despite a reasonable sample size in the year. No clear explanation for this anomaly was found, although it is possible that the aforementioned reasons might apply.
Almost no studies reported the characteristics of patients with failed LV lead placement. It was therefore not possible to assess if any particular features of this population might have predisposed them to experience a failed LV lead placement.
The rate of failure of LV lead implantation for CRT has fallen very significantly since the introduction of this procedure, but the rate of change seems to have slowed or stopped. The reasons for failure to place a lead have changed from technical limitations to those predominantly due to adverse anatomy. Lead revision rates have changed little since the introduction of CRT.
COMPETENCY IN MEDICAL KNOWLEDGE: The rate of failure to place a left ventricular lead for CRT in contemporary trials and series is 2.4%, and the rate of lead revision is 5.7 per 100 patient-years. The rate of failure to place a lead has decreased very significantly from the rates reported in earlier trials of CRT. About two-thirds of contemporary failures to place a left ventricular lead are due to adverse distal venous anatomy, and the remaining one-third are due to challenging CS anatomy.
TRANSLATIONAL OUTLOOK: The incidence of failure to place a LV lead is now so low that second- and third-line surgical or interventional procedures should be centralized into highly specialist centers. Further developments in LV lead design should focus on facilitating lead deployment into challenging distal venous anatomy.
The authors thank Dr. Jacqueline Birks of the Oxford Centre for Statistics in Medicine for invaluable assistance with the statistical analysis.
No funding was received for this study. Dr. Gamble has received research fellowship funding from St. Jude Medical Ltd. Dr. Herring acknowledges support from the British Heart Foundation Centre of Research Excellence (RE/08/004) and is a BHF Intermediate Fellow at the University of Oxford. Dr. Betts has received research funding from St. Jude Medical Ltd. and honoraria for product development and speaker fees from Boston Scientific Ltd. and Medtronic Ltd.; and he acknowledges support from the U.K. National Institute of Health Research Oxford Biomedical Research Centre. Drs. Gamble, Herring, Ginks, Rajappan, and Bashir have received educational support from St. Jude Medical Ltd., Medtronic Ltd., and Boston Scientific Ltd. Drs. Gamble and Betts acknowledge support from Heart Research U.K.
- Abbreviations and Acronyms
- confidence interval
- cardiac resynchronization therapy
- cardiac sinus
- left ventricular
- randomized controlled trial
- Received March 23, 2015.
- Revision received July 20, 2015.
- Accepted August 13, 2015.
- American College of Cardiology Foundation
- Cleland J.G.,
- Abraham W.T.,
- Linde C.,
- et al.
- Van Rees J.B.,
- de Bie M.K.,
- Thijssen J.,
- Borleffs C.J.,
- Schalij M.J.,
- van Erven L.
- León A.R.,
- Abraham W.T.,
- Curtis A.B.,
- et al.
- Moher D.,
- Liberati A.,
- Tetzlaff J.,
- Altman D.G.,
- the PRISMA Group
- Vardas P.E.,
- Auricchio A.,
- Blanc J.J.,
- et al.
- R Core Team
- Giannola G.,
- Iacopino S.,
- Lombardo E.,
- et al.
- Ahsan S.Y.,
- Saberwal B.,
- Lambiase P.D.,
- et al.
- Macías A.,
- García-Bolao I.,
- Díaz-Infante E.,
- et al.
- Gras D.,
- Böcker D.,
- Lunati M.,
- et al.
- Boriani G.,
- Connors S.,
- Kalarus Z.,
- et al.
- Behar J.M.,
- Bostock J.,
- Zhu Li A.P.,
- et al.
- Betts T.R.,
- Gamble J.H.,
- Khiani R.,
- Bashir Y.,
- Rajappan K.
- Bordachar P.,
- Derval N.,
- Ploux S.,
- et al.
- Singh J.P.,
- Klein H.U.,
- Huang D.T.,
- et al.
- Ypenburg C.,
- van Bommel R.J.,
- Delgado V.,
- et al.
- Khan F.Z.,
- Virdee M.S.,
- Palmer C.R.,
- et al.
- Saba S.,
- Marek J.,
- Schwartzman D.,
- et al.
- Dickstein K.,
- Bogale N.,
- Priori S.,
- et al.
- Borne R.T.,
- Peterson P.N.,
- Greenlee R.,
- et al.