Key Findings and Strength of Evidence

The key findings of this review for the outcomes identified as being most clinically important are summarized in the evidence summary tables (Tables 31–38) and factors used to determine the overall strength of evidence are summarized in Appendix G.

Due to methodological limitations, only one randomized controlled trial (RCT) was considered to be high quality.⁴⁵

For the comparison of radiofrequency ablation (RFA) with medical therapy, for the primary clinical outcomes (mortality, stroke, myocardial infarction (MI), chronic heart failure (CHF) and health related quality of life (HRQOL), all ratings were low or insufficient. Strength of evidence was low or insufficient for most intermediate outcomes as well. Exceptions were freedom from recurrence of any atrial arrhythmia in patients with paroxysmal atrial fibrillation (AF) in both the short term (≤12 months) and longer term (>12 months) and for the pooled estimate across all studies regardless of AF type for which the strength of evidence was rated as moderate for the comparison of radiofrequency ablation with medical therapy. For many outcomes for a specific AF type, only single studies were available and conclusions were not possible secondary to study limitations (i.e. methodology), small sample sizes resulting in imprecise estimates and/or limited data from these single studies leading to a strength of evidence of insufficient. Overall, findings from observational studies did not alter conclusions or impact the strength of evidence for any of the specified outcomes.

Evidence was most robust for the comparison of RFA with medical therapy. Overall, data were sparse for primary clinical outcomes such as mortality and stroke for both the short term and long term. The differences in results for long versus short term, particularly with respect to freedom from recurrence and reablation, may be due to a variety of factors including fewer studies reporting longer-term outcome and the short-term followup being too short to capture later relapses, thus longer-term benefit appears lower. There were insufficient data to evaluate the extent to which differences in technique or population characteristics may have impacted outcomes.

Given that long-term outcomes (≥12 months) and findings from studies in the Medicare population are of primary importance to Centers for Medicare and Medicaid Services (CMS), these are highlighted. The only information available on populations 65 years of age and older was from two observational studies comparing RFA with medical therapy.^{48, 53}

Longer-Term Efficacy and Effectiveness

Long-term outcomes were only available for the comparison of RFA with medical therapy. Outcomes for which there were longer-term data are summarized in Table 31.

Table 31

Key findings and strength of evidence for Key Question 1: Long-term (>12 months) comparative efficacy and effectiveness of RFA versus medical therapy for AF.

Medicare-Relevant Population

In the Medicare-relevant population, there were no RCTs to provide evidence for efficacy and included studies did not provide analyses of this subpopulation. Definitive conclusions cannot be drawn from the two comparative observational studies identified that reported on patients aged 65 years and older. One fair-quality study with mean of 60 months followup⁴⁸ was in patients with persistent AF and the other was a poor-quality case-control study with a mean of 69 months followup in patients with various AF types,⁵³ although approximately 70 percent were classified by the authors as having “nonparoxysmal” AF. This latter study was in a slightly younger population mean age 67.2 years) compared with the study in those with persistent AF (mean age, 75 years). Across these studies, results were conflicting with regard to the impact of RFA on all-cause mortality, with one in patients with persistent AF reporting comparable mortality between treatment groups (RFA: 1.3%; medical therapy: 1.9%); and the other in the mixed population reporting more than a fourfold greater risk of mortality in the medical treatment group (RFA: 2.1%; medical therapy: 16.5%) that was statistically significant. Similarly, in the study of those with persistent AF, the development of CHF was comparable between groups (RFA: 0%; medical therapy: 1.2%) in contrast to the significantly higher frequency in medical treatment group in the study of the mixed population (RFA: 0.7%; medical therapy: 9.8%). The case-control study did not provide information on the timing of outcomes such as mortality and stroke during the followup period. Neither study reported on HRQOL. It is not clear to what extent population differences, control for confounding, study design and/or technical differences across these studies might have influenced the results. In patients with persistent AF sinus rhythm was maintained in 58 percent of the RFA group versus 43 percent and 18.3 percent of patients required reablation.⁴⁸ Evidence was insufficient for all outcomes due to methodological concerns, imprecision, particularly for rare outcomes and unknown consistency from single studies. The inconsistency of the findings from the two studies and insufficient strength of evidence may make it challenging to weigh overall benefits and risks of RFA compared with medical therapy in this population from an evidence-based perspective.

General Population

In the general population, firm conclusions regarding the effect of RFA on the primary long-term clinical outcomes (e.g., mortality) were limited given the paucity of data available. Three trials of RFA versus medical therapy in patients with paroxysmal AF reported on outcomes >12 months^{35, 46, 83} with longest followup of 48 months in one study. No randomized controlled trials reporting on longer-term outcomes in patients with persistent AF were identified. One comparative observational study which contained those with either paroxysmal or persistent AF reported followup means of 16 months.⁵¹

In patients with paroxysmal AF, the strength of evidence was low that RFA did not appear to affect long-term all-cause mortality in those with paroxysmal AF across two trials and no statistical difference between treatments was reported.^{35, 42} Death was most commonly due to fatal MI. No differences between groups were seen for the outcomes of stroke or MI, but these were rare outcomes across studies which may have been insufficiently powered. Across AF types, observational studies with >12 months followup may suggest lower risk of stroke and CHF for RFA versus medical therapy, however, the limitations of these studies weaken confidence in these findings and more robust evidence is needed. SF-36 MCS and PCS were similar for treatment groups at 24 months³⁵ and 48 months⁸³ in patients with paroxysmal AF in two separate studies; however, firm conclusions regarding the impact of RFA on HRQOL in the long term are not possible; only two small studies measuring SF-36 at different times provided data. Up to 48 months, RFA remained superior to medical therapy at improving freedom from recurrence of any atrial arrhythmia in patients with paroxysmal AF (pooled RR 1.24, 95% CI 1.11 to 1.47), however, reablation was common (pooled estimate 24%, 95% CI 12.6 to 41.5). Rehospitalization for cardiac causes was more common in the medical treatment group in one trial of patients with paroxysmal AF followed to 48 months which included repeat procedures and crossover to ablation.⁸³ Another trial only reported hospitalization for heart failure in two patients in the medical therapy group by 24 months and none in the ablation arm.³⁵

In the one comparative observational study of mixed AF types,⁵¹ mortality was comparable between treatments at 16 months in one observational study, stroke was more common in the medical treatment group and a higher proportion of RFA patients maintained sinus rhythm compared with medical therapy. Conclusions regarding RFA superiority in this population are not possible due to study limitations and imprecision and the strength of evidence was insufficient for all outcomes.

Short-Term (≤12 Months) Efficacy and Effectiveness

Data were sparse for the primary short-term efficacy outcomes comparing catheter ablation (RFA or cryoballoon ablation) with medical therapy which limits the conclusions that can be drawn (Tables 32 and 33). None of the observational studies contributed data on outcomes ≤12 months.

Table 32

Key findings and strength of evidence for Key Question 1: Short-term (≤12 months) comparative efficacy and effectiveness of RFA versus medical therapy for AF.

Table 33

Key findings and strength of evidence for Key Question 1: Short-term (≤12 months) efficacy and effectiveness outcomes comparing cryoballoon ablation with medical therapy for AF.

Medicare-Relevant Population

None of the included RCTs were in a Medicare-eligible population and no subanalyses by age were provided. Neither of the observational studies in people ≥65 years provided data on short-term effectiveness.

General Population

For the comparison of RFA with medical therapy, the incidence of all-cause mortality >30 days, stroke >30 days and MI >30 days was low for both treatment groups regardless of AF type and no distinct risk patterns based on treatment were seen. While there were seven small RCTs which reported on all-cause mortality >30 days,^{38, 39, 41, 43, 45–47} there was a paucity of trials reporting on stroke or MI >30 days regardless of AF type.

In patients with paroxysmal AF, evidence was low across three trials that there were no statistical differences between RFA and medical therapy for all-cause mortality past 30 days and across two trials for MI >30 days.^{38, 46, 47} There is insufficient evidence regarding stroke from one trial, which reported none for RFA or medical therapy and no conclusions are possible.

Conclusions regarding the impact of RFA on HRQOL in both the short- or long-term were not possible. In general, for most HRQOL measures, findings were similar between treatment groups at six to 12 months followup with the few exceptions noted in the summary tables below.

This is not to say that there is not improvement in HRQOL or symptoms following ablation, but there are challenges in confirming this with the currently available evidence from comparative studies that met the inclusion criteria. While individual studies reported statistically significant results favoring RFA over medical therapy for specific measures or isolated domains of measures, others did not. The variety of measures used, varied measurement timing and extensive cross-over make it difficult to draw firm conclusions regarding the impact of catheter ablation on HRQOL compared with use of antiarrhythmic drugs (AADs) as it is difficult to effectively evaluate consistency across measures and effect sizes. This was the basis for the rating of insufficient evidence. Although the Short Form-36 (SF-36) questionnaire was employed most frequently, measurements occurred at a variety of time points across studies and data were presented in different ways. These factors, combined with possible differences in patients receiving catheter ablation as a first line treatment versus a second line treatment precluded meaningful pooling of data. Where data were pooled, for the SF-36 Physical Component Score (PCS), different analysis methods yielded differing results, calling the stability of the estimate into question. Across four clinically heterogeneous trials with discrepant time frames (6–48 months), the SF-36 bodily pain domain was significantly improved in the RFA groups in three^{36, 37, 46, 83} of the four trials that provided data, however, given the differences in clinical populations and time frames, definitive conclusions are not advisable. The Minnesota Living with Heart Failure Questionnaire (MLHFQ) was reported in three small trials in patients with persistent AF and concomitant heart failure. Two of three studies reported improvement favoring RFA at 6 months; the third didn't reach significance but tended to favor RFA (low strength of evidence). One of the trials reported that the difference was also evident at 12 months (insufficient strength of evidence). For all measures, there was likely insufficient sample size in many studies to detect differences between treatments. Clinically, a primary indication for performing RFA is to improve symptoms. While it may be that patients experience relief of symptoms following RFA, the majority of the HRQOL measures reported may not capture the type and range of symptoms experienced by those with AF or the potential impact of RFA on patient symptoms. There were inadequate data to evaluate the influence of use of RFA as a first versus second-line treatment on HRQOL.

There was moderate evidence that RFA was superior to medical therapy at improving freedom from recurrence (pooled RR 3.06, 95% CI 2.35 to 3.90 across 4 RCTs)^{38, 44, 46, 47} of any atrial arrhythmia in patients in patients with paroxysmal AF and low strength of evidence that maintenance of sinus rhythm was significantly better following RFA compared with medical therapy in the short term, based in intention-to-treat analysis. In interpreting these findings, differences in definitions, methods of measurement and symptom status across studies need to be considered as does the substantial crossover which occurred in most studies. The need for reablation by 12 months varied widely across studies in patients with paroxysmal AF (0% to 43%). Overall, hospitalization was more frequent in patients who received medical therapy versus RFA. In those with paroxysmal AF, rehospitalization was more common in the medical treatment groups by 12 months across two studies^{44, 46} and the trial population followed to 48 months previously described.⁸³ Studies did not provide detail regarding reasons for hospitalization and the extent to which hospitalization for reablation procedures or crossover from medical therapy to ablation were included. Variability in reporting across trials makes it difficult to synthesize information.

In patients with persistent AF, strength of evidence for all outcomes was low or insufficient. Similar to findings in those with paroxysmal AF, there was low evidence that there was no difference between treatments for all-cause mortality and insufficient information for no difference for stroke. There is insufficient evidence on the impact of RFA on HRQOL compared with medical therapy. RFA was again superior to medical therapy for keeping patient free from recurrent atrial arrhythmia with rates of 58 to 74 percent in the RFA group versus 4 to 58 percent in the medical therapy group across three trials^{35, 42, 83} but strength of evidence was considered low as was the case for improved maintenance of sinus rhythm following RFA. The need for reablation by 12 months in those with persistent AF was variable (pooled estimate 20.1%, 95% CI 11 to 34).

Data for the comparison of cryoballoon ablation versus medical therapy was limited; only one moderate-sized RCT which combined patients with different AF types was identified and provided data on short-term outcomes only (Table 33).⁸⁶ Freedom from protocol-defined treatment failure, which was defined as absence of detectable AF after the blanking period, use of a nonstudy AAD, or any nonprotocol intervention (e.g., RFA), was significantly greater in the cryoballoon ablation group compared with the group treated medically (cryoballoon ablation: 69.9%; medical therapy: 7.3%). The strength of evidence for this outcome was low and strength of evidence was insufficient for other outcomes. Again, there was substantial crossover from medical therapy to cryoballoon ablation.

Conclusions regarding the primary outcomes of interest for this report are not possible for the comparison of cryoballoon ablation with RFA (Tables 34 and 35). Neither of the identified trials, one in those with paroxysmal AF and one in a mixed population reported on these outcomes. Both trials suggest that freedom from recurrence is lower and need for reablation is higher following cryoablation compared with RFA, but small trial sizes may preclude demonstration of statistical significance and strength of evidence is insufficient. Observational studies reported no significant differences between treatments for these outcomes and didn’t alter the strength of evidence.

Table 34

Key findings and strength of evidence for Key Question 1: Long-term (>12 months) efficacy and effectiveness outcomes comparing cryoballoon ablation with RFA for AF.

Table 35

Key findings and strength of evidence for Key Question 1: Short-term (≤12 months) efficacy and effectiveness outcomes comparing cryoballoon ablation with RFA for AF.

Harms

Primary clinically important outcomes related to harms were mortality <30 days, stroke <30 days, atrial fibrillation <3 months post-ablation, cardiac tamponade, pericardial effusion, pulmonary vein stenosis, and drug intolerance requiring discontinuation.

Harms were reported for RFA versus medical therapy by 13 RCTs^{35–47, 83} and six comparative observational studies (Table 36).^48–53 In order to better assess the risk of rare adverse events, data from 16 case series of at least 1000 patients specifically designed to evaluate the incidence of adverse events following catheter ablation were also included.^{55–65, 67–71} In addition, one meta-analysis of case series was also included.⁶⁶ While most case series used RFA, some used a mix of different energy sources.

Table 36

Key findings and strength of evidence for Key Question 2: Comparative short-term and long-term complications and harms associated with RFA versus medical therapy for AF.

Medicare-Relevant Population

One observational study in patients with persistent AF reported on outcomes for 412 patients over the age of 70, thus pertinent to the Medicare population, however no differences were seen in harms rates from this study compared to those from other studies of the general population.⁴⁸ A significantly higher stroke risk following RFA (2.6%) versus medical therapy (0.4%) and significantly lower risk of antiarrhythmic drug intolerance requiring discontinuation following RFA (2.6%) versus medical therapy (12.7%) were reported. Due to study limitations and the fact that these data were from a single study, it is difficult to draw definitive conclusions.

General Population

Overall in the general population, risk of any major adverse event within 30 days (all-cause mortality, stroke, myocardial infarction, major bleeding/hemorrhage, need for transfusion and heart failure) from RCTs ranged from 0 to 6.3 percent in those receiving RFA and from 0 to 5.7 percent for those receiving medical therapy. Cardiac tamponade risk ranged from 0 to 9.5 percent following RFA. In studies of patients with heart failure and AF risks were higher; overall risk of any major complication was 14.8 percent of all procedures in one RCT and risk of cardiac tamponade ranged from 3.8 to 9.5 percent across three RCTs. These overall ranges are from controlled studies conducted in tertiary referral centers with procedures performed by experienced personnel and may not reflect broader clinical practice outside of such centers.

With regard to individual harms, both all-cause mortality and stroke within 30 days of treatment were relatively rare events as reported in the comparative studies, and no patterns were detected according to type of AF. Overall 30-day mortality rates ranged from 0 to 0.7 percent in the RFA group and 0 to 4.2 percent in the medical therapy group as reported by five RCTs^{35, 37, 41, 42, 47} and one cohort study.⁴⁸ Data from 11 case series supported the conclusion that 30-day mortality rates are low following ablation, with mortality reported in 0 to 0.8 percent of procedures or patients.^{55–57, 59–62, 64, 65, 67, 70} Stroke within 30 days of treatment was reported to occur in 0 to 4.8 percent of RFA patients (three strokes total) and 0 percent of medical therapy patients based on data from eight RCTs.^{35, 37, 40–42, 45, 46} Data from eight case series and one meta-analysis of case series reported low 30-day stroke rates that ranged from 0.08 to 0.8 percent patients or procedures.^{55, 57, 59, 61, 64, 66–68, 70}

There was no difference in the 3-month AF recurrence rate in two RCTs of paroxysmal AF patients.^{35, 46} Two RCTs of persistent AF patients (with heart failure) suggested a lower 2- to 3-month risk of AF following RFA versus medical therapy.^{37, 43}

The risk of cardiac tamponade following RFA was moderate in four RCTs of paroxysmal AF patients (pooled risk 1.7%, 95% CI 0.8 to 3.6).^{35, 38, 42, 47} Although risk was higher in three very small RCTs (i.e., 73 patients total) of persistent AF patients (pooled risk 5.5%, 95% CI 2.1 to 13.7), small sample sizes may preclude accurate determination of risk.^{39, 40, 60} No cases were reported in two comparative observational studies.^{48, 52} Across eight case series plus one meta-analysis of case series, cardiac tamponade was reported in 0.6 to 6.7 percent of ablation patients or procedures.^{55–57, 59–61, 66, 69} Comparative studies with greater sample size are needed to determine whether the risk of this complication is higher in persistent versus paroxysmal AF patients.

Pericardial effusion risk ranged from 0.5 to 4.5 percent as reported by six RCTs^{35, 41, 42, 45, 47, 83} and were less than 2 percent in one cohort study⁴⁸ and four case series.^{55, 58, 64, 65} Pulmonary vein stenosis developed in 0 to 3 percent of RFA patients as reported by seven RCTs; there was not a clear difference in this risk in RCTs of paroxysmal versus persistent AF patients or between those with 12 versus 24 months of followup. Pulmonary vein stenosis was reported in 0 to 7.1 percent (i.e., 6/85) of patients across three comparative observational studies, though definitive conclusions could not be made due to study limitations,^{48, 51, 52} and in 0 to 1.5 percent of ablation patients or procedures across six case series.^{55, 57–60, 64}

Drug intolerance requiring discontinuation was reported in three RCTs, in which 0 to 23 percent of medical therapy patients discontinued antiarrhythmic medication due to adverse events such as pro-arrhythmia, bradycardia, thyroid dysfunction, and sexual impairment.^{36, 44, 47} However, due to limited duration and/or usage of medical therapy in the RFA group it is difficult to make comparative conclusions.

Data on harms for the comparison of cryoballoon ablation versus medical therapy were limited as a single study was available (Table 37).⁸⁶ There were no deaths in either group and one stroke in the cryoballoon ablation group within 30 days of treatment initiation. The incidence of cardiac tamponade following cryoballoon ablation was also low (0.9%).⁸⁶ Limited data on harms for the comparison of cryoballoon ablation with RFA were available; with two small RCTs^{88, 89} and two comparative observational studies^{76, 79} identified (Table 38), very few definitive conclusions can be made. Due to limited reporting on harms from the RCTs^{74, 75} and general poor quality of the comparative observational studies^78–81 reporting major outcomes, no definitive conclusions can be made.

Table 37

Key findings and strength of evidence for Key Question 2: Comparative short-term and long-term complications and harms associated with cryoballoon ablation versus medical therapy for AF.

Table 38

Key findings and strength of evidence for Key Question 2: Comparative short-term and long-term complications and harms associated with cryoballoon ablation versus RFA for AF.

Findings in Relationship to What is Already Known

Findings in this review are generally consistent with prior systematic reviews which included evaluation of catheter ablation versus medical therapy.^{22, 24}

This review expands information available from previous reviews in a number of ways. First, more recent publications comparing RFA with medical therapy with information on longer-term outcomes (>12 months) were included as were studies comparing different energy sources. An attempt to specifically identify and summarize studies focused on the Medicare population was made. Based on input from Key Informants during Topic Refinement, analyses were stratified based on AF type (paroxysmal and persistent) when possible, as these are clinically different populations. Analyses to evaluate differential efficacy and harms for use of catheter ablation as a first-line treatment versus a second-line treatment, age, patient characteristics and comorbidities, provider characteristics and other factors were considered; however, data were too sparse to draw conclusions. This report also expands analyses of ablation-specific adverse events and includes a broader spectrum of outcomes such as reablation, echocardiographic parameters and biomarkers.

The two previous reviews yielded similar findings to the current review with respect to the benefit of RFA over medical therapy in improving freedom from recurrence of AF and/or maintenance of sinus rhythm.^{22, 24} The current review, however, extends the findings of both reviews report to include longer-term outcome data and a broader spectrum of endpoints, stratifying them by AF type. Similar to this current report, the Washington State Health Technology Assessment (2013)²⁴ found low frequency of mortality or stroke not attributed to the procedure, with no differences between the ablation and medical therapy groups and there was low quality of evidence that there was no difference between RFA (or cryoablation) and medical therapy for the primary harms of interest. Both previous reports focused on the general population and reported a paucity of data in Medicare populations and for important clinical endpoints such as all-cause mortality, stroke, and heart failure, particularly in the long term.

Comparison of catheter ablation approaches/techniques to each other is an important question of interest to CMS but was beyond the scope of this review. There was substantial heterogeneity across included studies with respect to ablation techniques and sites which precluded their assessment in this review. Previous reviews provide some insight into this question. In the 2013 AHRQ review, information on the following aspects of the procedure was captured: type of ablation catheter used, whether cavo-tricuspid isthmus (CTI) ablation was performed in conjunction with the PVI, whether ablation of complex fractionated atrial electrograms (CFAEs) was done, and whether sites other than CTI and CFAEs were ablated. However, the data could not be pooled due to the insurmountable heterogeneity observed in different aspects of the procedure. RCTs that compared different approaches (i.e., PVI, wide area circumferential ablation [WACA], addition of right or left lines, CFAE) were included in the Washington State HTA.²⁴ In terms of freedom from recurrence, there was low quality evidence that WACA was favored over PVI, moderate-quality evidence that PVI plus CFE favored PVI alone, and moderate-quality evidence that there was no difference in recurrence following PVI versus PVI plus left lines or plus right lines. The HTA did not compare different ablation techniques.

Applicability

The applicability of the findings from this review is described below.

Implications for Clinical and Policy Decisionmaking

RFA is increasingly being used to treat AF. The bulk of the available evidence compares RFA with medical therapy and in patients with paroxysmal AF. Evidence in this report provides insights that may be useful for clinical and policy decisionmaking on use of RFA compared with medical therapy to treat patients with paroxysmal AF and persistent AF as distinct clinical populations and on use of RFA as a first or second-line treatment. Evidence for shorter-term and longer-term efficacy, effectiveness, and safety is also valuable to decisionmaking. For clinical outcomes, there is insufficient evidence for the use of cryoballoon ablation compared with either medical therapy or RFA to support evidence-based decisionmaking in the general population and no evidence for these comparisons in the Medicare population.

Limitations of the Review Process

The findings presented have limitations related to the approach and scope of this review. First, comparative evaluation of ablation techniques and approaches was beyond the scope of this review. There was substantial heterogeneity across included studies with regard to techniques and approaches that precluded comparative evaluation of studies. Though evaluation of mapping modalities and strategies was also beyond the scope of this review, we found insufficient information from included studies to assess mapping.

Stratification by AF type was felt to be clinically important and stratification to assess data at followup at >12 months was important to answering Key Questions. This resulted in fewer studies available for pooling within followup strata. Profile likelihood methods were used to provide more conservative estimates and confidence intervals given the small number of studies. This, combined with sparse data for many outcomes, may have limited the ability to explore statistical heterogeneity and precluded ability for further subgroup analyses.

Non-English studies were excluded and searches for studies published only as abstracts were not conducted. Formal assessment of publication bias was not conducted as there were fewer than 10 studies available for outcomes based on AF type, and research indicates that such methods can be misleading with smaller numbers of studies.⁹⁴

Every attempt was made to assure that variables and outcomes were assessed and abstracted accurately; however, wide variability across studies (in the quality of reporting of study methods, in how outcomes were defined, and in which patients were included) has the potential for introducing inaccuracies.

Limitations of the Evidence Base

Important limitations of the evidence base include the sample size of the available trials, limited data available on primary clinical outcomes particularly at followup times >12 months, and the substantial crossover from medical therapy to catheter ablation in most trials. These factors make it difficult to draw strong conclusions regarding the effects and benefits of catheter ablation. Only one RCT comparing cryoballoon ablation with medical therapy was identified and two small trials comparing cryoballoon ablation with RFA were identified but didn’t provide data on primary outcomes of interest. This precludes drawing conclusions regarding the comparative effectiveness of various energy sources.

The evidence base was constrained by the methodological limitations of the included studies. Common methodological shortcomings included unclear allocation concealment (only one trial documented concealed allocation) and lack of assessor blinding for primary outcomes. Four studies did not report information on random sequence generation.^{38, 41, 44, 83}

Although not a factor for determination of individual study quality or overall strength of evidence, the high frequency of crossover from medical therapy to ablation in most included studies may hinder drawing definitive conclusions regarding the full benefits and harms of catheter ablation compared with medical therapy.

Study sizes were likely insufficient to effectively determine risk of the primary clinical outcomes (e.g., mortality,) for either group or to detect statistical differences between treatment groups. Two recent large observational studies that reported on the primary outcomes of interest in this report were identified but were excluded as there was insufficient information on use of AADs (the focus of this review) versus rate control medications or no treatment in the control groups.^{95, 96} Both studies suggest that risk of stroke was significantly lower in patients receiving catheter ablation for AF compared with those who did not receive ablation in followup to 3.5 years. One study, based on administrative data from a large regional registry stratified stroke risk by CHADS2 score and age reported that reduced stroke risk was present in all age groups and across all CHADS2 risk profiles.⁹⁵ The other study based on Taiwan’s National Health Insurance claims database, reported that there was no association between RFA and lower mortality or hospitalization for heart failure.⁹⁶ Although both studies attempted to control for confounding, the possibility of residual confounding from unmeasured factors should be considered as should the limitations of administrative data (such as misclassification) when interpreting these findings.⁹⁷ Findings should be confirmed in additional high-quality studies that provide specifics regarding treatments received.

A variety of HRQOL measures were used at varied time frames across trials; this, along with small sample sizes, limited the ability to pool data or draw firm conclusions regarding the impact of catheter ablation on HRQOL, as discussed previously.

For the freedom from recurrence outcome, the definitions used by each study were accepted. This outcome was variably defined across studies, with variations in the type of arrhythmia (i.e., AF, AF or atrial flutter, or any atrial arrhythmia) and whether study reports were limited to symptomatic or asymptomatic AF or not. Maintenance of sinus rhythm appeared to be used interchangeably with freedom from recurrence in many studies, although these two outcomes are not the same.

Most studies focused on the intermediate outcome related to freedom from recurrence. This was variably defined and adjudicated across studies; there was heterogeneity across studies regarding whether recurrence included any atrial arrhythmia or AF only, whether symptomatic and asymptomatic recurrences were included, and whether characteristics related to duration were considered. In addition, blanking periods ranged from 1 to 3 months across 11 RCTs. This variability in study protocols likely introduces variation in the cross-study calculations of the proportion of those free from AF after the blanking period.

There is less evidence on the effectiveness of catheter ablation in patients with persistent AF and limited evidence on use of catheter ablation as a first-line treatment, limiting the conclusions that can be drawn in these instances. No data were available to assess the differential efficacy and harms of catheter ablation by patient or provider characteristics.

Research Gaps

Gaps in evidence were identified for each Key Question.

Key Question 1: RFA or cryoablation versus medical therapy and comparison of cryoablation versus RFA

• There were limited data on the impact of RFA compared with medical therapy on final clinical outcomes such as mortality, particularly in the long term. Long-term data are particularly sparse those with persistent AF.
• Data were sparse for the comparison of cryoballoon ablation versus medical therapy and for the comparison of cryoballoon ablation with RFA for all clinically-relevant outcomes.
• No data from high-quality comparative studies on Medicare-relevant populations were identified.
• Across studies, there was lack of a standardized, consistent method of measuring, monitoring and reporting “freedom from recurrence” or maintenance is sinus rhythm.
• Conclusions related to HRQOL and symptom relief were not possible from the included studies due the variety of HRQOL measures reported across different time frames and inconsistency in statistical significance.
• Ideally future studies would conceptualize strategies for treatment of AF, such as catheter ablation and medical therapy, and evaluate the impact of such strategies on hard clinical outcome such as death and stroke.

Key Question 2: Harms

• Comparative data on rare harms were limited by study sample sizes and/or study quality particularly in those with persistent AF. This was true for the comparisons of RFA versus medical therapy as well as for the comparisons of cryoballoon ablation with medical therapy or RFA.

Key Question 3: Differential efficacy, effectiveness, harms

• Available RCTs did not have sufficient power to evaluate differential efficacy or harm of catheter ablation (RFA or cryoballoon ablation) verses medical therapy or for comparison of cryoballoon ablation with RFA for specific patient subgroups or provider settings. No conclusions regarding which patients may benefit most are possible or regarding which patients may not benefit from catheter ablation are possible with current evidence.
• Limited data on use of RFA as the first treatment of choice (first-line therapy) versus a second-line therapy following failure of antiarrhythmic medications were available. No data for cryoballoon ablation were available to evaluate its use as first versus second-line therapy.

Some of these gaps may be addressed via the CABANA Trial, (https://www.cabanatrial.org/), which is scheduled for completion in March 2018. This RCT conceptualizes ablation and medical therapy as strategies for treatment of AF and compares first-line ablation with pharmacologic therapy for reducing the composite endpoint of total mortality, disabling stroke, serious bleeding, or cardiac arrest in patients with atrial fibrillation. The target sample size is 2200 and current enrollment is about 1650 patients across 126 study sites in 10 countries. The trial will likely have good representation of Medicare patients, as to be eligible for the trial, patients must be either 65 years of age or older or have at least one risk factor for stroke. In addition, inclusion of patients with specific risk factors for stroke (e.g. hypertension, diabetes) may facilitate better understanding of the impact of catheter ablation in subpopulations with specific risk factors. One potential limitation of the trial include the use of a composite outcome, which may preclude adequate characterization and representation of important individual outcomes. This may limit conclusions which can be drawn regarding the efficacy of catheter ablation versus medical therapy to affect clinical outcomes such as mortality, stroke, and development of CHF if there is insufficient power to evaluate these as separate outcomes although such outcomes are listed as s intermediate outcomes for this study. Another limitation is its span over a very long period of time (started in 2009) during which technology is evolving quickly. Thus, equipment and techniques that were used earlier in the trial may not be relevant to clinical practice when the trial ends. . It is not clear if there will be sufficient data to evaluate efficacy separately by AF type or to compare ablation techniques. The planned followup of approximately 5 years will provide additional evidence on the longer-term impact of catheter ablation on clinical outcomes. Results from CABANA will enhance the current evidence base and provide much needed information on efficacy and safety particularly in those >65 years of age. In addition to CABANA, several other trials are currently evaluating aspects of catheter ablation for the treatment of AF, including a total of 14 relevant ongoing clinical trials were identified using the United States National Institute of Health clinical registry (www.ClinicalTrials.gov) (Appendix, Table H8). There are three current ongoing trials comparing cryoablation to radiofrequency ablation which are estimated to be completed between June 2015 and January 2017. One study will examine the efficacies of two cryoablation procedures, standard and short, in comparison to radiofrequency ablation (NCT01913522). One study will utilize RFA in both treatment groups (NCT01521988). There are two current ongoing trials comparing cryoablation to medical therapy. Both trials are for first line treatments and are expected to be completed between January 2016 and June 2017. Patients in both trials have paroxysmal AF. Six trials comparing radiofrequency ablation to medical therapy are currently ongoing. The trials are expected to be completed between May 2015 and August 2019. All but one study includes a special patient population (NCT01341353). In one trial, the medical therapy group undergoes EEC in addition to antiarrhythmic drug treatment (NCT01850277). In this same study, the medical treatment group has the option to undergo ablation. Only one study is non-randomized (NCT01341353). Two studies appear to compare medical therapy to either cryoablation or radiofrequency ablation. The trials are expected to be completed between March and June of 2018. One study includes a special patient population that is over the age of 60 years old and has paroxysmal AF (NCT01570361). The second study includes a mixed AF patient population (NCT00911508). In addition to the evidence the CABANA trial and other currently open trials may yield, there is a need to evaluate effectiveness across clinical settings and provider skill levels outside of randomized trials. There are several key elements that might augment the productivity of future research. For example, high-quality clinical registries, designed a priori to address specific clinical questions, could provide an important source for addressing some research gaps. Such registries could be designed to extend the observations from randomized trials; this approach might differ from currently available study results by allowing for evaluation of the durability of the effects seen with longer followup, including a larger sample size to examine rare events, and including a broader range of patients to facilitate evaluation of how well patient groups are represented in clinical trials. Specific components that could be considered for capture include: more detailed information on patient and provider characteristics; AF details (duration, prior treatments, and episode severity); procedural details (e.g., mapping, catheter use, and technique/approach); management of patients before, during, and after procedure; and acute procedural outcomes as well as longer-term outcomes and complications. Such a registry could be used to address important issues such as how outcomes in clinical practice compare with outcomes observed in clinical trials and how outcomes are associated with characteristics of patients and providers. Registries could help determine performance measures and assist with quality improvement, post-marketing safety public reporting as well. There are, however, a number of limitations and factors to consider for registries. One challenge of creating and maintaining a registry is that research questions may change after data collection begins which may shift the importance of some of the elements collected. Other challenges include potential difficulty in balancing the burden of high-quality data collection with acceptable registry size. Additionally, registries designed to test specific medical devices face practical problems including variability in insurance coverage for specific devices, influence of operator characteristics while using the same device, and difficulty in differentially identifying unique devices in use.⁹⁸

Registry studies would not replace the need for high-quality comparative studies. Pragmatic trials and comparative observational studies that are designed to reduce bias may facilitate evaluation of effectiveness and safety in the “real world” and over the long term in particular. Such studies could provide a foundation for registry development. Methodologically rigorous prospective cohort studies comparing treatments based on standardized protocols can potentially provide high-quality data. Such studies might be complemented with data from registries.

There may be some value in performing meta-analysis on individual patient data from relevant clinical trials to more effectively stratify outcomes by patient characteristics, comorbidities, and factors such as use of catheter ablation as an initial first-line treatment versus a second-line treatment. The small sample sizes of individual trials may, however, limit such exploration.

In order to evaluate the extent to which there is differential efficacy or harm for specific subpopulations, clinical trials need to have sufficient statistical power to stratify by important groups and to test for statistical interaction. Future studies, regardless of design, would benefit from the use of standardized definitions and methods of measuring salient outcomes (e.g., freedom from recurrence)and detailed reporting of co-interventions that may influence outcomes (e.g., use of antiarrhythmic medications in ablation groups and use of anticoagulation).

The technology and strategies related to catheter ablation (and mapping) continues to evolve. Information on newer technologies including phased radiofrequency ablation (phased RFA), laser balloon ablation, contact force-guided radiofrequency (CF-guided RFA), and mesh ablation is emerging in the peer-reviewed literature. Present literature indicates they may be effective alternatives to currently available technologies; however additional evidence is needed to verify early findings.^{88, 89, 99–102}

Future systematic reviews will need to include technologies as they are approved and become widely available as well as updated studies of newer technologies such as cryoballoon ablation that are already approved but have a limited evidence base. Given the rapid evolution of catheter ablation, it will be important to update reviews such as this in the near future.