Page 251 - Read Online

Lam et al. Vessel Plus 2019;3:27 I http://dx.doi.org/10.20517/2574-1209.2019.009 Page 3 of 9

moderate PVL, severe PVL, any PVL, 30-day mortality, renal failure, stroke, major bleeds, hospital LOS,
ICU LOS, fluoroscopic time, contrast volume, and procedure time.

Analysis
A random effects meta-analysis was performed with R packages meta and metafor (R version 3.5.0; R
Foundation for Statistical Computing, Vienna, Austria). If only the median and quartiles were provided
in a study, mean and standard deviation were estimated based on protocol found in the literature [11,12] .
Risk ratios (RRs) weighted based on number of events and sample size were calculated using the Mantel-
Haenszel method for perioperative dichotomous outcomes. The event rates and percentages for all
subgroups were calculated based on raw data for perioperative dichotomous outcomes. Mean differences
(MD) were used to analyze continuous variables as calculated by the inverse variance method. Means were
2
also calculated based on raw data for continuous variables. Heterogeneity was reported as low (I = 0%-
[13]
2
2
25%), moderate (I = 26%-50%), or high (I > 50%), as reported previously in the literature . Data was
analyzed as a pooled group then as 2 sub-groups: matched and unmatched observational studies.
One additional post hoc analysis was conducted. It compared studies that only used TTE in the MAC
group to studies that had mixed TTE and TEE usage in the MAC group on the outcome of PVL.

RESULTS
After searching in Ovid and Embase, 1,133 citations were screened. Sixteen citations (n = 3510 patients)
fulfilled the inclusion and exclusion criteria. There were 10 unmatched (n = 2037 patients) [7,14-22] and 6
matched (n = 1473 patients) [4,23-27] retrospective observational studies [Supplementary Figure 1].

The studies were rated at low or moderate risk of bias [Supplementary Table 1]. All studies had concurrent
controls. Twelve studies [4,14-18,20,21,24-27] did not report follow up rates. One study lost more than 10% of
[7]
patients in the 1 year follow-up . TTE-TAVI and TEE-TAVI shared similar baseline characteristic,
including predicted mortality risk scores [Supplementary Tables 2 and 3].

Primary outcome: PVL
Event rates and relative risks for the primary outcome and periprocedural events are summarized in Table 1.
Corresponding forest plots are shown in Figure 1 and Supplementary Figures 2-8. The incidence rate in
pooled results for mild PVL was not significantly different when TTE-TAVI was compared to TEE-TAVI
2
(32.1% vs. 34.0%, RR: 0.95, 95%CI: 0.72 to 1.25, P = 0.71, I = 54%). Similar non-significant results were
2
found for moderate PVL (6.6% vs. 6.7%, RR: 0.87, 95%CI: 0.55 to 1.38, P = 0.56, I = 0%), severe PVL (0.9%
2
vs. 0.3%, RR: 1.39, 95%CI: 0.06 to 33.21, P = 0.84, I = 53%), and any PVL ³ mild (18.4% vs. 21.4%, RR: 1.01,
2
95%CI: 0.83 to 1.23, P = 0.92, I = 36%). All subgroups also did not show any significant difference between
TTE-TAVI and TEE-TAVI.

Periprocedural events
When comparing patients that had TTE and TEE at the time of TAVI, there was no significant difference
in pooled event rates for 30-day mortality (4.0% vs. 5.0%, RR: 0.69, 95%CI: 0.49 to 1.13, P = 0.14, I = 0%)
2
[Table 1]. Similarly, when TTE-TAVI was compared to TEE-TAVI, no significant differences were found
for renal failure (5.4% vs. 4.0%, RR: 0.86, 95%CI: 0.58 to 1.26, P = 0.43, I = 0%), stroke (2.8% vs. 2.7%, RR:
2
2
0.85, 95%CI: 0.52 to 1.39, P = 0.52, I = 0%), and major bleed (2.8% vs. 5.1%, RR: 0.58, 95%CI: 0.28 to 1.22,
2
P = 0.15, I = 45%) [Table 1]. All subgroups also did not show a significant difference when TTE-TAVI was
compared to TEE-TAVI.

Hospital resources
The weighted means and MDs for the continuous outcomes are summarized in Table 2. Corresponding
forest plots are depicted in Figure 2 and Supplementary Figures 9-12. Hospital LOS was significantly