Results
Of 246 patients in the review, 33 (13%) were PC-ECMO patients. This cohort were supported on ECMO a total of 179 days. The median age was 66 years old, and 73% of patients were male. The most common operations were coronary artery bypass (30%), followed by combination of major cardiac procedures (i.e.: CABG/AVR, AVR/MVR, etc.) cases (21%) and aortic surgery (18%). VA ECMO represented 82% of cases (n 27), with the remaining six cases (18%) of cases being VV support. The median duration of ECMO support was 4 days and median length of stay was 23 days. Median operative (OR) time was 478 (range 192-1262) minutes, median CPB time was 233 (range 102-899) minutes, and median XC time was 125 (range 34-388) minutes. On average, patients went back on pump one time before transitioning to ECMO (range 0-6 times) and received a median of 2 units of red blood cells (range 0-20 units) in the operating room. ECMO was initiated immediately in the operating room in 22 patients (66%), and 11 patients (34%) had ECMO initiated later, at a median of 18 hours postoperatively (range 1-71). This data included four patients who were transferred to our institution following their index operation.
Overall survival data were as follows: overall ability to wean from ECMO 61%, hospital survival 55%. One-month survival was 45%. Long term survival was captured for all patients who were discharged, representing a total of 15,068 days of follow up data. The median follow up time for surviving patients was 1071 days [Range: 640-1996 days] or 35.2 months after discharge. The estimated 12 month survival for all PC-ECMO patients was 40%, and was 33% for both 24 and 36 months. The estimated 12 month survival for patients surviving their index hospitalization was 66% (95%CI: 40%, 83%), and survival was 60% at both 24 and 36 months (95% CI: 33%, 78%). Survival data are further described via Figure 1 below.
Data was examined in multiple ways to determine differences between groups. Discussion will focus on both pertinent positives and negatives of our investigation. Basic demographic information was examined to determine if any effect on survival or ability to wean from ECMO existed. There was a borderline, statistically significant difference in age for 30-day survivors versus those who expired (median 60 versus 71 years old, p 0.053) but otherwise there were no differences in the basic demographics collected (Figures 2-4).
Operative, CPB, and XC times were all examined for impact on length of stay, duration of support, ability to wean from ECMO and hospital survival. Total OR time includes all time in the OR from “wheels in to wheels out” including induction of anesthesia, line placement and all associated activities prior to and after surgery. Operative time is the “skin to skin” time of the actual procedure performed. Interestingly, longer CPB and XC times were associated with increased ability to wean from ECMO (Figure 2) though this did not translate into hospital or 30-day survival differences (Figures 3-4).
The type of operation was also examined to determine if this played a role in survival outcomes. There was no difference in the types of operations performed when compared to ability to wean from ECMO (p 0.27), hospital survival (p 0.16) or 30 day survival (p 0.59). Likewise we compared open chest status, reoperation for postoperative hemorrhage and cannulation location (central versus peripheral). In terms of how patients were placed on ECMO, 36% remained with central cannulation, and 64% were cannulated via peripheral technique. More patients in the centrally cannulated group were also able to wean from ECMO (central = 83% versus peripheral = 48%; p 0.07). This trend did continue through 30 day survival after hospital discharge (central = 67% versus peripheral = 33%; p 0.08) but did not reach statistical significance (Figures 3-4).
The chest was left open following surgery in 39% of cases, with a primary closure performed in the remainder of cases. Twelve patients were taken back to the operating room specifically for hemorrhage or tamponade and this did not correlate with initial open chest status or site of cannulation (p 0.14). A hemorrhagic event was interestingly associated with an increased chance of weaning from ECMO (92% versus 43%, p 0.01), and longer length of stay (18 versus 30 days, p 0.026) though this difference did not translate to a statistically significant difference in hospital survival or 30 day survival (p 0.15, p 0.08 respectively) (Figures 3-4).
Subgroup analysis was performed on several groups of interest: the frozen mediastinum patients, patients with immediate versus delayed ECMO initiation and patients supported with VA versus VV ECMO.
Of the 33 patients reviewed, there were 16 patients (48%) who met criteria for frozen mediastinum which included 14 patients with prior sternotomy and 2 patients with chest radiation in the past.
As one might expect, those patients with a frozen mediastinum status required more time for the operative procedure, but their ability to wean from ECMO and survival was not affected (Figure 5).The type of ECMO support required (VV versus VA), and location of cannulation site (peripheral versus central) was similar between these two groups (p 0.65, 0.28 respectively). There was also no difference in rates of re-operation for hemorrhage or for open chest status (p 1.00 and 0.48 respectively). These patients did not require more units of RBCs in the operating room (p 0.13) contrary to what one might suspect.
Long-term survival of patients who were discharged from the hospital was not significantly altered for patients with frozen mediastinum versus normal chest (55% versus 65% p 0.55).
Outcomes were compared to determine if a survival benefit might exist for patients who were placed on ECMO earlier in their PC course versus at some point after leaving the OR. We compared the same values with the results summarized in Figure 6. If ECMO was not initiated immediately in the operating room following the index procedure, the median time to ECMO was 18 hours (range 1-71 hours).
Patients who were placed on ECMO immediately spent more time in the operating room and more time on pump. They also required more RBC transfusions during the case. Patients who were placed on ECMO immediately were more likely to have an open chest at the end of the operation (59% versus 0%, p 0.002) and were also more likely to return to the operating room for hemorrhage (50% versus 9%, p 0.03). They were also more likely to be cannulated centrally compared to the delayed group (50% versus 9%, p 0.27). There was not a difference in type of support required, with VA representing the vast majority of patients in both groups (86% versus 73%, p 0.38).
There was no difference in ability to wean from ECMO, hospital or 30-day survival between the immediate or delayed ECMO groups. Likewise, there was not a statistically significant difference in long-term survival (immediate 50%, delayed 85% p 0.24).
Patients supported with VA versus VV ECMO were compared. There were six patients included in this study who required VV support within 72 hours of cardiac surgery. Four of the six patients were supported with an RVAD with in-line oxygenation for profound right ventricular dysfunction following surgery, and two were purely VV support. All of the above analysis were performed both including and excluding the six VV patients and the results were nearly identical. Of all of the comparisons made between the VA versus VV cohorts, there were no statistically significant findings (likely related to sample size) other than number of units of RBCs in the OR with the VA group requiring more blood (median 2.5 units versus 0 units, p 0.01). The remainder of the findings are summarized in Figure 7.
There was no difference in type of surgery performed, rates of reoperation for hemorrhage, site of cannulation, open chest status, or amount of time to ECMO initiation.
Thirty-six month follow up for hospital survivors was 75% for VV and 55% for VA, and did not reach statistical significance for variation (p 0.29), again likely due to a small sample size.