The major findings of this study are that PAC placement was not associated with a survival benefit in critically ill patients with cardiac diseases, either in surgical or non-surgical patients. Furthermore, those who received PAC insertion also had fewer vasopressor- and ventilation-free days than the non-PAC group. In surgical patients, PAC insertion had no effect on AKI occurrence but was associated with a higher volume of fluid administration. In non-surgical patients, PAC insertion was associated with a higher risk of AKI occurrence, while no association with fluid input was observed.
Since its introduction, PAC has been used in the management of critically ill patients. Importantly, PAC measures, particularly in critically ill patients, are frequently inaccurate[11]. Misinterpretation of data is generally accepted to be common and can potentially lead to mismanagement of the patients. Thus, PAC-directed therapy, including the use of fluids and inotropes, may be ineffective or harmful in some statuses[12].
Though Shaw and colleagues showed that PAC use was associated with an overall potential benefit during adult cardiac surgery [7], the majority of studies have shown no mortality benefit or higher mortality when assessing the influence of PAC placement on critically ill patients with cardiac surgeries[12–15]. Despite the debate about PAC use in the management of cardiac surgery, an international survey showed that practitioners preferred using PAC for most cardiac surgeries[16]. In the present study, no significant difference was observed between the PAC group and the non-PAC group in patients undergoing cardiac surgeries. Similarly, no survival benefit has been definitely confirmed in non-surgical patients with cardiac diseases. Previously, the ESCAPE trial found that PAC use had no effects on overall mortality and hospitalization in patients with congestive heart failure[3]. Gabriel et al. conducted a retrospective cohort study and found that PAC use was associated with higher mortality in patients with heart failure[17]. In those with cardiogenic shock, PAC was associated with lower mortality and in-hospital cardiac arrest[17]. Recently, another retrospective study evaluating PAC use in patients with acute myocardial infarction-cardiogenic shock excluding cardiac surgery showed that the PAC cohort had a higher in-hospital mortality, longer length of stay, higher hospitalization costs, and fewer discharges to home[18]. Here, in this study, no mortality benefit of PAC use was identified in non-surgical patients with cardiac diseases. However, as PAC is not routinely indicated in these patients and there is currently no consensus on the use of PAC in these patients[19, 20], future studies are warranted.
In this study, patients in the PAC group had higher severity of illness scores and were more likely to receive mechanical ventilation and vasopressors at admission. Although these factors indicated a more severe condition, we used PSM analysis and found no mortality benefit of PAC insertion after adjustment for confounding factors. Moreover, the initial timing of PAC insertion had little effect on clinical outcomes. Of note, a longer indwelling time of PAC was associated with a worse outcome.
PAC measurement of real-time parameters can guide timely and accurate interventions, including the use of fluids, vasopressors, and inotropes. On the other hand, it may also lead to a state of overtreatment. We found that PAC placement was associated with a longer support of ventilation and vasopressors. Differences were observed between surgical and non-surgical patients regarding the management of fluid and occurrence of AKI. Consistent with a previous cohort study[8], PAC insertion had no effect on AKI occurrence in cardiac surgery patients. While in non-surgical patients, PAC insertion was associated with a higher risk of AKI occurrence. Similar results were demonstrated in patients with heart failure[17].
When comparing mortality rates for different subgroups considering whether or not they underwent surgeries, a seemingly strange pattern can be observed: for surgery and non-surgery subgroups, mortality rates in PAC group were higher than those in the non-PAC group, but the total mortality rate was higher in the non-PAC group. This constitutes a textbook example of a statistical phenomenon known as Simpson’s paradox which refers to the observation that aggregating data across subpopulations may yield opposite trends (and thus lead to reversed conclusions) when considering subpopulations separately[21]. The key to understanding the phenomenon lies in the fact that we are dealing with relative frequencies. However, such percentages conceal the absolute numbers of patients within each age group. Considering these absolute numbers sheds light on how the phenomenon can arise: the distribution of cases with PAC or without PAC differs significantly between the two subgroups. Particularly, the surgery group recorded a much higher proportion of patients with PAC than the non-surgery group. In summary, the larger share of PAC insertion among surgery patients combined with the fact that the non-surgery patients are at higher risk of death, explains the mismatch between total and subgroup mortality rates and thus gives rise to Simpson’s paradox in the data.
Several limitations in the present study should not be neglected. First, this is an observational single-center study retrospectively performed on electronic health record data. The generalizability of these findings to other institutions requires more investigation. Second, compared with the non-PAC group, patients in the PAC group had a higher severity of illness. Though PSM was used to balance these covariates in terms of their baseline risk, there may still be unmeasured confounders that bias the study findings. Well-designed RCTs with large sample sizes are needed in the future. Third, data surrounding pulmonary artery pressures were limited in this dataset; thus, we failed to perform further analysis. Fourth, in addition to PAC utilization, less invasive CO measurement techniques, such as FloTrac/Vigileo and PiCCO monitoring, have been increasingly used in clinical practice. However, there were no data available in the mimic database, and thus we could not perform further analysis.