The frequency and potential sequelae of NOA and potential hemodynamic benefits of vasopressin in septic shock patients are well described; however, evidence describing the relationship between vasopressin’s catecholamine-sparing effects and the incidence on NOA are scarce. This retrospective cohort study compared the incidence of NOA between early and late vasopressin initiation in septic shock patients with no prior history of cardiac arrhythmias that required norepinephrine and vasopressin for hemodynamic support. The results suggest early vasopressin initiation does not reduce the incidence of NOA.
Results of this study were consistent with prior data observing use of vasopressin alone or in conjunction with other catecholamines in septic shock results in no difference in the incidence of arrhythmias, with observed rates of 9.1% and 6.9% in the early and late vasopressin groups, respectively. A subgroup analysis of only sepsis randomized-control trials, performed by McIntyre et al. observed no difference in the relative risk of atrial fibrillation in patients receiving catecholamines and vasopressin analogues compared to catecholamines alone (10 vs 15%; RR 0.76; 95% CI 0.55-1.05; p = 0.09) (22). Similarly, the subgroup of only vasopressin randomized-controlled trials observed no difference in the relative risk of ventricular arrhythmias (8 vs 9%; RR 0.88; 95% CI 0.56-1.38; p = 0.57). The slightly lower observed incidence in our study may be explained by the inclusion of patients with no history of cardiac arrhythmias and/or a definition of NOA requiring clinical intervention. McIntyre et al. acknowledged limitations surrounding catecholamine and vasopressin timing, dosing, and titration strategies, which the presented study described. In addition, exclusion of patients requiring additional/alternative vasoactive agents was not done by McIntyre et al. and minimizes the potential for confounding in the presented study.
In contrast, a retrospective cohort study by Reardon et al., demonstrated initiation of vasopressin within six hours of septic shock onset was associated with a significantly lower incidence of NOA compared to later initiation (37 vs 63%; p < 0.01) (23). The stark difference in the incidence of NOA between Reardon et al. and our study may be secondary to differences in patient population and definitions of history of arrhythmias and/or NOA. Higher APACHE II scores and rates of alternative vasoactive agent use highlight important differences between Reardon et al. and our study. The presented study focused on the most common vasoactive agents in septic shock, norepinephrine and vasopressin, to minimize confounders. Further, Reardon et al. provided minimal detail surrounding identification of patients’ history of cardiac arrhythmias. In contrast, the presented study, first, utilized historical ICD information to exclude patients with a history of NOA. Further, inclusion of pertinent medication administration data before septic shock onset allowed exclusion of patients not identified via ICD review or who developed a NOA secondary to an unrelated cause. Finally, Reardon et al. similarly classified NOA; however, our study required EKG or telemetry evidence and was reviewed and confirmed by a board-certified cardiologist, in addition to administration of an anti-arrhythmic or rate control agent. This was done to ensure a focus on clinically relevant NOA and minimize potential charting or administrative errors. Notably, variable incidence of NOA has been observed across prior critical care and sepsis literature due to heterogeneity of study design, patient populations, disease severity, and arrhythmia identification strategies (7).
We conducted a multivariable logistic regression analysis to control for confounders and identify potential predictors of NOA in septic shock patients with no history of cardiac arrhythmias. This analysis failed to reveal a significant association between time to vasopressin initiation and the incidence of NOA. To our knowledge, this is the first study to incorporate time to vasopressin initiation in a predictive model for NOA. In a recent meta-analysis, Bosch et al. observed vasopressor use was associated with a greater than 50% change in odds of NOAF (7). Several other sepsis-related risk factors were identified as potential targets for NOAF prevention. Our study’s inclusion of vasopressin initiation and timing built upon their findings and should serve as a basis for future study.
Although our study was not powered to detect a mortality difference, we observed a statistically significant association between early vasopressin initiation and reduced in-hospital mortality. Prospective and retrospective evaluations of norepinephrine with or without adjunctive vasopressin as well as norepinephrine vs vasopressin monotherapy, which are not directly applicable to our study, have yielded conflicting results related to mortality. Reardon and colleagues reported a higher overall incidence of in-hospital mortality, but no significant difference between early vs late vasopressin administration (89% vs. 89%, p = 1.0) (23). These results contradict our study’s findings but are likely driven by prior described differences in severity of illness and vasopressor prescribing practices. Patients in this study represented an acute-on-chronically ill, refractory shock population, as evidenced by median Charlson Comorbidity Index and mAPACHE II score of 5 and 21 in both groups, respectively, and baseline rates of mechanical ventilation over 90% in both groups. While well-matched in these respects, numeric differences in MAP at septic shock onset, adjunctive corticosteroid use, and need for new RRT receipt may suggest a more severely ill patient population in the late vasopressin group and account for the observed mortality difference between groups.
Alternatively, the significantly longer time to initiation of vasopressin and total norepinephrine exposure may reflect a cohort receiving vasopressin as salvage therapy later during septic shock. Furthermore, the higher incidence of RRT in the late vasopressin group may relate to the higher mortality observed reflecting the hypothesis of a more severely ill population in the late vasopressin group. While study design precludes the ability to determine causality in this regard, the finding associated with this secondary outcome and others remains hypothesis-generating.
The current study had several strengths. To date, it was the largest study to assess the effect of early administration of vasopressin on NOA in septic shock patients with no history of cardiac arrhythmias treated with only norepinephrine and vasopressin. It attempted to minimize and control for factors known or likely to contribute to NOA development in sepsis/septic shock via stringent inclusion and exclusion criteria and the use of a multivariable logistic regression analysis inclusive of novel details surrounding vasopressin timing. This study’s internal validity was strengthened by the novel approach employed to ensure NOA events were both clinically relevant and truly new onset. The use of a historical ICD-9/10 code query combined with medication administration, clinical, EKG, and/or telemetry correlation by a board-certified cardiologist reflects methodology not described previously.
Limitations of this study include first, the observed low incidence of the primary outcome compared to some prior literature, which limited the scope of our statistical analysis. In order to avoid co-linearity and overfitting the regression model, investigators limited the number of included covariates, despite it being likely that many factors contributed to NOA in septic shock patients. Second, the utilized definitions for history of cardiac arrhythmias and NOA may still have allowed for inclusion of patients inappropriately and/or failure to capture NOA events because patients did not have prior documented ICD codes within the 15-hospital health system, did not require clinical intervention, and/or lacked telemetry, EKG, or other clinical support on manual chart review. Third, the study spanned eight years in which temporal changes in vasopressin prescribing and clinical outcomes were possible. While an expected trend toward reduced mortality was observed during the study timeframe, no noticeable temporal patterns related to vasopressin timing or incidence of NOA were observed. Other limitations of this study remained inherent to its retrospective, single-center design. These included limited generalizability to patients less severely ill, admitted to non-medical or surgical ICUs or requiring additional/alternative vasopressor or inotropic agents. Lastly, the study design precluded the ability to control for all confounding and/or biases that may occur within an institution that lacked any formal restrictions surrounding vasopressin administration.