Carbon Dioxide, Blood Pressure, and Stroke: Reply.

Abstract

We thank Dr. Sessler1 for his interest in our article.2 Stroke is a devastating complication for surgical patients and a challenging outcome to study. Dr. Sessler raises important points, and we welcome the opportunity to engage in constructive dialogue to raise clinical and scientific awareness of this important topic.We think it is important to first describe for readers the context surrounding perioperative stroke research. First, the majority of postoperative strokes evade clinical detection (i.e., covert stroke, as mentioned by Dr. Sessler).3 As such, these strokes will not be recorded in routinely collected data (electronic health record, clinical registry, or billing databases) and, thus, they are unavailable for analysis. Second, intraoperative physiologic contributions to postoperative stroke remain incompletely understood. This is because cerebral perfusion is a complex process regulated by multiple physiologic variables (e.g., blood pressure, metabolic demand, autoregulation, chemoregulation). This physiologic complexity might explain, in part, why intraoperative hypotension is inconsistently identified as a risk factor. We do point out, however, that other studies not discussed in Dr. Sessler’s letter have described an association between intraoperative hypotension and stroke.4,5 Third, commonly used modeling techniques (e.g., logistic or binomial regression)6,7 cannot disambiguate the relationship between physiologically related variables and stroke; more sophisticated approaches such as the seemingly unrelated regression technique used in the current analysis may be more appropriate.8 Thus, all researchers in this field face a complex picture, including physiologic knowledge gaps that might account for inconsistent findings across studies.Our investigation was not designed to establish or approximate perioperative stroke incidence, which has been estimated at 0.1% in a broad noncardiac, nonneurologic, non-major-vascular surgical population.9 Rather, our goal was to implement a rigorous case-control matching strategy for testing physiologic hypotheses. To achieve this goal, manual identification of stroke cases was first required, which necessitated a detailed chart review across multiple study sites. During this activity, additional surgical cases were added to the Multicenter Perioperative Outcomes Group database, since data are regularly uploaded from existing and newly contributing sites. As expected, the available denominator of control cases grew during the course of the study. Because incidence estimation was not a goal of the study, the exact number of eligible control cases available at the initial time of stroke case identification was not recorded. For these reasons, the true incidence of perioperative stroke is not calculable from this study, nor should it be inferred from the flow diagram. By extension, calculating a number needed to treat or harm from this study would not be possible, given the lack of an accurate control event rate. However, given the more than 300 million surgeries performed around the globe each year,10 we believe that identifying modifiable risk factors for perioperative stroke that are under the control of anesthesia clinicians is worth pursuing.Dr. Sessler raises a valid point regarding selection bias given the broader population available for control case matching. To mitigate population-based selection bias, stroke cases and controls were matched to the same hospital, eligible surgeries were restricted to the same time range, and models were adjusted for time (year). We also point out that this larger pool of control cases provided more opportunities for optimal patient matching. Approximately 25% of control cases from the broader population were hand-reviewed to confirm the absence of stroke; typically, control cases are not manually reviewed in these retrospective studies.6,7 This would require labor-intensive efforts to evaluate large volumes of cases, which is impractical and can be prohibitive for large-scale database research. Nonetheless, the control cases in our study screened negative based on billing code data, and no strokes were found in the final control cases that were manually reviewed. Overall, the results only apply to recognized stroke cases, but, as was noted, this is a problem for all researchers in this field.We do agree that confounding is a concern with all retrospective, observational studies, including previously published work that has not identified an association between intraoperative hypotension and stroke.6,7 This is precisely why seemingly unrelated regression modeling was used to account for the two primary (and related) exposure variables of blood pressure and carbon dioxide. In addition to adjusting for relevant comorbidities, we matched and adjusted for total body blood volume and intraoperative hemorrhage, because anemia can increase stroke risk by challenging autoregulatory mechanisms.11 These adjustments were made in addition to the Mahalanobis distance case-control matching process described in our prespecified analysis plan.12 Thus, while residual confounding is certainly a possibility, we used mitigation strategies targeted to both comorbid and physiologic confounding where possible.Histograms of all available mean arterial pressure (MAP) and end-tidal (etco2) values are presented in figure 1. Of note, 72 of 122 (59%) stroke patients experienced MAPs less than 55 mmHg for at least 10 min, and 65 of 122 (53%) experienced at least 10 min with a MAP of 50 mmHg and etco2 of 28 mmHg. In terms of statistical approach, given adequate sample size, multivariable analysis should not be gated by nonsignificant bivariable associations. Furthermore, seemingly unrelated regression was not used in the bivariable analyses. The choice of seemingly unrelated regression was motivated by the physiologic relationship between blood pressure and carbon dioxide; this more rigorous and sensitive approach then revealed significant associations in the multivariable analysis.We do agree that there were likely multiple stroke etiologies, particularly for strokes occurring several days after surgery. Of note, however, delayed stroke recognition is common in hospitalized patients,13 and the temporal relationship to surgery remains unclear in many cases. Hypotension and dyscarbia may also contribute to embolic stroke by exacerbating cerebral hypoperfusion,14 and these physiologic disturbances may persist postoperatively, where it is generally more challenging to monitor blood pressure and ventilation. As previously stated, many physiologic knowledge gaps remain with respect to perioperative stroke.Last, it is important that we provide readers with the complete last sentence of our published article: “These physiologic perturbations may serve as risk factors that can be modified to reduce the incidence of postoperative stroke.”2 This sentence was modified and misquoted in Dr. Sessler’s letter to the editor in a way that misrepresents the causal significance we ascribed to the findings. Indeed, hypotension and dyscarbia may serve as stroke risk factors, but confirmatory, prospective trials are ultimately required, as we originally stated in our article.Supported by the National Institutes of Health (Bethesda, Maryland) grants L30GM116069 and K23GM126317 (Dr. Vlisides). Funding was also provided by departmental and institutional resources at each contributing site. In addition, partial funding to support underlying electronic health record data collection into the Multicenter Perioperative Outcomes Group registry was provided by Blue Cross Blue Shield of Michigan/Blue Care Network (Detroit, Michigan) as part of the Blue Cross Blue Shield of Michigan/Blue Care Network Value Partnerships program. Although Blue Cross Blue Shield of Michigan/Blue Care Network and Multicenter Perioperative Outcomes Group work collaboratively, the opinions, beliefs, and viewpoints expressed by the authors do not necessarily reflect the opinions, beliefs, and viewpoints of Blue Cross Blue Shield of Michigan/Blue Care Network or any of its employees.Dr. Vlisides receives support from Blue Cross Blue Shield of Michigan for research unrelated to this work. The other authors declare no competing interests.