Time-scale-dependent association between deep hypnotic exposure and intraoperative hypotension burden: a case-level, event-aligned, and within-case analysis using the VitalDB database

<title>Abstract</title> <p> <bold>Background:</bold> Intraoperative hypotension and deep hypnotic states are both clinically important during general anesthesia. Although deeper hypnosis may plausibly contribute to arterial pressure reduction through anesthetic-related cardiovascular depression, observational relationships between hypnotic depth and arterial pressure in routine practice are shaped by anesthetic titration, hemodynamic treatment, and time-dependent clinical feedback. Using the VitalDB database, we examined the association between Bispectral Index (BIS) &lt;40 exposure and intraoperative hypotension burden at three complementary levels: case-level, event-aligned, and within-case analyses. <bold>Methods:</bold> This was a retrospective observational database study. We screened VitalDB cases with both BIS and invasive arterial pressure recordings. Deep hypnotic exposure was defined as the proportion of valid BIS monitoring time with BIS &lt;40. Hypotension burden was defined as the proportion of valid mean arterial pressure (MAP) monitoring time with MAP &lt;65 mmHg. The primary case-level outcome was substantial hypotension burden, defined as MAP &lt;65 mmHg for at least 10% of valid monitoring time. Multivariable logistic regression was used to assess the association between BIS &lt;40 exposure and substantial hypotension burden, adjusting for age, sex, body mass index, American Society of Anesthesiologists physical status, preoperative hypertension, and preoperative diabetes. To further explore temporal relationships, we performed event-aligned analyses around each case's first qualified hypotension onset and within-case paired-window analyses comparing an earlier stable window with the pre-hypotension window. The first qualified hypotension event was selected to avoid statistical dependence from multiple events within the same case and to prevent a small number of event-rich cases from dominating the average trajectory.We prespecified a confounding framework that distinguished baseline patient-level confounders from intraoperative management variables. The primary model adjusted for baseline variables, whereas intraoperative variables such as blood loss, fluid administration, transfusion, and vasopressor use were examined in extended sensitivity models because they may represent surgical complexity, evolving patient status, or treatment responses to hypotension. <bold>Results:</bold> Among 2103 prescreened case identifiers, 1428 had complete BIS and arterial pressure tracks. After signal quality control, 1394 cases were included in the final case-level analysis, of which 1380 had complete covariates for adjusted models. Substantial hypotension burden occurred in 390 of 1394 cases. At the case level, cases with substantial hypotension burden had a lower BIS &lt;40 time proportion than cases with lower hypotension burden (32.22% [interquartile range, 10.77-61.19] vs 47.74% [22.61-68.88]; P&lt;0.001). The BIS &lt;40 time proportion was weakly and inversely correlated with the MAP &lt;65 time proportion (Spearman rho=-0.187; P&lt;0.001). In adjusted logistic regression, each 10% increase in BIS &lt;40 time was associated with lower odds of substantial hypotension burden (adjusted odds ratio, 0.885; 95% confidence interval, 0.846-0.926; P&lt;0.001). The direction of association was unchanged when a stricter hypotension definition (MAP &lt;65 mmHg for &gt;=20% of valid monitoring time) or adjusted linear regression with continuous hypotension burden was used. In contrast, time-series analyses identified local peri-event changes. In the event-aligned analysis, mean MAP decreased from 78.7 mmHg during the pre-event window (-15 to -5 min) to 65.7 mmHg around hypotension onset (-1 to +1 min; P&lt;0.001) and subsequently recovered to 74.6 mmHg during the post-event window (+5 to +10 min; P&lt;0.001). Mean BIS decreased modestly from 44.8 before the event to 42.2 near onset (P&lt;0.001), but did not increase significantly after MAP recovery (41.9; P=0.153). The probability of BIS &lt;40 increased from 35.0% before the event to 41.5% near onset (P&lt;0.001) and remained similar after the event (42.7%; P=0.211). In the strict within-case paired-window analysis, mean MAP decreased from 83.1 mmHg in the stable window to 75.2 mmHg in the pre-hypotension window (mean difference, -7.85 mmHg; P&lt;0.001), mean BIS decreased from 43.9 to 41.8 (mean difference, -2.03; P&lt;0.001), and the BIS &lt;40 time proportion increased from 37.5% to 44.4% (mean difference, 6.83 percentage points; P&lt;0.001). <bold>Conclusions:</bold> At the case level, greater cumulative BIS &lt;40 exposure was not associated with greater intraoperative MAP &lt;65 hypotension burden; instead, a weak inverse association was observed. However, event-aligned and within-case analyses showed a modest decrease in BIS and a higher probability of BIS &lt;40 around specific hypotension events. These findings suggest that the relationship between hypnotic depth and arterial pressure is time-scale dependent: cumulative case-level metrics may obscure short-term peri-event BIS-MAP coupling. The inverse case-level association should not be interpreted as evidence that deep hypnosis protects against hypotension, but rather as a reflection of time-dependent physiology, clinical management, and aggregation effects. These findings should be interpreted as associations rather than causal effects because BIS values, arterial pressure, and intraoperative treatments are dynamically co-managed during anesthesia. Clinical trial number: Not applicable. </p>