Differential effects of calcium channel blockers combined with various opioids on length of stay in hypertensive cardiac surgery patients: analysis of the MIMIC-IV database Authors: Xinglu Li, Haitao Liu, Qian Lin, Yuanhui Zhang Published: 2026 ABSTRACT AbstractObjectiveThe co-administration of calcium channel blockers (CCB) and opioids is common in hypertensive patients undergoing cardiac surgery. This study aimed to investigate the association between combinations of CCB with different opioids and length of stay (LOS) in this population.MethodsThis retrospective study included 2,093 hypertensive adult patients who underwent cardiac surgery, drawn from the MIMIC-IV database. Potential covariates were systematically screened through univariate logistic regression and LASSO regression. Subsequently, generalized linear model (GLM) and multivariable logistic regression analyses were employed to evaluate the association between different CCB-opioids combinations and LOS. To further explore heterogeneity, subgroup analyses and interaction analyses were conducted.ResultsGLM revealed that the concomitant use of CCB with morphine (β = 4.354, 95%CI: 2.535–6.173), oxycodone (β = 2.418, 95%CI: 0.943–3.893), and tramadol (β = 2.382, 95%CI: 1.340–3.427) was significantly associated with prolonged LOS compared to CCB use alone. Notably, among these three opioid drugs, the CCB-oxycodone combination showed a robust association with prolonged hospital stay across the majority of adjusted logistic regression models. Subgroup analyses identified significant interaction effects (P for interaction < 0.001), showing that the association between CCB-oxycodone and extending LOS was substantially stronger in patients undergoing valve surgery and those with a CCI score > 5.ConclusionCo-administration of CCB and oxycodone is robustly associated with prolonged LOS in hypertensive cardiac surgery patients. These findings highlight the necessity of individualized analgesic strategies to mitigate drug interaction risks.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12872-026-06042-x.Keywords: Calcium channel blockers, Cardiac surgery, Opioid, Length of stay Introduction Cardiac surgery is a frequently performed major surgical procedure worldwide, and effective postoperative pain management constitutes a critical element of perioperative care [1–3]. Adequate analgesia plays a pivotal role in facilitating patient recovery, minimizing postoperative complications, and optimizing healthcare resource utilization, including reducing the length of hospital stay (LOS) [4]. Hypertension is among the most prevalent comorbidities in patients undergoing cardiac surgery [5–7], with many requiring ongoing antihypertensive therapy. Calcium channel blockers (CCB), as first-line agents for hypertension management, are commonly continued throughout the perioperative period [8]. Concurrently, opioids remain the standard of care for managing moderate to severe postoperative pain due to their potent analgesic efficacy [9]. Consequently, the concomitant use of CCB and opioids in hypertensive patients undergoing cardiac surgery represents a common yet complex clinical scenario that may predispose patients to potential drug–drug interactions (DDIs). Pharmacologically, CCB—particularly dihydropyridine derivatives—are recognized inhibitors of cytochrome P450 3A4 (CYP3A4) [10]. Importantly, several widely used opioids, including fentanyl and oxycodone, are predominantly metabolized via the CYP3A4 pathway [11–13]. The co-administration of CYP3A4-inhibiting CCB with these opioids may theoretically impair opioid metabolism, resulting in elevated systemic drug concentrations and an increased risk of adverse effects such as excessive sedation and respiratory depression [14, 15]. These adverse events may contribute to organ dysfunction and potentially prolong hospitalization. Despite this mechanistic plausibility, the real-world clinical impact of this interaction on key patient outcomes—particularly LOS—remains inadequately characterized. Furthermore, significant inter-agent variability exists in opioid metabolic pathways; for instance, morphine and hydromorphone are primarily metabolized through glucuronidation, a process largely independent of CYP3A4 activity [16]. This pharmacokinetic distinction suggests that the magnitude and clinical consequences of CCB-opioids interactions may differ substantially across opioid classes. To date, there is a paucity of direct comparative studies examining such differential effects, representing a critical knowledge gap. Elucidating drug-specific interactions is essential for informing evidence-based analgesic selection, improving patient safety, and enhancing overall perioperative care efficiency. Accordingly, this study aims to leverage a large-scale intensive care database to investigate, for the first time, the differential effect of combining CCB with various opioid classes—including fentanyl, hydromorphone, morphine, oxycodone, and tramadol—on LOS in hypertensive patients undergoing cardiac surgery. We hypothesize that, compared to opioids with minimal reliance on CYP3A4 metabolism (e.g., hydromorphone), those primarily metabolized by CYP3A4 (e.g., fentanyl and oxycodone) will be associated with prolonged hospital stays when used concomitantly with CCB. Methods Data source and study participants The Medical Information Mart for Intensive Care IV (MIMIC-IV) is a large-scale, publicly available, and de-identified database that contains comprehensive clinical data from patients admitted to the intensive care units (ICUs) and emergency department at the Beth Israel Deaconess Medical Center in Boston, Massachusetts, between 2008 and 2019. Developed and maintained by the MIT Laboratory for Computational Physiology, the database includes a wide array of information such as patient demographics, vital signs, laboratory measurements, medication records, diagnoses, procedures, and clinical notes. Given that all personally identifiable information has been removed from the dataset, this study qualifies for exemption from individual informed consent requirements under established ethical guidelines governing the secondary use of de-identified data. The study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement and adhered to the ethical principles set forth in the Declaration of Helsinki. The study initially identified 2,267 hypertensive patients who underwent cardiac surgery from the MIMIC-IV database. After excluding two patients with the LOS < 24 h, 2,265 patients remained. Subsequent exclusions were applied for severe liver disease (n = 11), severe renal insufficiency (n = 14), malignant tumors (n = 54), non-admission to the ICU (n = 67), and in-hospital mortality (n = 26). Ultimately, a total of 2,093 patients were included in the final analysis (Fig. 1). Flow chart of participants selection Exposure and clinical outcome The primary exposure of interest was the concomitant use of calcium channel blockers (CCB)—a class of antihypertensive medications—and various opioid analgesics, including fentanyl, hydromorphone, morphine, oxycodone, and tramadol. The primary outcome was LOS. Data collection The study included a range of potential covariates for analysis, including age, gender, race, insurance, marital status, smoking, drinking, body mass index (BMI), admission type, sentiment polarity, and sentiment subjectivity. Sentiment polarity and subjectivity were derived from natural language processing[NLP] of clinical notes. Polarity reflects the positive or negative tone of the clinical narrative[e.g., patient distress vs. stable recovery], while subjectivity measures the degree of personal emotion or opinion versus objective facts. In the absence of continuous structured pain scores, these were included as surrogate indicators of subjective patient comfort. Clinical measurements covered heart rate, respiratory rate, oxygen saturation (SpO2), Sequential Organ Failure Assessment (SOFA) score, Simplified Acute Physiology Score II (SAPSII), and Charlson Comorbidity Index (CCI). Laboratory values comprised partial pressure of oxygen (pO2), anion gap, glucose, creatinine, platelets, potassium, sodium, calcium, magnesium, red blood cells (RBC), red cell distribution width (RDW), white blood cells (WBC), monocytes percentage, lymphocytes percentage, neutrophils percentage, and urine output. Furthermore, interventions such as the type of surgery (coronary artery bypass grafting [CABG)] or valve), blood transfusion, mechanical ventilation (MV), non-invasive ventilation, and antibiotic use were considered, along with perioperative complications and comorbidities including anxiety, depression, coronary heart disease (CHD), chronic obstructive pulmonary disease (COPD), diabetes, myocardial infarction (MI), anemia, heart failure, cerebrovascular disease (CVD), and renal disease. Statistical analysis The normality of continuous variables was first assessed using the Shapiro-Wilk test. Normally distributed variables are presented as mean ± standard deviation (SD) and were compared between groups using the independent samples t-test. Non-normally distributed data are reported as median (interquartile range [IQR]) and were compared using the Mann-Whitney U test. Categorical variables are summarized as frequency (percentage) and were analyzed using either the chi-square or Fisher’s exact test, as appropriate. All statistical analyses were performed in R software (version 4.2.1), with a P-value < 0.05 considered statistically significant. First, covariates were systematically selected through a stepwise approach incorporating baseline characteristic analysis, univariate logistic regression, and LASSO regression. To evaluate the stability of the regression estimates across conceptually distinct clinical domains and to strictly prevent overadjustment, the covariates were purposefully partitioned into four mutually exclusive, parallel adjustment models rather than being entered simultaneously. Specifically, these independent models were defined as follows: Model 1 was adjusted exclusively for baseline admission and surgical characteristics (admission type, type of surgery, subjective status, and heart rate); Model 2 evaluated demographic and lifestyle factors independently (race, marital status, and alcohol consumption); Model 3 accounted solely for clinical severity and comorbidities (SOFA, CCI, MI, anemia, heart failure, and CVD); and Model 4 evaluated laboratory parameters as a distinct set (anion gap, glucose, creatinine, platelet count, calcium, magnesium, RBC, RDW, WBC, and monocytes). This domain-specific modeling strategy mitigated the risk of over-parameterization and overadjustment that could arise from combining numerous variables in a single model, thereby validating the robustness of the associations. Next, to assess the association between concomitant use of CCB and different opioid analgesics and LOS—a continuous outcome—generalized linear regression models were employed in hypertensive patients undergoing cardiac surgery. Subsequently, the sample was dichotomized into two groups: LOS < 7 days and LOS ≥ 7 days. This 7-day threshold was selected because it not only represents the median LOS in this cohort but also serves as a widely recognized clinical milestone in cardiac surgery. Typically, patients with an uncomplicated postoperative course are discharged within 5 to 7 days [17]; thus, a hospitalization extending beyond 7 days is commonly adopted in prior studies to define ‘prolonged LOS’, reflecting delayed recovery or the presence of postoperative complications. Logistic regression analysis was then conducted to re-examine the association between drug combination use and prolonged hospitalization. To assess the robustness of our findings against potential unmeasured confounding (e.g., unmeasured pain severity or intraoperative anesthetic details), we calculated the E-value for the primary significant associations. The E-value quantifies the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome to fully negate the observed association, conditional on the measured covariates. To further investigate potential mechanisms underlying the observed association, three additional analyses were performed: (1) comparison of perioperative complications across different medication combination groups to evaluate clinical burden; (2) subgroup analyses to identify potentially susceptible populations; and (3) assessment of interaction effects for stratification variables that showed differential associations across subgroups. Furthermore, to account for the increased risk of Type I error, the Benjamini-Hochberg False Discovery Rate (FDR) correction was also applied to the P-values in analyses. Statistical significance was defined as FDR-adjusted P-values (q-values) < 0.05. Results Characteristics of patients A total of 2093 patients were included in the analysis, categorized into a short LOS (< 7 days, n = 1139) group and a prolonged LOS (≥ 7 days, n = 954) group, with baseline comparisons presented in Table 1. Patients in the prolonged LOS group were significantly older, had a lower proportion of males, a higher proportion of non-white patients, a lower proportion of married individuals, and a higher prevalence of drinking history (all P < 0.05). This group was also more likely to be admitted on an urgent or emergency basis and exhibited lower sentiment polarity with higher sentiment subjectivity (all P < 0.05). Clinically, patients with a prolonged LOS presented with a more severe profile, characterized by a higher heart rate, higher respiratory rate, lower SpO2, and significantly higher SOFA, SAPSII, and CCI scores (all P < 0.05). This was mirrored in laboratory findings, where the prolonged LOS group showed lower pO2, higher anion gap, higher glucose, higher creatinine, higher platelets, lower sodium, higher calcium, higher magnesium, higher RBC, higher RDW, lower WBC, higher monocytes, and lower lymphocytes; their potassium levels also differed significantly (all P < 0.05). In terms of interventions and outcomes, the prolonged LOS group had a higher rate of CABG surgery, a higher rate of antibiotic use, and a greater incidence of perioperative complications. Finally, they carried a significantly higher burden of comorbidities, including CHD, diabetes, MI, anemia, heart failure, CVD, and renal disease. Conversely, no statistically significant differences were observed between the groups in terms of BMI, insurance type, smoking status, neutrophil, urine output, blood transfusion, MV, non-invasive ventilation, or in the prevalence of anxiety, depression, and COPD (all P > 0.05). Table 1Comparison of demographic and clinical dataVariableOverall (n = 2093)LOS < 7 days (n = 1139)LOS ≥ 7 days (n = 954)P-valueDemographicsAge, years68.000 (60.000, 75.000)67.000 (60.000, 74.000)69.000 (60.000, 76.000)0.020BMI, kg/m ²28.900 (25.600, 32.900)28.600 (25.600, 32.500)29.200 (25.700, 33.400)0.079Gender, n (%)0.004 Male1478 (70.616)834 (73.222)644 (67.505) Female615 (29.384)305 (26.778)310 (32.495)Race, n (%)< 0.001 White1497 (71.524)862 (75.680)635 (66.562) Non-white596 (28.476)277 (24.320)319 (33.438)Marital status, n (%)< 0.001 Married1260 (64.815)739 (68.236)521 (60.511) Other684 (35.185)344 (31.764)340 (39.489)Insurance, n (%)0.242 Medicare910 (43.478)482 (42.318)428 (44.864) Other1183 (56.522)657 (57.682)526 (55.136)Smoking0.907 No2035 (97.229)1107 (97.191)928 (97.275) Yes58 (2.771)32 (2.809)26 (2.725)Drinking< 0.001 No2014 (96.226)1111 (97.542)903 (94.654) Yes79 (3.774)28 (2.458)51 (5.346)Admission type, n (%)< 0.001 Elective296 (14.142)181 (15.891)115 (12.055) Urgent477 (22.790)148 (12.994)329 (34.486) Emergency204 (9.747)67 (5.882)137 (14.361)Sentiment polarity0.084 (0.069,0.097)0.086 (0.071,0.099)0.082 (0.068,0.096)< 0.001Sentiment subjectivity0.328 (0.306,0.352)0.324 (0.303,0.348)0.333 (0.310,0.358)< 0.001Vital Signs & Clinical ScoresHeart rate, beats/min80.000 (72.000, 85.000)80.000 (71.000, 83.000)80.000 (75.000, 86.000)< 0.001Respiratory rate, breaths/min16.000 (14.000, 17.000)16.000 (14.000, 17.000)16.000 (14.000, 18.000)< 0.001SpO2, %100.000 (99.000, 100.000)100.000 (99.000, 100.000)100.000 (98.000, 100.000)< 0.001SOFA2.000 (1.000, 4.000)2.000 (1.000, 4.000)3.000 (1.000, 4.000)< 0.001SAPSII34.000 (28.000, 40.000)33.000 (28.000, 39.000)35.000 (29.000, 41.000)< 0.001CCI5.000 (4.000, 6.000)4.000 (3.000, 5.000)5.000 (4.000, 6.000)< 0.001Laboratory ValuespO2, mmHg358.000 (279.000, 415.000)368.000 (294.000, 425.000)344.000 (262.000, 401.000)< 0.001Anion gap, mEq/L13.000 (11.000, 15.000)12.000 (10.000, 14.000)14.000 (12.000, 16.000)< 0.001Glucose, mg/dL111.000 (98.000, 141.000)108.000 (97.000, 135.000)115.000 (100.000, 149.000)< 0.001Creatinine, mg/dL0.900 (0.700, 1.000)0.800 (0.700, 1.000)0.900 (0.800, 1.100)< 0.001Platelets, K/µL170.000 (131.000, 215.000)152.000 (122.000, 194.000)189.000 (146.000, 233.000)< 0.001Potassium, mEq/L4.200 (3.900, 4.500)4.200 (4.000, 4.500)4.200 (3.900, 4.500)< 0.001Sodium, mEq/L139.000 (137.000, 141.000)139.000 (137.000, 141.000)139.000 (137.000, 141.000)0.004Calcium, mg/dL8.500 (8.100, 9.000)8.400 (8.000, 8.800)8.800 (8.200, 9.200)< 0.001Magnesium, mg/dL2.300 (2.100, 2.700)2.300 (2.100, 2.600)2.400 (2.100, 2.700)0.002RBC, m/µL3.740 (3.120, 4.370)3.430 (3.000, 4.130)4.060 (3.450, 4.530)< 0.001RDW, %13.200 (12.700, 14.100)13.100 (12.600, 13.800)13.500 (12.900, 14.600)< 0.001WBC, K/µL9.000 (6.900, 12.700)9.600 (7.200, 13.700)8.300 (6.700, 11.200)< 0.001Monocytes, %3.900 (2.800, 5.400)3.700 (2.700, 5.000)4.200 (2.900, 5.800)< 0.001Lymphocytes, %17.000 (12.400, 22.500)17.200 (13.100, 22.500)16.400 (11.500, 22.700)0.003Neutrophils, %76.500 (70.200, 81.000)76.400 (70.600, 80.600)76.700 (69.600, 81.700)0.680Urine output, mL125.000 (75.000,205.000)125.000 (75.000,200.000)125.000 (75.000,220.000)0.842Intervention, yes, n (%)Type of surgery0.004 CABG1186 (56.665)613 (53.819)573 (60.063) Valve907 (43.335)526 (46.181)381 (39.937)Blood transfusion82 (3.918)37 (3.248)45 (4.717)0.085MV1290 (64.759)689 (63.502)601 (66.262)0.199Non-invasive ventilation1277 (64.106)685 (63.134)592 (65.270)0.322Antibiotic2069(98.853)1121(98.420)948(99.371)0.042Perioperative complications135(6.450)58(5.092)77(8.071)0.006Comorbidities, yes, n (%)Anxiety312 (14.907)155 (13.608)157 (16.457)0.068Depression255 (12.183)134 (11.765)121 (12.683)0.522CHD1558 (74.439)823 (72.256)735 (77.044)0.012COPD190 (9.078)91 (7.989)99 (10.377)0.058Diabetes795 (37.984)391 (34.328)404 (42.348)< 0.001MI578 (27.616)244 (21.422)334 (35.010)< 0.001Anemia1340 (64.023)656 (57.594)684 (71.698)< 0.001Heart failure353 (16.866)108 (9.482)245 (25.681)< 0.001CVD167 (7.979)61 (5.356)106 (11.111)< 0.001Renal disease132 (6.307)40 (3.512)92 (9.644)< 0.001Median (P25, P75) for continuous variables and counts (percentage) for categorical variablesAbbreviations: LOS Length of stay, BMI Body mass index, SpO2 Oxygen saturation, SOFA Sequential organ failure assessment, SAPSII Simplified acute physiology score II, CCI Charlson Comorbidity Index, pO2 partial pressure of oxygen, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, CABG Coronary artery bypass grafting, MV Mechanical ventilation, CHD Coronary heart disease, COPD Chronic obstructive pulmonary disease, MI Myocardial infarction, CVD Cerebrovascular disease Comparison of demographic and clinical data Median (P25, P75) for continuous variables and counts (percentage) for categorical variables Abbreviations: LOS Length of stay, BMI Body mass index, SpO2 Oxygen saturation, SOFA Sequential organ failure assessment, SAPSII Simplified acute physiology score II, CCI Charlson Comorbidity Index, pO2 partial pressure of oxygen, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, CABG Coronary artery bypass grafting, MV Mechanical ventilation, CHD Coronary heart disease, COPD Chronic obstructive pulmonary disease, MI Myocardial infarction, CVD Cerebrovascular disease Comparison of medication use We further analyzed the utilization patterns of CCB and several opioids between the two LOS groups (Fig. 2). Patients in the prolonged LOS group had a significantly higher rate of CCB administration (P < 0.05). With regard to opioid analgesics, excluding hydromorphone, the use of fentanyl, morphine, oxycodone, and tramadol was significantly more frequent in the group with a prolonged hospital stay (all P < 0.05). Differences in medication use among patients stratified by LOS. (A) CCB; (B) Fentanyl; (C) Hydromorphone; (D) Morphine; (E) Oxycodone; (F) Tramadol. Abbreviations: CCB, calcium channel blockers; LOS, length of stay Crucially, when examining the concomitant use of CCB and opioids (Fig. 3), we found that the combined administration of CCB with each of the following opioids—fentanyl, hydromorphone, morphine, oxycodone, and tramadol—was significantly more prevalent in patients with prolonged LOS. This suggests a strong association between the co-administration of CCB and these opioid agents and extended hospitalization duration. Differences in the use of combination medications among patients stratified by LOS. (A) CCB and Fentanyl; (B) CCB and Hydromorphone; (C) CCB and Morphine; (D) CCB and Oxycodone; (E) CCB and Tramadol. Abbreviations: CCB, calcium channel blockers; LOS, length of stay Covariates screening Prior to conducting the regression analysis, we performed covariate screening. First, all variables with statistically significant differences in the baseline comparison were included in a univariate logistic regression analysis (Supplementary Table 1). Subsequently, the significant variables from the univariate analysis were entered into LASSO regression for further screening to reduce the number of variables and prevent multicollinearity (Fig. 4a&b). We applied the standard error of the minimum distance rule to select the most parsimonious and effective model, which corresponded to an optimal lambda value of 0.021. Ultimately, the LASSO regression selected 23 covariates for inclusion in the final multivariable regression model (Supplementary Table 2). These variables were: admission type, type of surgery, race, marital status, drinking, MI, anemia, heart failure, CVD, heart rate, anion gap, glucose, creatinine, platelets, calcium, magnesium, RBC, RDW, WBC, monocytes, SOFA, CCI, and sentiment subjectivity. The LASSO regression analysis for identifying key features. (a) Coefficient profile plot; (b) 10-fold cross-validation curve plot. Abbreviations: LASSO, least absolute shrinkage and selection operator Association between combinations of CCB with different opioids and LOS We utilized generalized linear models to assess the association between combinations of CCB and different opioids with the LOS (Table 2). Using the “CCB use without an opioid” group as the reference in each analysis, the results consistently show that concomitant use of CCB with several opioids was associated with a significantly prolonged LOS. The combination of CCB and morphine exhibited the strongest and most robust association, yielding the largest effect size after adjustment for laboratory values (Model 4: β = 4.354, 95% CI [2.535, 6.173], P < 0.001). Similarly, combinations of CCB with oxycodone (Model 4: β = 2.418, 95% CI [0.943, 3.893], P = 0.001) and tramadol (Model 4: β = 2.382, 95% CI [1.340, 3.427], P < 0.001) were significantly associated with longer LOS across all models. The CCB–fentanyl combination also showed a significant association with prolonged LOS after adjusting for clinical severity (Model 3: β = 2.566, 95% CI [0.959, 4.174], P = 0.002) and laboratory values (Model 4: β = 4.055, 95% CI [1.480, 6.630], P = 0.002). In contrast, the combination of CCB and hydromorphone was not associated with prolonged LOS in any model. These research findings reveal a significant and stable association between the prolonged hospital stay and the combined use of CCB and morphine, oxycodone, and tramadol in patients with hypertension after cardiac surgery. Importantly, after applying the FDR correction for multiple testing across the five opioid classes, the associations between the concomitant use of CCB with morphine, oxycodone, and tramadol and prolonged LOS remained statistically significant. Table 2Generalized linear regression analysis of drug combinations and LOSModel 1Model 2Model 3Model 4β (95% CI)P-valueq-valueβ (95% CI)P-valueq-valueβ (95% CI)P-valueq-valueβ (95% CI)P-valueq-valueCCB and Fentanyl CCB (+), Fentanyl (-)Reference-Reference-Reference-Reference- CCB (+), Fentanyl (+)-0.083 (-1.951, 1.786)0.9310.9310.972 (-0.733, 2.676)0.2640.3522.566 (0.959, 4.174)0.0020.0054.055 (1.480, 6.630)0.0020.005 CCB (-), Fentanyl (+)-2.018 (-3.835, -0.201)0.0300.055-1.136 (-2.771, 0.499)0.1730.2410.998 (-0.557, 2.553)0.2080.2842.740 (0.200, 5.279)0.0340.06 CCB (-), Fentanyl (-)-2.071 (-4.126, -0.016)0.0480.08-1.836 (-3.695, 0.023)0.0530.084-1.511 (-3.221, 0.199)0.0830.1241.262 (-1.590, 4.114)0.3860.473CCB and Hydromorphone CCB (+), Hydromorphone (-)Reference-Reference-Reference-Reference- CCB (+), Hydromorphone (+)0.099 (-1.020, 1.218)0.8620.8910.397 (-0.791, 1.584)0.5130.561.039 (-0.023, 2.102)0.0550.0850.639 (-0.502, 1.780)0.2720.355 CCB (-), Hydromorphone (+)-2.011(-2.783, -1.239)< 0.0010.003-1.959 (-2.770, -1.148)< 0.0010.003-0.913 (-1.648, -0.177)0.0150.03-1.038 (-1.849, -0.226)0.0120.025 CCB (-), Hydromorphone (-)-1.866 (-2.566, -1.166)< 0.0010.003-1.945 (-2.683, -1.208)< 0.0010.003-1.433 (-2.088, -0.778)< 0.0010.003-0.986 (-1.726, -0.245)0.0090.019CCB and Morphine CCB (+), Morphine (-)Reference-Reference-Reference-Reference- CCB (+), Morphine (+)4.325 (2.548, 6.101)< 0.0010.0034.858 (2.976, 6.740)< 0.0010.0034.182 (2.492, 5.872)< 0.0010.0034.354 (2.535, 6.173)< 0.0010.003 CCB (-), Morphine (+)0.394 (-0.780, 1.567)0.5110.560.532 (-0.718, 1.781)0.4040.4760.475 (-0.642, 1.592)0.4050.4760.098 (-1.132, 1.328)0.8760.891 CCB (-), Morphine (-)-1.718 (-2.321, -1.114)< 0.0010.003-1.794 (-2.429, -1.159)< 0.0010.003-1.331 (-1.898, -0.763)< 0.0010.003-0.888 (-1.525, -0.251)0.0060.013CCB and Oxycodone CCB (+), Oxycodone (-)Reference-Reference-Reference-Reference- CCB (+), Oxycodone (+)2.629 (1.213, 4.044)< 0.0010.0033.109 (1.644, 4.574)< 0.0010.0032.247 (0.897, 3.596)0.0010.0032.418 (0.943, 3.893)0.0010.003 CCB (-), Oxycodone (+)-0.979 (-1.848, -0.109)0.0270.051-0.899 (-1.805, 0.006)0.0520.084-0.639 (-1.467, 0.189)0.1300.186-0.415 (-1.340, 0.511)0.3800.473 CCB (-), Oxycodone (-)-1.671 (-2.303, -1.040)< 0.0010.003-1.709 (-2.378, -1.040)< 0.0010.003-1.348 (-1.942, -0.755)< 0.0010.003-0.939 (-1.600, -0.278)0.0050.012CCB and Tramadol CCB (+), Tramadol (-)Reference-Reference-Reference-Reference- CCB (+), Tramadol (+)2.672 (1.660, 3.684)< 0.0010.0032.879 (1.805, 3.952)< 0.0010.0032.704 (1.751, 3.656)< 0.0010.0032.382 (1.340, 3.427)< 0.0010.003 CCB (-), Tramadol (+)-0.385 (-1.207, 0.437)0.3580.457-0.357 (-1.219, 0.505)0.4170.4810.102 (-0.663, 0.868)0.7930.8350.307 (-0.562, 1.176)0.4890.554 CCB (-), Tramadol (-)-0.856 (-1.653, -0.059)0.0350.06-0.962 (-1.794, -0.129)0.0240.046-0.607 (-1.348, 0.133)0.1080.158-0.243 (-1.092, 0.607)0.5760.617Model 1 adjusted for admission type, type of surgery, subjectivity, heart rateModel 2 adjusted for race, marital status, drinkingModel 3 adjusted for SOFA, CCI, MI, anemia, heart failure, CVDModel 4 adjusted for anion gap, glucose, creatinine, platelets, calcium, magnesium, RBC, RDW, WBC, monocytesAbbreviations: CCB Calcium channel blockers, SOFA Sequential organ failure assessment, CCI Charlson Comorbidity Index, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, MI Myocardial infarction, CVD Cerebrovascular disease, LOS Length of stay Generalized linear regression analysis of drug combinations and LOS Model 1 adjusted for admission type, type of surgery, subjectivity, heart rate Model 2 adjusted for race, marital status, drinking Model 3 adjusted for SOFA, CCI, MI, anemia, heart failure, CVD Model 4 adjusted for anion gap, glucose, creatinine, platelets, calcium, magnesium, RBC, RDW, WBC, monocytes Abbreviations: CCB Calcium channel blockers, SOFA Sequential organ failure assessment, CCI Charlson Comorbidity Index, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, MI Myocardial infarction, CVD Cerebrovascular disease, LOS Length of stay We further conducted supplementary analyses by redefining the reference group as “opioid use only, no CCB use” to evaluate the association of adding CCB to those three opioid therapies with LOS (Supplementary Table 3). The results indicated that, when using “use of opioid alone” as the reference, concomitant CCB use was strongly associated with a significantly prolonged hospital stay across all three opioids, and this association remained highly consistent across all four adjustment models. The combination of CCB with morphine showed the greatest effect on LOS prolongation (Model 4: β = 4.256, 95% CI [2.206, 6.306], P < 0.001), followed by oxycodone (Model 4: β = 2.833, 95% CI [1.289, 4.376], P < 0.001). We also conducted a multivariate logistic regression analysis to evaluate the association between the combination of CCB and those three opioid drugs and the risk of prolonged hospital stay (Table 3). Using “CCB use without opioid” as the reference group, the results showed that the combination of CCB and oxycodone also exhibited a robust association with prolonged hospitalization, with consistently significant effects across Models 1, 2, and 3, and the strongest effect in Model 3 (OR = 2.059, 95% CI [1.113, 3.808], P = 0.021). Similarly, the co-administration of CCB and morphine significantly increased the risk of prolonged stay after adjusting for demographic characteristics (Model 2) and clinical severity (Model 3). In contrast, the combination of CCB and tramadol reached statistical significance only in Model 2, following adjustment for demographic factors. Similarly, we changed the reference group to “opioid use only” and conducted the logistic regression analysis again (Supplementary Table 4). The results showed that, except for morphine, the additional use of CCB among patients using oxycodone or tramadol was strongly associated with a significantly increased risk of prolonged LOS, and this finding was very robust in all four adjusted models. Similarly, after applying the FDR correction for multiple testing, the robust associations between the combinations of CCB with oxycodone and prolonged hospitalization remained statistically significant across the adjusted logistic regression models. Table 3Multivariable logistic regression analysis of the association between drug combinations and LOSModel 1Model 2Model 3Model 4OR (95% CI)P-valueq-valueOR (95% CI)P-valueq-valueOR (95% CI)P-valueq-valueOR (95% CI)P-valueq-valueCCB and Morphine CCB (+), Morphine (-)Reference-Reference-Reference-Reference- CCB (+), Morphine (+)2.002 (0.921, 4.350)0.0800.1112.744 (1.224, 6.149)0.0140.0252.253 (1.021, 4.974)0.0440.0661.748 (0.724, 4.220)0.2140.249 CCB (-), Morphine (+)0.894 (0.575, 1.390)0.6190.6370.848 (0.545, 1.319)0.4640.5060.904 (0.575, 1.420)0.6610.6610.869 (0.514, 1.467)0.5980.633 CCB (-), Morphine (-)0.440 (0.350, 0.552)< 0.0010.0030.415 (0.330, 0.522)< 0.0010.0030.463 (0.368, 0.581)< 0.0010.0030.614 (0.469, 0.805)< 0.0010.003CCB and Oxycodone CCB (+), Oxycodone (-)Reference-Reference-Reference-Reference- CCB (+), Oxycodone (+)1.884 (1.029, 3.450)0.0400.0632.104 (1.168, 3.790)0.0130.0252.059 (1.113, 3.808)0.0210.0361.775 (0.856, 3.679)0.1230.161 CCB (-), Oxycodone (+)0.619 (0.448, 0.856)0.0040.0090.600 (0.435, 0.826)0.0020.0050.659 (0.473, 0.918)0.0140.0250.776 (0.526, 1.146)0.2030.249 CCB (-), Oxycodone (-)0.451 (0.356, 0.571)< 0.0010.0030.422 (0.332, 0.536)< 0.0010.0030.472 (0.373, 0.598)< 0.0010.0030.623 (0.472, 0.824)0.0010.003CCB and Tramadol CCB (+), Tramadol (-)Reference-Reference-Reference-Reference- CCB (+), Tramadol (+)1.353 (0.920, 1.991)0.1250.1611.506 (1.019, 2.224)0.0400.0631.416 (0.959, 2.090)0.0800.1111.261 (0.801, 1.986)0.3170.357 CCB (-), Tramadol (+)0.564 (0.416, 0.766)< 0.0010.0030.550 (0.405, 0.747)< 0.0010.0030.617 (0.455, 0.837)0.0020.0050.790 (0.545, 1.144)0.2120.249 CCB (-), Tramadol (-)0.465 (0.345, 0.626)< 0.0010.0030.447 (0.332, 0.602)< 0.0010.0030.474 (0.353, 0.638)< 0.0010.0030.594 (0.414, 0.853)0.0050.011Model 1 adjusted for admission type, type of surgery, subjectivity, heart rateModel 2 adjusted for race, marital status, drinkingModel 3 adjusted for SOFA, CCI, MI, anemia, heart failure, CVDModel 4 adjusted for anion gap, glucose, creatinine, platelets, calcium, magnesium, RBC, RDW, WBC, monocytesAbbreviations: CCB Calcium channel blockers, SOFA Sequential organ failure assessment, CCI Charlson Comorbidity Index, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, MI Myocardial infarction, CVD Cerebrovascular disease, LOS Length of stay Multivariable logistic regression analysis of the association between drug combinations and LOS Model 1 adjusted for admission type, type of surgery, subjectivity, heart rate Model 2 adjusted for race, marital status, drinking Model 3 adjusted for SOFA, CCI, MI, anemia, heart failure, CVD Model 4 adjusted for anion gap, glucose, creatinine, platelets, calcium, magnesium, RBC, RDW, WBC, monocytes Abbreviations: CCB Calcium channel blockers, SOFA Sequential organ failure assessment, CCI Charlson Comorbidity Index, RBC Red blood cell, RDW Red cell distribution width, WBC White blood cell, MI Myocardial infarction, CVD Cerebrovascular disease, LOS Length of stay Sensitivity analysis using the E-value demonstrated that for the CCB and oxycodone combination, an unmeasured confounder would need to be associated with both the treatment and prolonged LOS by a risk ratio of at least 2.09, 2.26, and 2.22—depending on the specific model—to explain away the observed association. This suggests our findings are relatively robust against unmeasured confounding. Further exploratory analysis Based on all previous analyses, we found that the combination of CCB and oxycodone was consistently and significantly associated with prolonged length of stay (LOS) across both generalized linear regression and logistic regression models, irrespective of whether the reference group was defined as CCB use alone or opioid analgesic use alone. This association remained robust in the vast majority of adjusted models. Therefore, oxycodone may represent a critical anesthetic agent warranting closer clinical attention in patients with hypertension following cardiac surgery. To further investigate the clinical nuances of this association, we conducted exploratory analyses to identify potential effect modifiers rather than mechanistic validation. While we initially hypothesized that this drug combination might induce more adverse events, thereby leading to a higher incidence of perioperative complications. However, this hypothesis was not supported by our analysis, which revealed no significant difference in the distribution of perioperative complications among the groups (P = 0.239). Subsequently, we performed a subgroup analysis to identify whether this association was more pronounced in specific susceptible populations (Table 4). The cohort was stratified by several key clinical variables, including the presence of perioperative complications, type of surgery, CCI score, and age. The results indicated that the association between the drug combination and prolonged LOS remained significant regardless of the presence of complications or the age stratum (all subgroup P < 0.05), suggesting these were not effect modifiers. In contrast, the type of surgery and CCI score significantly altered the strength of this association. The highly significant association was predominantly observed in the subgroup of patients who underwent valve surgery (β = 6.243, 95% CI [3.631, 8.856], P < 0.001) and in those with a CCI score > 5 (β = 6.852, 95% CI [3.870, 9.834], P < 0.001). To confirm this, we conducted formal interaction analyses for these two stratifying variables, both of which demonstrated a significant interaction effect (all P for interaction < 0.001). It is important to note that some subgroups (e.g., specific surgical types or complication strata) exhibited extremely large effect sizes and wide confidence intervals, which are likely attributable to the relatively small sample sizes (n) within those specific strata. Therefore, these estimates should be interpreted with caution. Table 4Subgroup analysis of the association between CCB-Oxycodone combinations and LOSSubgroup N Combined groupβ (95% CI)P-valueq-valuePerioperative complications No1958CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)2.774 (1.314, 4.233)< 0.0010.003CCB (-), Oxycodone (+)-0.726 (-1.630, 0.178)0.1160.174CCB (-), Oxycodone (-)-1.483 (-2.126, -0.839)< 0.0010.003 Yes135CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)11.685 (4.471, 18.899)0.0020.004CCB (-), Oxycodone (+)-4.287 (-8.135, -0.440)0.0310.050CCB (-), Oxycodone (-)-3.572 (-6.709, -0.434)0.0270.046Type of surgery CABG1186CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)0.765 (-0.763, 2.293)0.3270.374CCB (-), Oxycodone (+)-1.440 (-2.356, -0.524)0.0020.004CCB (-), Oxycodone (-)-2.274 (-2.940, -1.608)< 0.0010.003 Valve907CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)6.243 (3.631, 8.856)< 0.0010.003CCB (-), Oxycodone (+)-0.325 (-1.967, 1.317)0.6980.728CCB (-), Oxycodone (-)-0.722 (-1.891, 0.447)0.2260.271Age ≤ 68 years1104CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)3.147 (1.102, 5.192)0.0030.006CCB (-), Oxycodone (+)-0.846 (-2.006, 0.313)0.1530.215CCB (-), Oxycodone (-)-1.575 (-2.443, -0.708)< 0.0010.003 > 68 years989CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)3.346 (1.294, 5.397)0.0010.003CCB (-), Oxycodone (+)-0.975 (-2.337, 0.387)0.1610.215CCB (-), Oxycodone (-)-1.536 (-2.474, -0.597)0.0010.003CCI ≤ 5558CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)-0.464 (-1.903, 0.975)0.5280.576CCB (-), Oxycodone (+)-1.020 (-1.796, -0.245)0.0100.018CCB (-), Oxycodone (-)-1.599 (-2.169, -1.029)< 0.0010.003 > 5431CCB (+) Oxycodone (-).Reference-CCB (+), Oxycodone (+)6.852 (3.870, 9.834)< 0.0010.003CCB (-), Oxycodone (+)-0.117 (-2.377, 2.143)0.9190.919CCB (-), Oxycodone (-)-0.982 (-2.530, 0.566)0.2140.270Abbreviations: LOS Length of stay, CCB Calcium channel blockers, CABG Coronary artery bypass grafting, CCI Charlson Comorbidity Index Subgroup analysis of the association between CCB-Oxycodone combinations and LOS Abbreviations: LOS Length of stay, CCB Calcium channel blockers, CABG Coronary artery bypass grafting, CCI Charlson Comorbidity Index Discussion In this retrospective study of hypertensive patients undergoing cardiac surgery, we sought to elucidate the differential effects of concomitant CCB and opioid analgesic use on LOS. Our primary finding demonstrates that the co-administration of CCB with several opioids is associated with a significant prolongation of LOS. Notably, the combination of CCB and oxycodone exhibited the most robust association with extended hospitalization, whereas hydromorphone was the only opioid for which no such association was observed when used concurrently with CCB. Furthermore, subgroup and interaction analyses revealed that the association of CCB–oxycodone with LOS was significantly amplified among patients undergoing valve surgery or those with a CCI > 5. These findings underscore the critical importance of adopting individualized analgesic strategies in the perioperative management of hypertensive patients undergoing cardiac surgery. Our research findings fill an important knowledge gap in a key clinical area. Although risk factors for prolonged hospital stay, such as advanced age, surgical complexity, and multiple comorbidities, have been well established [18–23], studies on the impact of specific drug combinations on postoperative outcomes after cardiac surgery, particularly in the large patient population with hypertension, are relatively scarce. Although the pharmacokinetic interaction between CCB and opioid based on the CYP450 enzyme system is known in pharmacology [24, 25], this study is the first to confirm the clinical relevance of this interaction in a large real-world cohort and to reveal significant differences among different opioids. Our findings identify inappropriate choice of analgesic drugs as a new, potentially modifiable risk factor, providing a new perspective for optimizing the management of patients after cardiac surgery. The differential effects observed among different opioid drugs in this study are most likely rooted in their respective metabolic pathways. Oxycodone is mainly metabolized by CYP3A4 and CYP2D6, while CCB (particularly dihydropyridines) are well-known inhibitors of CYP3A4 [26–28]. The inhibitory effect of CCB on this enzyme slows down the clearance of oxycodone, leading to an increase in its blood concentration and thereby raising the risk of adverse events such as excessive sedation, respiratory depression, postoperative ileus, and delirium, all of which are known drivers of prolonged hospital stays [29–31]. In sharp contrast, hydromorphone is mainly metabolized by glucuronidation in the liver, a process that does not rely on the CYP450 enzyme system [32–34]. Therefore, its metabolic process is not affected by CCB, avoiding the risk of drug accumulation, which explains why no prolongation of hospital stay was observed when it was used in combination with CCB. This striking contrast provides strong support for the hypothesis that CYP450-mediated drug interactions are the core mechanism leading to adverse outcomes. However, it is crucial to emphasize that while this CYP3A4-based explanation is biologically plausible, our retrospective database analysis cannot directly verify this pharmacokinetic mechanism. Our subgroup findings (e.g., the potentiated effects in valve surgery and high CCI patients) merely generate hypotheses regarding clinically vulnerable populations. Future formal pharmacokinetic studies measuring systemic drug concentrations are required to confirm this proposed mechanism. The findings of this study have significant clinical practice implications. For hypertensive patients receiving CCB treatment, when choosing an analgesic regimen after cardiac surgery, hydromorphone should be considered as a potentially safer option. Our subgroup analysis further identified the high-risk groups that require the most vigilance. Patients undergoing valve surgery typically experience more complex and longer surgeries, with lower physiological reserves, and thus may be more sensitive to adverse reactions caused by elevated drug concentrations [35–38]. Similarly, patients with high CCI scores, due to their heavy burden of comorbidities, may have impaired intrinsic drug clearance capabilities, making them particularly vulnerable to the consequences of drug interactions [39–41]. Therefore, for these two groups of patients, avoiding the use of opioids such as oxycodone that are metabolized by CYP3A4 may be a simple yet effective strategy to optimize perioperative management and facilitate more individualized and safer pain treatment. When interpreting the findings of this study, several limitations must be acknowledged. First, as an observational study, we cannot fully establish causality or completely rule out residual confounding. While our multivariable models and LASSO selection rigorously adjusted for known variables, critical unmeasured confounders such as specific intraoperative anesthetic regimens, exact duration/dose of opioid exposure, and dynamic pain severity scores were unavailable. Patients receiving strong opioids like oxycodone or morphine might inherently have greater surgical trauma or pain, leading to prolonged LOS (confounding by indication). Although our E-value sensitivity analysis suggests that a single unmeasured confounder would need a substantial effect to negate our findings, the results must still be interpreted as robust associations rather than definitive causal effects. Second, the data were collected from a single center (Beth Israel Deaconess Medical Center), intrinsically reflecting specific institutional practices regarding surgery, analgesia protocols, and discharge criteria. Furthermore, because the MIMIC-IV database is fundamentally ICU-focused, the exclusion of non-ICU patients inherently introduces selection bias toward a sicker cohort with higher baseline acuity. These factors potentially restrict the generalizability of our results to other healthcare settings, diverse patient populations, or less complex surgical pathways. External validation in prospective, multicenter, and non-ICU datasets is an important future direction. Third, a central analytical challenge lies in isolating the independent effect of any individual opioid within the complex intensive care environment, where patients commonly receive multimodal analgesia or sedation regimens involving multiple agents. Importantly, however, this limitation is non-differential, as it uniformly affects all opioid groups under comparison. That is, all opioids are evaluated within a comparable context of polypharmacy. Therefore, despite the potential for unmeasured drug interactions, the observed differences in effect sizes across opioid classes are likely robust and clinically meaningful. Fourth, the exclusion of patients who died in the hospital (n = 26) may have introduced survivor bias. Because in-hospital mortality inherently truncates the length of stay, these patients typically exhibit the shortest hospital stays but represent the most severe clinical outcomes. While their exclusion was methodologically necessary to prevent death from artificially skewing the ‘short LOS’ group, it inherently removes the sickest patients from the cohort. Consequently, this survivor bias might lead to an underestimation of the true adverse effects and clinical severity associated with certain drug combinations. Finally, exposure to CCB and opioids was operationalized purely as binary variables (i.e., used vs. not used). This simplification is a significant limitation, as it does not account for the exact dosage, route of administration (e.g., intravenous versus oral), duration of therapy, or cumulative drug exposure. Consequently, patients receiving a single, transient postoperative dose were grouped with those on sustained, high-dose therapy. This lack of pharmacological precision may lead to exposure misclassification and precludes an assessment of dose-response relationships or differences between short-term and long-term combination use. Future studies should prioritize multicenter prospective cohort designs to confirm the reproducibility of these findings and explore the feasibility of randomized controlled trials with detailed pharmacological data collection, enabling direct comparisons of different opioids on clinical outcomes among hypertensive patients receiving CCB, thereby strengthening causal inference. Conclusion For hypertensive patients undergoing cardiac surgery, co-administering CCB with oxycodone is significantly associated with prolonged hospitalization, an association that is stronger in patients undergoing valve surgery or with high comorbidity burdens. These findings highlight the critical need for personalized analgesic selection in this high-risk population. Supplementary Information Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Acknowledgements Not applicable. Clinical trial number Not applicable. Authors’ contributions XLL and HTL contributed to the conception and design. XLL and QL contributed to the collection and assembly of data. HTL and YHZ analyzed and interpreted the data. All authors wrote and approved the final manuscript. Funding No funding was received for this research. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The Ethics Committee of Longyan First Affiliated Hospital of Fujian Medical University deemed that this research is based on open-source data, so the need for ethics approval was waived and no informed consent required. The study was performed in accordance with the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xinglu Li and Haitao Liu contributed equally to this work and should be regarded as co-first author. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.