ABSTRACT
Purpose
This study aimed to evaluate the association between ramelteon use and mortality outcomes in patients with sepsis and cancer. The research question was whether ramelteon use is associated with reduced mortality in this high‐risk patient population.
Methods
A retrospective analysis of the MIMIC‐IV database was conducted. We included 3283 patients with sepsis and cancer, with 424 receiving ramelteon and 2859 not receiving ramelteon. Missing data (< 20%) were imputed using random forest. Propensity score matching (PSM) was used to balance baseline characteristics between groups (1:1 matching, caliper width of 0.1), resulting in 400 patients in each group. Kaplan–Meier curves were used to compare survival probabilities, and Cox regression and logistic regression analyses were performed to assess the association between ramelteon and mortality outcomes.
Results
Ramelteon use was significantly associated with lower 30‐day (HR: 0.37, 95% CI: 0.28–0.50, p < 0.001), 90‐day (HR: 0.39, 95% CI: 0.29–0.52, p < 0.001), ICU (OR: 0.29, 95% CI: 0.18–0.46, p < 0.001), and in‐hospital mortality (OR: 0.50, 95% CI: 0.35–0.72, p < 0.001) in the fully adjusted model after PSM. In exploratory duration‐related analyses, ramelteon use for ≤ 7 days was associated with lower mortality, whereas no additional significant association was observed with longer duration.
Conclusion
Ramelteon use was associated with lower mortality among patients with sepsis and cancer. However, given the observational design and potential residual confounding, these findings should be interpreted as hypothesis‐generating and require confirmation in prospective studies.
Keywords: cancer, mortality, propensity score matching, ramelteon, sepsis, survival analysis
1. Introduction
Sepsis often leads to inflammatory damage across multiple organ systems. The co‐occurrence of cancer and sepsis poses a significant clinical challenge, often leading to poor patient outcomes [1]. Factors such as compromised immunity, invasive treatments, and prolonged hospital stays predispose cancer patients to a higher risk of developing sepsis, which, in turn, significantly increases their mortality risk [2, 3]. Although previous studies have demonstrated that early recognition and standardized treatment protocols can improve outcomes for sepsis patients, effective intervention strategies specifically targeting sepsis in cancer patients remain limited.
Melatonin, an endogenous hormone, plays a crucial role in regulating circadian rhythms. Ramelteon, a synthetic melatonin receptor agonist with high affinity for Melatonin Receptor (MT)‐1 and MT2 receptors, can mimic the functions of endogenous melatonin. Clinically, ramelteon is recommended for treating sleep disorders in hospitalized and critically ill patients at a typical dose of 8 mg, with a longer half‐life than melatonin [4]. While existing literature on ramelteon has predominantly concentrated on its utility in the context of sleep‐related ailments, its potential value in cancer‐related fields has garnered increasing attention in recent years. Existing evidence indicates that ramelteon may play a role in cancer‐related neuroendocrine disorders, anti‐tumor activity, and the management of postoperative complications [5, 6]. For instance, studies suggest that ramelteon suppresses the proliferation and invasiveness of endometrial cancer cells [7]. Furthermore, ramelteon has shown potential in reducing the incidence and severity of delirium in elderly patients following lung or esophageal cancer surgery [8, 9]. The existing research also suggests that ramelteon may exert anti‐inflammatory, antioxidant, and antibacterial effects [10]. Animal experiments have demonstrated that ramelteon can improve survival in a model of polymicrobial sepsis and may exert antiviral effects against Severe Acute Respiratory Syndrome Coronavirus 2 infection in mice [11, 12]. Previous observational studies also suggest that ramelteon may be associated with lower mortality in sepsis patients, suggesting its potential prognostic relevance in this population [13, 14].
However, the direct studies on ramelteon in cancer patients with sepsis were limited. Therefore, this study aimed to evaluate the association between ramelteon use and mortality in patients with both sepsis and cancer, using a propensity score‐matched analysis of the MIMIC‐IV database to provide new insights to inform future research on clinical outcomes for this vulnerable population. We hypothesize that ramelteon use will be associated with lower mortality. The results of this study may provide preliminary evidence to inform future prospective studies evaluating ramelteon in this population.
2. Methods
2.1. Study Design and Population
This retrospective study was conducted using the Medical Information Mart for Intensive Care IV database (MIMIC‐IV 3.0) [15, 16]. Authorization for this study was granted by the Institutional Review Boards (IRB) of Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC), which waived the requirement for informed consent. Access to the database was authorized for a single researcher (Mingjie Liu), who completed the Collaborative Institutional Training Initiative (CITI) program (Record ID: 66311890). All cases were diagnosed according to the International Classification of Diseases, 9th and 10th Revisions (ICD‐9 and ICD‐10) coding systems. Inclusion criteria were as follows: (1) age ≥ 18 years; (2) first ICU admission; (3) diagnosis of sepsis according to the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis‐3) [17]; and (4) presence of a documented malignancy. Presence of malignancy was determined using ICD‐9 and ICD‐10 diagnosis codes recorded during the index hospitalization. Malignancy was defined as any documented malignant tumor, including lymphoma and leukemia, while excluding non‐melanoma malignant neoplasms of the skin. The overall malignancy definition included ICD‐9 code prefixes 2386, 140–172, 1740–1958, and 200–208, and ICD‐10 code prefixes C43, C88, C00–C26, C30–C34, C37–C41, C45–C58, C60–C76, C81–C85, and C90–C97. The detailed ICD code groupings used for malignancy ascertainment and subtype classification are provided in Table S3. Exclusion criteria were as follows: (1) ICU stay of < 48 h; and (2) use of melatonin analogs other than ramelteon. The flowchart of sample selection is detailed in Figure 1. Patients were stratified into a ramelteon cohort and a non‐ramelteon cohort, according to whether they were administered ramelteon (8 mg) during their hospital stay, irrespective of the route of administration (oral or via gastric tube) [18]. For all survival analyses, time zero was defined as the time of hospital admission in MIMIC‐IV. The same time origin was applied consistently for the 30‐day and 90‐day mortality endpoints. For patients in the ramelteon group, exposure initiation was defined as the earliest recorded ramelteon prescription start time during the index hospitalization. The interval from hospital admission to first ramelteon prescription was calculated to describe exposure timing, as summarized in Table S4.
FIGURE 1.
Flowchart of sample selection. A total of 3283 eligible patients with sepsis and cancer were included in the unmatched cohort, including 424 ramelteon users and 2859 non‐users. After 1:1 propensity score matching, 800 patients were included, with 400 patients in each group.
2.2. Derivation of Ramelteon Treatment Duration and Cumulative Dose
All ramelteon prescription records during the index hospitalization were extracted. Prescription start and stop timestamps were standardized to a common datetime format. Records with missing start or stop timestamps were not used for duration or dose derivation. For prescription segments with inconsistent timestamps, defined as a start time later than the stop time, a prespecified timestamp correction was applied by swapping the start and stop times. Segment‐level treatment duration was calculated as the interval between stop and start times in days and rounded up to an integer number of days. Same‐day or zero‐length documented prescription segments were assigned a duration of 1 day to avoid classifying a recorded prescription as no exposure.
The segment‐level ramelteon dose was derived from available prescription dose fields, including dose value, dose unit, dispensed form, and product strength. Because ramelteon prescriptions were predominantly 8‐mg tablets, tablet strength was used when dose fields were incomplete. Cumulative dose was calculated as the sum of segment‐level dose multiplied by segment duration across all prescription segments during the hospitalization. Total treatment days were calculated as the sum of segment durations, and average daily dose was calculated as cumulative dose divided by total treatment days.
2.3. Data Extraction
We extracted data regarding ramelteon administration and mortality outcomes during the patients' hospitalization. The primary outcome of this study was the 30‐day mortality, while secondary outcomes included 90‐day mortality, in‐hospital mortality, and ICU mortality. Furthermore, we extracted a comprehensive set of clinical data from the database, encompassing key domains such as demographic characteristics, vital signs, clinical severity scores, laboratory parameters, comorbidities, and therapeutic interventions. Demographic characteristics included age, weight, and race (categorized as Asian/Hispanic or Latino/other; White; Black). Vital signs included heart rate, blood pressure parameters such as systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP). Additionally, respiratory rate (RR), body temperature (T), and peripheral capillary oxygen saturation (SPO2) were recorded. Clinical severity was assessed using the Sequential Organ Failure Assessment (SOFA) score, Glasgow Coma Scale (GCS), Charlson Comorbidity Index (CCI), and Simplified Acute Physiology Score II (SAPS II). Laboratory parameters included prothrombin time (PT) and partial thromboplastin time (PTT). Maximum values of vital signs and laboratory parameters were extracted from the first day of ICU admission. Comorbidities assessed included heart failure, atrial fibrillation, cerebrovascular disease, peripheral vascular disease, rheumatic disease, peptic ulcer disease, chronic kidney disease (CKD), acute kidney injury (AKI), diabetes mellitus, hypertension, chronic obstructive pulmonary disease (COPD), metastatic solid tumor, moderate to severe liver disease, dementia, and paraplegia/hemiplegia. Oncologic status was characterized using ICD‐based malignancy subtype and metastatic solid tumor status. Malignancy subtype was further classified as solid tumor, hematologic tumor, or both according to ICD code groupings. Solid tumors were identified using ICD‐9 code prefixes 140–172 and 1740–1958, and ICD‐10 code prefixes C00–C26, C30–C34, C37–C41, C43, C45–C58, and C60–C76. Hematologic tumors were identified using ICD‐9 code prefixes 200–208 and 2386, and ICD‐10 code prefixes C81–C85, C88, and C90–C96. Patients with both solid tumor and hematologic tumor codes were classified as having both. ICD‐10 code C97 was considered part of the overall malignancy definition as multiple primary/other unspecified malignancy, but was not assigned to either the solid tumor or hematologic tumor subtype. Metastatic solid tumor was defined separately using ICD‐9 code prefixes 196, 197, 198, and 199, and ICD‐10 code prefixes C77, C78, C79, and C80. Therapeutic interventions recorded included continuous renal replacement therapy (CRRT), mechanical ventilation, vasoactive agent, angiotensin‐converting enzyme inhibitors (ACEI)/angiotensin II receptor blockers (ARB), NSAIDs, opioids, benzodiazepines, propofol, and the administration of various antibiotic classes: aminoglycosides, beta‐lactams, macrolides, glycopeptides, sulfonamides, tetracyclines, quinolones, and other antibiotics. Available indicators of sepsis severity included vital signs, laboratory parameters, SOFA score, SAPS II, GCS, AKI stage, mechanical ventilation, CRRT, and vasoactive agent use. Vasoactive treatment was captured as a binary exposure variable indicating whether vasoactive agents were used during the relevant hospitalization period.
2.4. Statistical Analysis
For variables with < 20% missing data, we employed the random forest imputation method (using the missForest package) to impute missing values [19] (Table S1). To assess potential multicollinearity among variables, we used the variance inflation factor (VIF) for diagnosis, with VIF values > 10 considered indicative of significant multicollinearity [20] (Table S2). Variables with VIF > 10 were excluded from the regression analysis models. We used propensity score matching (PSM) to reduce imbalance in measured baseline covariates between the ramelteon and non‐ramelteon groups [21]. Propensity scores were estimated using the prespecified baseline covariates, and 1:1 nearest‐neighbor matching was performed with a caliper width of 0.1. The caliper width was selected a priori as a relatively stringent matching criterion to reduce the likelihood of poor matches while preserving an adequate number of ramelteon‐exposed patients for analysis. This approach was intended to balance matching quality and sample retention. Covariate balance before and after matching was assessed using standardized mean differences (SMDs). In descriptive statistics, continuous variables are presented as median (with interquartile range Q1 and Q3), while categorical variables are presented as percentages. We used Kaplan–Meier curves to compare the differences in survival probability between the ramelteon and non‐ramelteon groups. Furthermore, Cox regression analysis and logistic regression analysis were performed for different outcomes, and the results are presented as hazard ratios (HR) with 95% confidence intervals (CI) and odds ratios (OR) with 95% CI, respectively. The proportional hazards assumption for Cox regression models was formally assessed using Schoenfeld residuals. Both global and covariate‐specific diagnostics were examined. When evidence suggested that the proportional hazards assumption was not fully satisfied, hazard ratios were interpreted as average associations over the follow‐up period rather than as constant hazard ratios across all time points [22]. In the survival analyses, follow‐up time for 30‐day and 90‐day mortality was calculated from hospital admission, which served as the common time zero for both Kaplan–Meier curves and Cox regression models. Prior to multivariate analysis, a univariable Cox regression analysis was performed using the primary outcome of 30‐day mortality to screen variables, and variables with a p‐value < 0.05 were further included in the multivariate model. We sequentially constructed three models for analysis: Model 1 adjusted for demographic parameters, anthropometric data, vital signs, and laboratory data. Model 2 added comorbidities and severity scores to Model 1. Model 3 further included clinical interventions to Model 2. Duration‐ and dose‐related analyses were performed as exploratory analyses. Because treatment duration may be influenced by survival time, these analyses were interpreted cautiously and were not considered confirmatory evidence of a causal duration threshold. Subgroup analyses were performed in the propensity score‐matched cohort to evaluate whether the association between ramelteon use and 30‐day mortality differed across clinically relevant strata. HRs in the subgroup analyses were estimated using Cox regression models adjusted for the same covariates as Model 3, except for the variable used for stratification, which was omitted from the corresponding model. Interaction p‐values were calculated to assess potential effect modification across subgroups. Data processing and analysis were performed using R version 4.4.0 (2024‐04‐24), along with Zstats 1.0 (www.zstats.net). All statistical tests were two‐sided, and p < 0.05 was considered statistically significant.
3. Results
3.1. Baseline Characteristics
In this study, we included 3283 patients with sepsis and cancer. Before PSM, the median age of the overall sample was 69.34 years (Q1, Q3: 60.96, 78.35). The median age of patients not receiving Ramelteon was 68.74 years (Q1, Q3: 60.50, 77.93), while those receiving Ramelteon had a median age of 73.02 years (Q1, Q3: 65.24, 80.06), showing a significant difference (p < 0.001). Regarding gender distribution, females accounted for 37.8% (n = 1241) and males for 62.2% (n = 2042), with no significant differences between groups (p = 0.853). In laboratory tests, the median hemoglobin level was 9.70 g/dL in the non‐ramelteon group and 9.40 g/dL in the ramelteon group before PSM (p = 0.009). Serum creatinine levels were higher in the non‐ramelteon group at 1.30 mg/dL compared to 1.10 mg/dL in the ramelteon group (p < 0.001). The median SOFA score was 3.00 (Q1, Q3: 2.00, 4.00), indicating a significantly higher clinical severity in the non‐ramelteon group. Regarding comorbidities before PSM, the incidence of atrial fibrillation was 32.35% (n = 1062), with heart failure at 27.38% (n = 899) and myocardial infarction at 14.62% (n = 480). These comorbidities were significantly higher in the non‐ramelteon group compared to the ramelteon group (p < 0.001 for all). After PSM, the overall sample size was reduced to 800 patients, with no significant differences across these parameters (Table 1). Figures 2 and 3 illustrate the effectiveness of PSM. Figure 2 shows the probability density plots of the Ramelteon and non‐ramelteon groups before and after PSM, indicating that the characteristics of the two groups were well matched after matching. Figure 3 highlights the SMDs before and after PSM, demonstrating a significant reduction in SMD values, which indicates improved balance between the groups.
TABLE 1.
Baseline characteristics of patients before PSM and after PSM.
| Variable | Before PSM | After PSM | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Total (n = 3283) | Non‐ramelteon (n = 2859) | Ramelteon (n = 424) | p | SMD | Total (n = 800) | Non‐ramelteon (n = 400) | Ramelteon (n = 400) | p | SMD | |
| Demographic parameters | ||||||||||
| Age (years), M (Q1, Q3) | 69.34 (60.96, 78.35) | 68.74 (60.50, 77.93) | 73.02 (65.24, 80.06) | < 0.001 | 0.261 | 72.79 (64.68, 80.51) | 72.83 (64.48, 80.67) | 72.69 (64.99, 80.12) | 0.879 | 0.006 |
| Sex, n (%) | 0.853 | 0.942 | ||||||||
| Female | 1241 (37.8) | 1079 (37.74) | 162 (38.21) | 0.01 | 313 (39.12) | 157 (39.25) | 156 (39.00) | −0.005 | ||
| Male | 2042 (62.2) | 1780 (62.26) | 262 (61.79) | −0.01 | 487 (60.88) | 243 (60.75) | 244 (61.00) | 0.005 | ||
| Race, n (%) | 0.431 | 0.898 | ||||||||
| Asian/Hispanic or Latino/other | 661 (20.13) | 566 (19.80) | 95 (22.41) | 0.063 | 178 (22.25) | 88 (22.00) | 90 (22.50) | 0.012 | ||
| Black | 351 (10.69) | 309 (10.81) | 42 (9.91) | −0.03 | 86 (10.75) | 45 (11.25) | 41 (10.25) | −0.033 | ||
| White | 2271 (69.17) | 1984 (69.39) | 287 (67.69) | −0.036 | 536 (67) | 267 (66.75) | 269 (67.25) | 0.011 | ||
| Weight (kg), M (Q1, Q3) | 76.50 (64.50, 91.00) | 76.55 (64.70, 91.10) | 76.10 (63.74, 88.82) | 0.27 | −0.088 | 76.00 (63.58, 90.30) | 75.50 (63.58, 90.81) | 76.65 (63.65, 89.12) | 0.761 | −0.005 |
| Cancer category, n (%) | < 0.001 | 0.96 | ||||||||
| Hematologic tumors | 422 (12.85) | 296 (10.35) | 126 (29.72) | 0.424 | 225 (28.12) | 113 (28.25) | 112 (28.00) | −0.006 | ||
| Solid tumors | 2828 (86.14) | 2538 (88.77) | 290 (68.40) | −0.438 | 562 (70.25) | 281 (70.25) | 281 (70.25) | 0 | ||
| Both | 33 (1.01) | 25 (0.87) | 8 (1.89) | 0.074 | 13 (1.62) | 6 (1.50) | 7 (1.75) | 0.019 | ||
| Vital signs | ||||||||||
| Heart rate (bpm), M (Q1, Q3) | 111.00 (96.00, 127.00) | 111.00 (96.00, 127.00) | 112.00 (96.00, 131.00) | 0.471 | 0.054 | 110.00 (96.00, 130.00) | 111.00 (96.00, 131.00) | 110.00 (96.00, 129.00) | 0.54 | −0.041 |
| SBP (mmHg), M (Q1, Q3) | 143.00 (130.00, 160.00) | 143.00 (130.00, 160.00) | 146.50 (131.00, 161.00) | 0.117 | 0.075 | 143.50 (130.00, 160.00) | 142.00 (129.75, 157.00) | 146.00 (130.38, 161.00) | 0.142 | 0.106 |
| DBP (mmHg), M (Q1, Q3) | 86.00 (75.00, 99.00) | 86.00 (74.00, 98.00) | 87.50 (77.00, 101.25) | 0.002 | 0.171 | 88.00 (77.00, 101.00) | 88.00 (77.00, 101.00) | 87.00 (77.00, 101.25) | 0.848 | 0.03 |
| MAP (mmHg), M (Q1, Q3) | 101.00 (90.00, 114.00) | 100.00 (89.60, 114.00) | 103.00 (92.00, 115.00) | 0.025 | 0.089 | 102.00 (91.00, 115.00) | 101.00 (90.00, 115.00) | 103.00 (92.00, 115.00) | 0.304 | 0.083 |
| RR (bpm), M (Q1, Q3) | 29.00 (25.00, 34.00) | 29.00 (25.00, 34.00) | 30.00 (26.00, 35.00) | 0.009 | 0.126 | 30.00 (26.00, 35.00) | 30.00 (26.00, 35.00) | 30.00 (26.00, 35.00) | 0.977 | −0.016 |
| T (°C), M (Q1, Q3) | 37.33 (37.00, 37.89) | 37.33 (37.00, 37.90) | 37.22 (37.00, 37.78) | 0.085 | −0.077 | 37.22 (36.94, 37.83) | 37.28 (36.94, 37.89) | 37.22 (37.00, 37.78) | 0.849 | 0.02 |
| SPO2 (%), M (Q1, Q3) | 100.00 (100.00, 100.00) | 100.00 (100.00, 100.00) | 100.00 (99.00, 100.00) | 0.618 | −0.062 | 100.00 (99.00, 100.00) | 100.00 (99.00, 100.00) | 100.00 (99.00, 100.00) | 0.055 | 0.098 |
| Laboratory | ||||||||||
| Hemoglobin (g/dL), M (Q1, Q3) | 9.70 (8.50, 10.90) | 9.70 (8.50, 10.90) | 9.40 (8.20, 10.80) | 0.009 | −0.111 | 9.40 (8.30, 10.70) | 9.40 (8.30, 10.60) | 9.40 (8.20, 10.80) | 0.803 | 0.043 |
| Platelet (103/μL), M (Q1, Q3) | 176.00 (108.00, 263.30) | 175.00 (109.00, 264.10) | 186.50 (106.00, 260.00) | 0.933 | −0.004 | 177.90 (105.00, 262.25) | 169.50 (102.90, 266.50) | 187.00 (106.00, 256.05) | 0.827 | −0.002 |
| RBC (106/μL), M (Q1, Q3) | 3.24 (2.85, 3.70) | 3.24 (2.85, 3.69) | 3.21 (2.79, 3.76) | 0.861 | 0.008 | 3.21 (2.81, 3.73) | 3.21 (2.82, 3.70) | 3.21 (2.79, 3.76) | 0.859 | 0.027 |
| RDW (%), M (Q1, Q3) | 16.10 (14.80, 17.90) | 16.10 (14.80, 17.81) | 16.10 (14.70, 17.92) | 0.929 | 0.011 | 16.10 (14.70, 18.00) | 16.11 (14.80, 18.10) | 16.10 (14.70, 17.90) | 0.588 | −0.04 |
| WBC (103/μL), M (Q1, Q3) | 12.20 (8.01, 17.90) | 12.06 (8.00, 17.70) | 12.73 (8.30, 19.22) | 0.056 | 0.1 | 12.60 (8.30, 18.42) | 12.68 (8.50, 18.00) | 12.47 (8.28, 19.12) | 0.77 | −0.005 |
| AG (mmol/L), M (Q1, Q3) | 14.00 (12.00, 17.00) | 14.00 (12.00, 17.00) | 14.00 (12.00, 17.00) | < 0.001 | −0.167 | 14.00 (12.00, 16.00) | 14.00 (12.00, 16.00) | 14.00 (12.00, 16.10) | 0.313 | −0.034 |
| Bicarbonate (mmol/L), M (Q1, Q3) | 23.00 (20.00, 25.80) | 23.00 (20.00, 26.00) | 22.00 (20.00, 25.00) | 0.125 | −0.038 | 22.70 (20.00, 25.00) | 23.00 (20.00, 25.40) | 22.10 (20.00, 25.00) | 0.66 | 0 |
| BUN (mg/dL), M (Q1, Q3) | 25.00 (16.00, 40.00) | 24.00 (16.00, 39.00) | 29.00 (18.30, 45.05) | < 0.001 | 0.188 | 28.00 (18.00, 45.00) | 26.10 (18.00, 45.25) | 28.60 (18.00, 44.00) | 0.504 | 0.001 |
| Chloride (mmol/L), M (Q1, Q3) | 104.00 (101.00, 108.00) | 104.80 (101.00, 108.00) | 103.00 (99.00, 107.00) | < 0.001 | −0.248 | 103.00 (99.00, 107.00) | 103.00 (99.00, 106.35) | 103.00 (99.00, 107.00) | 0.684 | 0.009 |
| Serum creatinine (mg/dL), M (Q1, Q3) | 1.10 (0.80, 1.80) | 1.10 (0.80, 1.80) | 1.30 (0.90, 2.10) | < 0.001 | 0.187 | 1.20 (0.88, 2.10) | 1.20 (0.80, 2.08) | 1.30 (0.90, 2.10) | 0.22 | 0.047 |
| Glucose (mg/dL), M (Q1, Q3) | 137.00 (112.90, 179.00) | 136.00 (112.00, 178.00) | 138.60 (114.75, 180.70) | 0.304 | 0 | 138.80 (114.00, 184.00) | 140.00 (114.00, 187.00) | 137.00 (114.00, 180.15) | 0.758 | −0.029 |
| Sodium (mmol/L), M (Q1, Q3) | 138.00 (135.10, 141.00) | 138.00 (135.60, 141.00) | 138.00 (135.00, 141.00) | 0.958 | 0.012 | 138.00 (135.00, 141.00) | 138.00 (135.00, 141.00) | 138.00 (135.00, 141.00) | 0.803 | −0.015 |
| Potassium (mmol/L), M (Q1, Q3) | 4.30 (3.90, 4.70) | 4.30 (3.90, 4.70) | 4.40 (4.09, 4.80) | < 0.001 | 0.219 | 4.40 (4.00, 4.80) | 4.40 (4.00, 4.90) | 4.40 (4.03, 4.80) | 0.575 | −0.068 |
| PT (s), M (Q1, Q3) | 15.70 (13.70, 19.30) | 15.60 (13.63, 19.20) | 15.90 (13.70, 19.88) | 0.323 | 0.021 | 15.90 (13.69, 19.60) | 15.90 (13.60, 19.48) | 15.90 (13.80, 19.88) | 0.739 | −0.044 |
| PTT (s), M (Q1, Q3) | 33.90 (29.20, 46.38) | 33.90 (29.20, 46.38) | 33.20 (28.20, 46.30) | 0.163 | 0.061 | 33.90 (28.70, 47.42) | 34.25 (28.83, 48.55) | 33.42 (28.20, 45.83) | 0.278 | −0.013 |
| Co‐morbidities | ||||||||||
| Atrial fibrillation, n (%) | < 0.001 | 0.564 | ||||||||
| No | 2221 (67.65) | 1978 (69.19) | 243 (57.31) | −0.24 | 478 (59.75) | 243 (60.75) | 235 (58.75) | −0.041 | ||
| Yes | 1062 (32.35) | 881 (30.81) | 181 (42.69) | 0.24 | 322 (40.25) | 157 (39.25) | 165 (41.25) | 0.041 | ||
| Heart failure, n (%) | < 0.001 | 0.345 | ||||||||
| No | 2384 (72.62) | 2123 (74.26) | 261 (61.56) | −0.261 | 491 (61.38) | 239 (59.75) | 252 (63.00) | 0.067 | ||
| Yes | 899 (27.38) | 736 (25.74) | 163 (38.44) | 0.261 | 309 (38.62) | 161 (40.25) | 148 (37.00) | −0.067 | ||
| Myocardial infarction, n (%) | 0.027 | 0.926 | ||||||||
| No | 2803 (85.38) | 2456 (85.90) | 347 (81.84) | −0.105 | 659 (82.38) | 330 (82.50) | 329 (82.25) | −0.007 | ||
| Yes | 480 (14.62) | 403 (14.10) | 77 (18.16) | 0.105 | 141 (17.62) | 70 (17.50) | 71 (17.75) | 0.007 | ||
| CKD, n (%) | < 0.001 | 0.698 | ||||||||
| No | 2525 (76.91) | 2228 (77.93) | 297 (70.05) | −0.172 | 565 (70.62) | 280 (70.00) | 285 (71.25) | 0.028 | ||
| Yes | 758 (23.09) | 631 (22.07) | 127 (29.95) | 0.172 | 235 (29.38) | 120 (30.00) | 115 (28.75) | −0.028 | ||
| AKI stage, n (%) | 0.038 | 0.945 | ||||||||
| No AKI | 354 (10.78) | 319 (11.16) | 35 (8.25) | −0.105 | 64 (8) | 30 (7.50) | 34 (8.50) | 0.036 | ||
| I | 499 (15.2) | 448 (15.67) | 51 (12.03) | −0.112 | 96 (12) | 47 (11.75) | 49 (12.25) | 0.015 | ||
| II | 1289 (39.26) | 1113 (38.93) | 176 (41.51) | 0.052 | 337 (42.12) | 169 (42.25) | 168 (42.00) | −0.005 | ||
| III | 1141 (34.75) | 979 (34.24) | 162 (38.21) | 0.082 | 303 (37.88) | 154 (38.50) | 149 (37.25) | −0.026 | ||
| Diabetes, n (%) | 0.211 | 0.449 | ||||||||
| No | 2345 (71.43) | 2053 (71.81) | 292 (68.87) | −0.064 | 542 (67.75) | 266 (66.50) | 276 (69.00) | 0.054 | ||
| Yes | 938 (28.57) | 806 (28.19) | 132 (31.13) | 0.064 | 258 (32.25) | 134 (33.50) | 124 (31.00) | −0.054 | ||
| Hypertension, n (%) | < 0.001 | 1 | ||||||||
| No | 2037 (62.05) | 1740 (60.86) | 297 (70.05) | 0.201 | 552 (69) | 276 (69.00) | 276 (69.00) | 0 | ||
| Yes | 1246 (37.95) | 1119 (39.14) | 127 (29.95) | −0.201 | 248 (31) | 124 (31.00) | 124 (31.00) | 0 | ||
| Cerebrovascular disease, n (%) | 0.068 | 0.609 | ||||||||
| No | 2898 (88.27) | 2535 (88.67) | 363 (85.61) | −0.087 | 689 (86.12) | 347 (86.75) | 342 (85.50) | −0.036 | ||
| Yes | 385 (11.73) | 324 (11.33) | 61 (14.39) | 0.087 | 111 (13.88) | 53 (13.25) | 58 (14.50) | 0.036 | ||
| Peripheral vascular disease, n (%) | 0.005 | 0.431 | ||||||||
| No | 3010 (91.68) | 2636 (92.20) | 374 (88.21) | −0.124 | 711 (88.88) | 359 (89.75) | 352 (88.00) | −0.054 | ||
| Yes | 273 (8.32) | 223 (7.80) | 50 (11.79) | 0.124 | 89 (11.12) | 41 (10.25) | 48 (12.00) | 0.054 | ||
| Rheumatic disease, n (%) | 0.754 | 0.57 | ||||||||
| No | 3174 (96.68) | 2763 (96.64) | 411 (96.93) | 0.017 | 771 (96.38) | 384 (96.00) | 387 (96.75) | 0.042 | ||
| Yes | 109 (3.32) | 96 (3.36) | 13 (3.07) | −0.017 | 29 (3.62) | 16 (4.00) | 13 (3.25) | −0.042 | ||
| Dementia, n (%) | 0.251 | 0.863 | ||||||||
| No | 3181 (96.89) | 2774 (97.03) | 407 (95.99) | −0.053 | 765 (95.62) | 382 (95.50) | 383 (95.75) | 0.012 | ||
| Yes | 102 (3.11) | 85 (2.97) | 17 (4.01) | 0.053 | 35 (4.38) | 18 (4.50) | 17 (4.25) | −0.012 | ||
| Hemiplegia/paraplegia, n (%) | 0.62 | 0.718 | ||||||||
| No | 3128 (95.28) | 2722 (95.21) | 406 (95.75) | 0.027 | 768 (96) | 385 (96.25) | 383 (95.75) | −0.025 | ||
| Yes | 155 (4.72) | 137 (4.79) | 18 (4.25) | −0.027 | 32 (4) | 15 (3.75) | 17 (4.25) | 0.025 | ||
| Peptic ulcer disease, n (%) | 0.816 | 0.594 | ||||||||
| No | 3158 (96.19) | 2751 (96.22) | 407 (95.99) | −0.012 | 767 (95.88) | 382 (95.50) | 385 (96.25) | 0.039 | ||
| Yes | 125 (3.81) | 108 (3.78) | 17 (4.01) | 0.012 | 33 (4.12) | 18 (4.50) | 15 (3.75) | −0.039 | ||
| COPD, n (%) | 0.002 | 0.63 | ||||||||
| No | 2601 (79.23) | 2289 (80.06) | 312 (73.58) | −0.147 | 590 (73.75) | 292 (73.00) | 298 (74.50) | 0.034 | ||
| Yes | 682 (20.77) | 570 (19.94) | 112 (26.42) | 0.147 | 210 (26.25) | 108 (27.00) | 102 (25.50) | −0.034 | ||
| Moderate to severe liver disease, n (%) | 0.005 | 1 | ||||||||
| No | 2984 (90.89) | 2583 (90.35) | 401 (94.58) | 0.187 | 754 (94.25) | 377 (94.25) | 377 (94.25) | 0 | ||
| Yes | 299 (9.11) | 276 (9.65) | 23 (5.42) | −0.187 | 46 (5.75) | 23 (5.75) | 23 (5.75) | 0 | ||
| Severity scores | ||||||||||
| SOFA, M (Q1, Q3) | 3.00 (2.00, 4.00) | 3.00 (2.00, 4.00) | 3.00 (2.00, 5.00) | 0.439 | 0.047 | 3.00 (2.00, 5.00) | 3.00 (2.00, 5.00) | 3.00 (2.00, 5.00) | 0.71 | 0.001 |
| GCS, M (Q1, Q3) | 15.00 (13.00, 15.00) | 15.00 (13.00, 15.00) | 15.00 (13.00, 15.00) | 0.688 | 0.075 | 15.00 (14.00, 15.00) | 15.00 (14.00, 15.00) | 15.00 (13.00, 15.00) | 0.596 | −0.074 |
| CCI, M (Q1, Q3) | 8.00 (6.00, 10.00) | 8.00 (6.00, 10.00) | 8.50 (7.00, 11.00) | < 0.001 | 0.229 | 9.00 (7.00, 11.00) | 9.00 (7.00, 11.00) | 8.50 (7.00, 11.00) | 0.584 | −0.04 |
| SAPS II, M (Q1, Q3) | 46.00 (37.00, 56.00) | 45.00 (37.00, 56.00) | 47.50 (39.00, 58.00) | 0.003 | 0.143 | 47.00 (39.00, 56.00) | 47.00 (38.00, 56.00) | 47.00 (39.00, 57.00) | 0.595 | 0.039 |
| Clinical interventions | ||||||||||
| CRRT, n (%) | 0.01 | 0.816 | ||||||||
| No | 3029 (92.26) | 2651 (92.72) | 378 (89.15) | −0.115 | 718 (89.75) | 360 (90.00) | 358 (89.50) | −0.016 | ||
| Yes | 254 (7.74) | 208 (7.28) | 46 (10.85) | 0.115 | 82 (10.25) | 40 (10.00) | 42 (10.50) | 0.016 | ||
| Mechanical ventilation, n (%) | 0.378 | 0.394 | ||||||||
| No | 1337 (40.72) | 1156 (40.43) | 181 (42.69) | 0.046 | 358 (44.75) | 185 (46.25) | 173 (43.25) | −0.061 | ||
| Yes | 1946 (59.28) | 1703 (59.57) | 243 (57.31) | −0.046 | 442 (55.25) | 215 (53.75) | 227 (56.75) | 0.061 | ||
| Vasoactive agent, n (%) | < 0.001 | 1 | ||||||||
| No | 1666 (50.75) | 1411 (49.35) | 255 (60.14) | 0.22 | 478 (59.75) | 239 (59.75) | 239 (59.75) | 0 | ||
| Yes | 1617 (49.25) | 1448 (50.65) | 169 (39.86) | −0.22 | 322 (40.25) | 161 (40.25) | 161 (40.25) | 0 | ||
| ACEI/ARB, n (%) | < 0.001 | 0.867 | ||||||||
| No | 2677 (81.54) | 2357 (82.44) | 320 (75.47) | −0.162 | 616 (77) | 309 (77.25) | 307 (76.75) | −0.012 | ||
| Yes | 606 (18.46) | 502 (17.56) | 104 (24.53) | 0.162 | 184 (23) | 91 (22.75) | 93 (23.25) | 0.012 | ||
| NSAIDs, n (%) | 0.876 | 0.504 | ||||||||
| No | 2094 (63.78) | 1825 (63.83) | 269 (63.44) | −0.008 | 521 (65.12) | 265 (66.25) | 256 (64.00) | −0.047 | ||
| Yes | 1189 (36.22) | 1034 (36.17) | 155 (36.56) | 0.008 | 279 (34.88) | 135 (33.75) | 144 (36.00) | 0.047 | ||
| Opioids, n (%) | 0.034 | 0.349 | ||||||||
| No | 421 (12.82) | 353 (12.35) | 68 (16.04) | 0.101 | 138 (17.25) | 74 (18.50) | 64 (16.00) | −0.068 | ||
| Yes | 2862 (87.18) | 2506 (87.65) | 356 (83.96) | −0.101 | 662 (82.75) | 326 (81.50) | 336 (84.00) | 0.068 | ||
| Benzodiazepines, n (%) | 0.062 | 0.622 | ||||||||
| No | 3158 (96.19) | 2757 (96.43) | 401 (94.58) | −0.082 | 761 (95.12) | 379 (94.75) | 382 (95.50) | 0.036 | ||
| Yes | 125 (3.81) | 102 (3.57) | 23 (5.42) | 0.082 | 39 (4.88) | 21 (5.25) | 18 (4.50) | −0.036 | ||
| Propofol, n (%) | 0.017 | 0.257 | ||||||||
| No | 1649 (50.23) | 1459 (51.03) | 190 (44.81) | −0.125 | 382 (47.75) | 199 (49.75) | 183 (45.75) | −0.08 | ||
| Yes | 1634 (49.77) | 1400 (48.97) | 234 (55.19) | 0.125 | 418 (52.25) | 201 (50.25) | 217 (54.25) | 0.08 | ||
| Aminoglycosides, n (%) | 0.004 | 0.71 | ||||||||
| No | 3070 (93.51) | 2660 (93.04) | 410 (96.70) | 0.205 | 770 (96.25) | 384 (96.00) | 386 (96.50) | 0.027 | ||
| Yes | 213 (6.49) | 199 (6.96) | 14 (3.30) | −0.205 | 30 (3.75) | 16 (4.00) | 14 (3.50) | −0.027 | ||
| Beta‐lactams, n (%) | < 0.001 | 0.484 | ||||||||
| No | 610 (18.58) | 572 (20.01) | 38 (8.96) | −0.387 | 82 (10.25) | 44 (11.00) | 38 (9.50) | −0.051 | ||
| Yes | 2673 (81.42) | 2287 (79.99) | 386 (91.04) | 0.387 | 718 (89.75) | 356 (89.00) | 362 (90.50) | 0.051 | ||
| Macrolides, n (%) | < 0.001 | 0.36 | ||||||||
| No | 2636 (80.29) | 2341 (81.88) | 295 (69.58) | −0.267 | 550 (68.75) | 269 (67.25) | 281 (70.25) | 0.066 | ||
| Yes | 647 (19.71) | 518 (18.12) | 129 (30.42) | 0.267 | 250 (31.25) | 131 (32.75) | 119 (29.75) | −0.066 | ||
| Glycopeptides, n (%) | 0.123 | 0.635 | ||||||||
| No | 608 (18.52) | 541 (18.92) | 67 (15.80) | −0.086 | 133 (16.62) | 69 (17.25) | 64 (16.00) | −0.034 | ||
| Yes | 2675 (81.48) | 2318 (81.08) | 357 (84.20) | 0.086 | 667 (83.38) | 331 (82.75) | 336 (84.00) | 0.034 | ||
| Sulfonamides, n (%) | 0.59 | 0.762 | ||||||||
| No | 2766 (84.25) | 2405 (84.12) | 361 (85.14) | 0.029 | 685 (85.62) | 341 (85.25) | 344 (86.00) | 0.022 | ||
| Yes | 517 (15.75) | 454 (15.88) | 63 (14.86) | −0.029 | 115 (14.37) | 59 (14.75) | 56 (14.00) | −0.022 | ||
| Tetracyclines, n (%) | < 0.001 | 0.791 | ||||||||
| No | 3158 (96.19) | 2768 (96.82) | 390 (91.98) | −0.178 | 738 (92.25) | 368 (92.00) | 370 (92.50) | 0.019 | ||
| Yes | 125 (3.81) | 91 (3.18) | 34 (8.02) | 0.178 | 62 (7.75) | 32 (8.00) | 30 (7.50) | −0.019 | ||
| Quinolones, n (%) | < 0.001 | 0.677 | ||||||||
| No | 2213 (67.41) | 1886 (65.97) | 327 (77.12) | 0.266 | 611 (76.38) | 303 (75.75) | 308 (77.00) | 0.03 | ||
| Yes | 1070 (32.59) | 973 (34.03) | 97 (22.88) | −0.266 | 189 (23.62) | 97 (24.25) | 92 (23.00) | −0.03 | ||
| Other antibiotics use, n (%) | 0.101 | 0.944 | ||||||||
| No | 1840 (56.05) | 1618 (56.59) | 222 (52.36) | −0.085 | 421 (52.62) | 210 (52.50) | 211 (52.75) | 0.005 | ||
| Yes | 1443 (43.95) | 1241 (43.41) | 202 (47.64) | 0.085 | 379 (47.38) | 190 (47.50) | 189 (47.25) | −0.005 | ||
Note: Bold values indicate statistical significance (p < 0.05).
Abbreviations: ACEI/ARB, Angiotensin‐Converting Enzyme Inhibitors/Angiotensin II Receptor Blockers; AG, Anion Gap; AKI, Acute Kidney Injury; BUN, Blood Urea Nitrogen; CCI, Charlson Comorbidity Index; CKD, Chronic Kidney Disease; COPD, Chronic Obstructive Pulmonary Disease; CRRT, Continuous Renal Replacement Therapy; DBP, Diastolic Blood Pressure; GCS, Glasgow Coma Scale; MAP, Mean Arterial Pressure; NSAIDs, Nonsteroidal Anti‐Inflammatory Drugs; PSM, propensity score matching; PT, Prothrombin Time; PTT, Partial Thromboplastin Time; RBC, Red Blood Cell count; RDW, Red cell Distribution Width; RR, Respiratory Rate; SAPS II, Simplified Acute Physiology Score II; SBP, Systolic Blood Pressure; SMD, standardized mean differences; SOFA, Sequential Organ Failure Assessment Score; SPO2, Peripheral capillary oxygen saturation; T, Temperature; WBC, White Blood Cell count.
FIGURE 2.
Probability density plot of ramelteon and non‐ramelteon group before and after PSM. Before PSM, 424 ramelteon users and 2859 non‐users were included. After 1:1 PSM, 400 patients were included in each group. PSM, propensity score matching.
FIGURE 3.
Comparison of SMD before and after PSM. Covariate balance was assessed using SMDs, with smaller absolute SMD values indicating better balance between groups. ACEI/ARB, Angiotensin‐Converting Enzyme Inhibitors/Angiotensin II Receptor Blockers; AG, Anion Gap; AKI, Acute Kidney Injury; BUN, Blood Urea Nitrogen; CCI, Charlson Comorbidity Index; CKD, Chronic Kidney Disease; COPD, Chronic Obstructive Pulmonary Disease; CRRT, Continuous Renal Replacement Therapy; DBP, Diastolic Blood Pressure; GCS, Glasgow Coma Scale; MAP, Mean Arterial Pressure; NSAIDs, Nonsteroidal Anti‐Inflammatory Drugs; PSM, propensity score matching; PSM, propensity score matching; PT, Prothrombin Time; PTT, Partial Thromboplastin Time; RBC, Red Blood Cell count; RDW, Red cell Distribution Width; RR, Respiratory Rate; SAPS II, Simplified Acute Physiology Score II; SBP, Systolic Blood Pressure; SMD, standardized mean differences; SOFA, Sequential Organ Failure Assessment Score; SPO2, Peripheral capillary oxygen saturation; T, Temperature; WBC, White Blood Cell count.
3.2. Ramelteon and Mortality Outcomes
The timing of ramelteon initiation relative to hospital admission is summarized in Table S4. Among ramelteon‐exposed patients in the propensity score‐matched cohort, the admission‐to‐first prescription interval was computable for 399 of 400 patients. The median interval was 97.2 h (IQR, 30.2–261.3), corresponding to 4.05 days (IQR, 1.26–10.89). Ramelteon was initiated within 24 h in 86 patients (21.6%), between 24 and 48 h in 42 patients (10.5%), and after 48 h in 271 patients (67.9%). Univariable analysis (Table 2) showed that ramelteon use was significantly associated with lower 30‐day mortality, 90‐day mortality, ICU mortality, and in‐hospital mortality. Kaplan–Meier survival curves (Figure 4) further confirmed these findings, demonstrating significantly higher 30‐day and 90‐day survival probabilities in patients receiving ramelteon compared to those who did not, both in unmatched and PSM‐matched cohorts (all p < 0.0001). To select variables for inclusion in the multivariable Cox regression model, we performed a univariable Cox regression analysis, the results of which are shown in Table 3. Prior to the univariable analysis, we excluded covariates with a variance inflation factor (VIF) > 10. Because no deaths occurred within 30 days among patients not receiving opioids in the PSM‐matched cohort, opioid use was not included in the univariable analysis. Table 3 shows the results of the univariable Cox regression analysis of factors influencing 30‐day mortality after PSM. Table 4 presents the multivariable analysis results of the association of ramelteon and the risk of mortality. In the fully adjusted model (Model 3) after PSM, ramelteon use was significantly associated with lower 30‐day mortality (HR: 0.37, 95% CI: 0.28–0.50, p < 0.001), 90‐day mortality (HR: 0.39, 95% CI: 0.29–0.52, p < 0.001), ICU mortality (OR: 0.29, 95% CI: 0.18–0.46, p < 0.001), and in‐hospital mortality (OR: 0.50, 95% CI: 0.35–0.72, p < 0.001).
TABLE 2.
Univariable analysis of the association between ramelteon use and mortality outcomes.
| Total, n (%) | Non‐ramelteon, n (%) | Ramelteon, n (%) | HR/OR (95% CI) | p | |
|---|---|---|---|---|---|
| Before PSM | |||||
| Primary outcome | |||||
| 30‐day mortality | 803 (24.46) | 723 (25.29) | 80 (18.87) | 0.55 (0.44~0.69) | < 0.001 |
| Secondary outcomes | |||||
| 90‐day mortality | 866 (26.38) | 775 (27.11) | 91 (21.46) | 0.54 (0.43~0.67) | < 0.001 |
| ICU mortality | 569 (17.33) | 524 (18.33) | 45 (10.61) | 0.53 (0.38~0.73) | < 0.001 |
| In‐hospital mortality | 868 (26.44) | 777 (27.18) | 91 (21.46) | 0.73 (0.57~0.94) | 0.013 |
| After PSM | |||||
| Primary outcome | |||||
| 30‐day mortality | 201 (25.12) | 127 (31.75) | 74 (18.50) | 0.42 (0.31~0.56) | < 0.001 |
| Secondary outcomes | |||||
| 90‐day mortality | 217 (27.12) | 132 (33.00) | 85 (21.25) | 0.44 (0.33~0.58) | < 0.001 |
| ICU mortality | 137 (17.12) | 96 (24.00) | 41 (10.25) | 0.36 (0.24~0.54) | < 0.001 |
| In‐hospital mortality | 217 (27.12) | 132 (33.00) | 85 (21.25) | 0.55 (0.40~0.75) | < 0.001 |
FIGURE 4.
Kaplan–Meier survival curves for 30‐day and 90‐day mortality from hospital admission in unmatched and propensity score‐matched cohorts: (A) Before PSM (30‐day); (B) After PSM (30‐day); (C) Before PSM (90‐day); (D) After PSM (90‐day). Before PSM, 424 ramelteon users and 2859 non‐users were included; after PSM, 400 patients were included in each group. Survival curves were compared using the log‐rank test. Abbreviations: PSM, propensity score matching.
TABLE 3.
Univariable Cox regression analysis of factors influencing 30‐day mortality after PSM.
| Variable | HR (95% CI) | p |
|---|---|---|
| Ramelteon | ||
| No | 1.00 (reference) | |
| Yes | 0.42 (0.31~0.56) | < 0.001 |
| Sex | ||
| Female | 1.00 (reference) | |
| Male | 1.01 (0.76~1.34) | 0.935 |
| Race | ||
| Asian/Hispanic or Latino/other | 1.00 (reference) | |
| Black | 0.75 (0.45~1.25) | 0.269 |
| White | 0.87 (0.63~1.19) | 0.373 |
| Atrial fibrillation | ||
| No | 1.00 (reference) | |
| Yes | 0.98 (0.74~1.30) | 0.916 |
| Heart failure | ||
| No | 1.00 (reference) | |
| Yes | 1.39 (1.05~1.84) | 0.02 |
| Myocardial infarction | ||
| No | 1.00 (reference) | |
| Yes | 0.95 (0.65~1.39) | 0.788 |
| CKD | ||
| No | 1.00 (reference) | |
| Yes | 0.94 (0.69~1.27) | 0.681 |
| AKI stage | ||
| No AKI | 1.00 (reference) | |
| I | 2.38 (0.79~7.17) | 0.123 |
| II | 2.94 (1.07~8.07) | 0.036 |
| III | 5.50 (2.03~14.91) | < 0.001 |
| Diabetes | ||
| No | 1.00 (reference) | |
| Yes | 0.85 (0.62~1.15) | 0.289 |
| Hypertension | ||
| No | 1.00 (reference) | |
| Yes | 0.87 (0.64~1.18) | 0.364 |
| Cerebrovascular disease | ||
| No | 1.00 (reference) | |
| Yes | 0.86 (0.57~1.29) | 0.465 |
| Peripheral vascular disease | ||
| No | 1.00 (reference) | |
| Yes | 1.23 (0.82~1.86) | 0.318 |
| Rheumatic disease | ||
| No | 1.00 (reference) | |
| Yes | 1.48 (0.78~2.79) | 0.228 |
| Dementia | ||
| No | 1.00 (reference) | |
| Yes | 1.14 (0.58~2.22) | 0.71 |
| Hemiplegia paraplegia | ||
| No | 1.00 (reference) | |
| Yes | 0.34 (0.11~1.08) | 0.066 |
| Peptic ulcer disease | ||
| No | 1.00 (reference) | |
| Yes | 1.00 (0.51~1.96) | 0.995 |
| COPD | ||
| No | 1.00 (reference) | |
| Yes | 1.05 (0.77~1.45) | 0.744 |
| Moderate to severe liver disease | ||
| No | 1.00 (reference) | |
| Yes | 0.87 (0.47~1.60) | 0.654 |
| Metastatic solid tumor | ||
| No | 1.00 (reference) | |
| Yes | 1.66 (1.25~2.19) | < 0.001 |
| CRRT | ||
| No | 1.00 (reference) | |
| Yes | 1.24 (0.85~1.81) | 0.272 |
| Mechanical ventilation | ||
| No | 1.00 (reference) | |
| Yes | 1.13 (0.84~1.50) | 0.419 |
| Vasoactive agent | ||
| No | 1.00 (reference) | |
| Yes | 1.28 (0.97~1.69) | 0.076 |
| ACEI ARB | ||
| No | 1.00 (reference) | |
| Yes | 0.36 (0.23~0.56) | < 0.001 |
| NSAIDs | ||
| No | 1.00 (reference) | |
| Yes | 0.60 (0.44~0.82) | 0.001 |
| Benzodiazepines | ||
| No | 1.00 (reference) | |
| Yes | 0.64 (0.32~1.31) | 0.224 |
| Propofol | ||
| No | 1.00 (reference) | |
| Yes | 0.95 (0.72~1.27) | 0.745 |
| Aminoglycosides | ||
| No | 1.00 (reference) | |
| Yes | 0.74 (0.36~1.50) | 0.397 |
| Beta‐lactams | ||
| No | 1.00 (reference) | |
| Yes | 1.06 (0.63~1.80) | 0.827 |
| Macrolides | ||
| No | 1.00 (reference) | |
| Yes | 1.25 (0.94~1.67) | 0.126 |
| Glycopeptides | ||
| No | 1.00 (reference) | |
| Yes | 2.39 (1.30~4.40) | 0.005 |
| Sulfonamides | ||
| No | 1.00 (reference) | |
| Yes | 0.79 (0.53~1.18) | 0.245 |
| Tetracyclines | ||
| No | 1.00 (reference) | |
| Yes | 1.15 (0.72~1.84) | 0.564 |
| Quinolones | ||
| No | 1.00 (reference) | |
| Yes | 0.57 (0.40~0.82) | 0.002 |
| Other Antibiotics use | ||
| No | 1.00 (reference) | |
| Yes | 0.98 (0.74~1.30) | 0.885 |
| Age | 1.02 (1.01~1.03) | 0.006 |
| SOFA score | 1.03 (0.97~1.08) | 0.354 |
| Weight | 0.99 (0.99~0.99) | 0.018 |
| HR | 1.01 (1.01~1.01) | 0.012 |
| SBP | 0.99 (0.99~0.99) | 0.041 |
| DBP | 1.00 (0.99~1.00) | 0.411 |
| MAP | 1.00 (0.99~1.00) | 0.518 |
| RR | 1.04 (1.02~1.06) | < 0.001 |
| T | 0.81 (0.67~0.98) | 0.035 |
| SPO2 | 0.85 (0.76~0.94) | 0.002 |
| GCS | 0.97 (0.92~1.02) | 0.249 |
| SAPS II | 1.02 (1.01~1.03) | < 0.001 |
| Hemoglobin | 0.97 (0.89~1.04) | 0.378 |
| Platelet | 1.00 (1.00~1.00) | 0.376 |
| RBC | 0.92 (0.75~1.13) | 0.419 |
| RDW | 1.07 (1.02~1.12) | 0.003 |
| WBC | 1.00 (1.00~1.01) | 0.459 |
| AG | 1.01 (0.98~1.05) | 0.387 |
| Bicarbonate | 1.00 (0.97~1.03) | 0.913 |
| BUN | 1.01 (1.01~1.01) | 0.004 |
| Chloride | 0.95 (0.93~0.97) | < 0.001 |
| Serum creatinine | 1.01 (0.92~1.09) | 0.896 |
| Glucose | 1.00 (1.00~1.00) | 0.275 |
| Sodium | 0.94 (0.92~0.97) | < 0.001 |
| Potassium | 1.20 (1.01~1.44) | 0.049 |
| PT | 1.01 (1.00~1.02) | 0.115 |
| PTT | 1.00 (1.00~1.01) | 0.058 |
Abbreviations: ACEI/ARB, Angiotensin‐Converting Enzyme Inhibitors/Angiotensin II Receptor Blockers; AG, Anion Gap; AKI, Acute Kidney Injury; BUN, Blood Urea Nitrogen; CI, Confidence Interval; CKD, Chronic Kidney Disease; COPD, Chronic Obstructive Pulmonary Disease; CRRT, Continuous Renal Replacement Therapy; DBP, Diastolic Blood Pressure; GCS, Glasgow Coma Scale; HR, Hazard Ratio; MAP, Mean Arterial Pressure; NSAIDs, Nonsteroidal Anti‐Inflammatory Drugs; PSM, Propensity Score Matching; PT, Prothrombin Time; PTT, Partial Thromboplastin Time; RBC, Red Blood Cell count; RDW, Red cell Distribution Width; RR, Respiratory Rate; SAPS II, Simplified Acute Physiology Score II; SBP, Systolic Blood Pressure; SOFA Score, Sequential Organ Failure Assessment Score; SPO2, Peripheral capillary oxygen saturation; T, Temperature; WBC, White Blood Cell count.
TABLE 4.
Association of ramelteon and the risk of mortality.
| Outcome | Before PSM | After PSM | ||
|---|---|---|---|---|
| HR/OR (95% CI) | p | HR/OR (95% CI) | p | |
| Primary outcome | ||||
| 30‐day mortality | ||||
| Model 1 | 0.45 (0.36~0.57) | < 0.001 | 0.41 (0.31~0.55) | < 0.001 |
| Model 2 | 0.45 (0.35~0.57) | < 0.001 | 0.38 (0.28~0.51) | < 0.001 |
| Model 3 | 0.44 (0.35~0.56) | < 0.001 | 0.37 (0.28~0.50) | < 0.001 |
| Secondary outcomes | ||||
| 90‐day mortality | ||||
| Model 1 | 0.45 (0.36~0.56) | < 0.001 | 0.43 (0.32~0.56) | < 0.001 |
| Model 2 | 0.45 (0.36~0.57) | < 0.001 | 0.40 (0.30~0.54) | < 0.001 |
| Model 3 | 0.44 (0.35~0.55) | < 0.001 | 0.39 (0.29~0.52) | < 0.001 |
| ICU mortality | ||||
| Model 1 | 0.41 (0.29~0.57) | < 0.001 | 0.35 (0.23~0.53) | < 0.001 |
| Model 2 | 0.37 (0.26~0.53) | < 0.001 | 0.31 (0.20~0.48) | < 0.001 |
| Model 3 | 0.36 (0.25~0.52) | < 0.001 | 0.29 (0.18~0.46) | < 0.001 |
| In‐hospital mortality | ||||
| Model 1 | 0.60 (0.46~0.77) | < 0.001 | 0.54 (0.39~0.76) | < 0.001 |
| Model 2 | 0.57 (0.44~0.75) | < 0.001 | 0.51 (0.36~0.73) | < 0.001 |
| Model 3 | 0.57 (0.43~0.75) | < 0.001 | 0.50 (0.35~0.72) | < 0.001 |
Note: Model 1: Adjust: age, weight, heart rate, SBP, RR, T, SPO2, RDW, BUN, Chloride, Sodium, Potassium; Model 2: Adjust: Heart failure, AKI stage, Metastatic solid tumor, age, weight, heart rate, SBP, RR, T, SPO2, SAPS II, RDW, BUN, Chloride, Sodium, Potassium; Model 3: Adjust: Heart failure, AKI stage, Metastatic solid tumor, ACEI_ARB, NSAIDs, Glycopeptides, Quinolones, age, weight, heart rate, SBP, RR, T, SPO2, SAPS II, RDW, BUN, Chloride, Sodium, Potassium.
Abbreviations: ACEI/ARB, Angiotensin‐Converting Enzyme Inhibitors/Angiotensin II Receptor Blockers; AKI, Acute Kidney Injury; BUN, Blood Urea Nitrogen; CI, Confidence Interval; HR, Hazard Ratio; NSAIDs, Nonsteroidal Anti‐Inflammatory Drugs; OR, Odds Ratio; PSM, Propensity Score Matching; RDW, Red cell Distribution Width; RR, Respiratory Rate; SAPS II, Simplified Acute Physiology Score II; SBP, Systolic Blood Pressure; SPO2, Peripheral capillary oxygen saturation; T, Temperature.
3.3. Ramelteon Exposure Duration and Mortality Outcomes
Table 5 showed that, after PSM, patients receiving ramelteon for ≤ 3 days had significantly lower 30‐day mortality (HR: 0.38, 95% CI: 0.27–0.55, p < 0.001), 90‐day mortality (HR: 0.41, 95% CI: 0.29–0.57, p < 0.001), ICU mortality (OR: 0.30, 95% CI: 0.18–0.52, p < 0.001), and in‐hospital mortality (OR: 0.45, 95% CI: 0.29–0.69, p < 0.001) compared to those not receiving ramelteon. Patients receiving ramelteon for 3–7 days also had significantly lower 30‐day mortality (HR: 0.37, 95% CI: 0.23–0.61, p < 0.001), 90‐day mortality (HR: 0.40, 95% CI: 0.26–0.63, p < 0.001), ICU mortality (OR: 0.26, 95% CI: 0.12–0.55, p < 0.001), and in‐hospital mortality (OR: 0.53, 95% CI: 0.31–0.92, p = 0.023). However, no statistically significant differences were observed in any outcome measures for patients receiving ramelteon for > 7 days.
TABLE 5.
Ramelteon duration‐response relationship after PSM.
| Duration (days) | 30‐day mortality | 90‐day mortality | ICU mortality | In‐hospital mortality | ||||
|---|---|---|---|---|---|---|---|---|
| n (%) | HR (95% CI), p | n (%) | HR (95% CI), p | n (%) | OR (95% CI), p | n (%) | OR (95% CI), p | |
| Without ramelteon | 127 (31.75) | 1.00 (reference) | 132 (33.00) | 1.00 (reference) | 96 (24.00) | 132 (33.00) | ||
| ≤ 3 | 43 (18.86) | 0.38 (0.27~0.55), < 0.001 | 48 (21.05) | 0.41 (0.29~0.57), < 0.001 | 26 (11.40) | 0.30 (0.18~0.52), < 0.001 | 48 (21.05) | 0.45 (0.29~0.69), < 0.001 |
| 3–7 | 20 (15.62) | 0.37 (0.23~0.61), < 0.001 | 26 (20.31) | 0.40 (0.26~0.63), < 0.001 | 11 (8.59) | 0.26 (0.12~0.55), < 0.001 | 26 (20.31) | 0.53 (0.31~0.92), 0.023 |
| > 7 | 11 (25.00) | 0.74 (0.39~1.38), 0.338 | 11 (25.00) | 0.63 (0.34~1.19), 0.154 | 4 (9.09) | 0.33 (0.10~1.07), 0.064 | 11 (25.00) | 0.81 (0.36~1.83), 0.606 |
Note: Adjusted for variables in Model 3.
Abbreviations: CI, Confidence Interval; HR, Hazard Ratio; OR, Odds Ratio; PSM, Propensity Score Matching.
3.4. Subgroup Analysis
Subgroup analysis demonstrated that ramelteon use was associated with a reduced risk of 30‐day mortality across most subgroups (p < 0.001), and no significant interactions were observed (all p for interaction > 0.05), suggesting the robust nature of the observed association (Table 6).
TABLE 6.
Subgroup analysis of the association between ramelteon and 30‐day mortality after PSM.
| Variable | HR (95% CI) | p | p for interaction |
|---|---|---|---|
| Sex | 0.729 | ||
| Female | 0.36 (0.21, 0.59) | < 0.001 | |
| Male | 0.38 (0.25, 0.55) | < 0.001 | |
| Age | 0.715 | ||
| ≥ 65 | 0.40 (0.28, 0.56) | < 0.001 | |
| < 65 | 0.29 (0.14, 0.59) | < 0.001 | |
| Race | 0.642 | ||
| White | 0.42 (0.28, 0.61) | < 0.001 | |
| Asian/Hispanic or Latino/other | 0.20 (0.10, 0.41) | < 0.001 | |
| Black | 0.40 (0.09, 1.88) | 0.247 | |
| SOFA | 0.876 | ||
| < 5 | 0.34 (0.24, 0.50) | < 0.001 | |
| ≥ 5 | 0.43 (0.25, 0.75) | 0.003 | |
| Atrial fibrillation | 0.473 | ||
| No | 0.35 (0.24, 0.53) | < 0.001 | |
| Yes | 0.34 (0.21, 0.55) | < 0.001 | |
| Heart failure | 0.065 | ||
| No | 0.28 (0.19, 0.42) | < 0.001 | |
| Yes | 0.46 (0.29, 0.75) | 0.002 | |
| Myocardial infarction | 0.882 | ||
| Yes | 0.29 (0.11, 0.78) | 0.014 | |
| No | 0.36 (0.26, 0.50) | < 0.001 | |
| CKD | 0.36 | ||
| Yes | 0.29 (0.16, 0.53) | < 0.001 | |
| No | 0.40 (0.28, 0.58) | < 0.001 | |
| Diabetes | 0.387 | ||
| No | 0.41 (0.28, 0.58) | < 0.001 | |
| Yes | 0.26 (0.14, 0.50) | < 0.001 | |
| Hypertension | 0.435 | ||
| No | 0.36 (0.25, 0.51) | < 0.001 | |
| Yes | 0.35 (0.18, 0.66) | 0.001 | |
| Cerebrovascular disease | 0.575 | ||
| No | 0.37 (0.27, 0.51) | < 0.001 | |
| Yes | 0.50 (0.20, 1.25) | 0.137 | |
| Peripheral vascular disease | 0.677 | ||
| No | 0.37 (0.27, 0.51) | < 0.001 | |
| Yes | 0.41 (0.13, 1.27) | 0.121 | |
| COPD | 0.086 | ||
| No | 0.44 (0.31, 0.62) | < 0.001 | |
| Yes | 0.17 (0.08, 0.37) | < 0.001 | |
| Metastatic solid tumor | 0.764 | ||
| No | 0.37 (0.25, 0.55) | < 0.001 | |
| Yes | 0.39 (0.24, 0.63) | < 0.001 | |
| CRRT | 0.841 | ||
| No | 0.37 (0.27, 0.52) | < 0.001 | |
| Yes | 0.26 (0.09, 0.73) | 0.010 | |
| Mechanical ventilation | 0.844 | ||
| No | 0.41 (0.24, 0.70) | < 0.001 | |
| Yes | 0.38 (0.25, 0.56) | < 0.001 | |
| Vasoactive agent | 0.820 | ||
| No | 0.31 (0.20, 0.49) | < 0.001 | |
| Yes | 0.35 (0.22, 0.55) | < 0.001 | |
| ACEI/ARB | 0.641 | ||
| No | 0.36 (0.26, 0.50) | < 0.001 | |
| Yes | 0.39 (0.12, 1.26) | 0.116 | |
| NSAIDs | 0.978 | ||
| No | 0.36 (0.25, 0.51) | < 0.001 | |
| Yes | 0.37 (0.20, 0.69) | 0.002 |
Note: HRs were estimated using Cox regression models adjusted for the same covariates as Model 3, except for the variable used for stratification in each subgroup analysis.
Abbreviations: ACEI/ARB, Angiotensin‐Converting Enzyme Inhibitors/Angiotensin II Receptor Blockers; CI, Confidence Interval; CKD, Chronic Kidney Disease; COPD, Chronic Obstructive Pulmonary Disease; CRRT, Continuous Renal Replacement Therapy; HR, Hazard Ratio; NSAIDs, Nonsteroidal Anti‐Inflammatory Drugs; PSM, Propensity Score Matching; SOFA, Sequential Organ Failure Assessment score.
4. Discussion
This study aimed to evaluate the association between ramelteon use and mortality in patients with sepsis and cancer, including a total of 3283 patients for analysis, with 2859 (87.1%) in the non‐ramelteon group and 424 (12.9%) in the ramelteon group. Multiple statistical methods were employed for analysis. The results showed that ramelteon use was significantly associated with lower 30‐day, 90‐day, ICU, and in‐hospital mortality rates. The Kaplan–Meier survival curve analysis corroborated these findings. To control for potential confounding factors, we utilized PSM, and a significant reduction in SMD was observed post‐PSM (400 in the non‐ramelteon group and 400 in the ramelteon group, totaling 800), indicating good balance in baseline characteristics across groups. Multivariate Cox regression analysis, after adjusting for confounders such as age and comorbidities, continued to demonstrate a significant association between ramelteon use and reduced mortality risk. Further analysis revealed that ramelteon use for ≤ 7 days was associated with a decreased risk of mortality, whereas no significant association was observed with use beyond 7 days. Subgroup analyses showed generally consistent associations across subgroups. In conclusion, the findings of this study suggest that ramelteon use may have potential prognostic relevance in patients with sepsis and cancer, although causal effects cannot be inferred from this observational study.
Our study observed an association between ramelteon and improved survival outcomes in patients with sepsis and cancer, potentially mediated by diverse biological processes. While the precise pathophysiology remains incompletely understood, infection, inflammation, and oxidative stress are recognized as key components. Ramelteon may exert biological effects by modulating these core processes. Mechanistically, ramelteon's potential biological relevance is thought to involve activation of melatonin receptors, which mediate antioxidant and anti‐inflammatory responses. Previous studies demonstrated that melatonin can promote antioxidant stress responses through mechanisms such as electron transfer, hydrogen transfer, free radical adduct formation, and metal chelation [23]. Melatonin's role in reducing oxidative stress and inflammation has been shown to confer protection to vital organs like the lungs, kidneys, and liver, especially in conditions such as sepsis [24]. Experimental evidence demonstrates that melatonin receptor activation can attenuate organ damage, such as in bleomycin‐induced injury models, where protective effects on liver perfusion and tissue integrity were observed [25]. Ramelteon is capable of activating nuclear factor‐erythroid 2‐related factor 2 (Nrf2), a key regulator of the antioxidant response. Nrf2 activation by ramelteon may attenuate lipopolysaccharide (LPS)‐induced neuroinflammation [26]. Studies further indicate that ramelteon can modulate the Nrf2/heme oxygenase‐1 (HO‐1) signaling pathway, leading to reduced oxidative stress and decreased levels of pro‐inflammatory mediators, thereby alleviating LPS‐induced damage to human pulmonary microvascular endothelial cells [27]. Yan et al. [28] demonstrated that melatonin mitigates aflatoxin B1‐induced cardiotoxicity by inhibiting NOD‐like receptor family, pyrin domain containing 3 (NLRP3) inflammasome activation, primarily through the reduction of oxidative stress, thereby preventing myocardial injury.
The anti‐tumor properties of ramelteon and melatonin may also contribute to improved outcomes in patients with sepsis and cancer. In vitro experiments have demonstrated that ramelteon inhibits the proliferation of the ER‐positive endometrial cancer cell line HHUA to a similar extent as melatonin, and this inhibitory effect can be blocked by luzindole, suggesting that the anti‐tumor effects of ramelteon may be mediated by MT1/MT2 receptors [7]. Additionally, melatonin may suppress the proliferation of estrogen‐mediated cancer cells by modulating estrogen‐related signaling pathways [29]. Ramelteon may regulate the expression of matrix metalloproteinases (MMP)‐2 and MMP‐9 by modulating intracellular cyclic adenosine monophosphate (cAMP) concentrations through the MT1 receptor [30]. This action can lead to the inhibition of protein kinase A (PKA) activity, thereby suppressing tumor cell invasiveness [31]. Silent information regulator 2 (SIRT2) proteins, known as sirtuins, are a class of Nicotinamide Adenine Dinucleotide (NAD)+‐dependent deacetylases that regulate various critical cellular processes, and are considered potential targets for cancer therapy [32, 33]. SIRT1, a homolog of SIRT2 within this sirtuin family, has been demonstrated in prior research to have its abnormally elevated levels in tumor cells reduced by melatonin, leading to the inhibition of cell proliferation and tumor growth [10, 34]. Furthermore, Zhang et al. elucidated the molecular mechanisms underlying melatonin's protective effects against cisplatin‐induced oxidative stress in testicular tissue. Their findings indicate that activation of MT1/MT2 receptors promotes the SIRT1/Nrf2 signaling pathway, resulting in enhanced antioxidant defenses within Leydig cells [35]. In summary, ramelteon exhibits potential modulatory effects on the immune system and tumor microenvironment. The combined action of these mechanisms may explain the observed association with lower mortality of ramelteon in patients with sepsis and cancer. The results of this study suggest the potential application of ramelteon in patients with sepsis and cancer, warranting further prospective studies and in‐depth mechanistic investigations.
In exploratory duration‐related analyses, ramelteon use for ≤ 7 days was associated with a lower risk of mortality, whereas no additional significant association was observed with longer treatment duration. This pattern should be interpreted cautiously, as treatment duration is intrinsically dependent on survival time: patients who die early have less opportunity to receive prolonged therapy, whereas those who survive longer are more likely to accumulate longer exposure. Therefore, survivor/time‐related or reverse‐causation bias may have contributed to the observed duration–outcome pattern, and the absence of an additional association beyond 7 days should not be interpreted as evidence of a definitive duration threshold. Subject to these limitations, the association observed with shorter courses remains biologically plausible and may be partly explained by the anti‐inflammatory and chronobiological properties of ramelteon. The early phase of sepsis is characterized by excessive inflammatory activation, immune dysregulation, oxidative stress, and acute organ dysfunction, all of which may contribute to increased mortality [36, 37]. In this context, ramelteon may have potential biological relevance through melatonin receptor‐mediated modulation of inflammatory and oxidative pathways. In addition, sepsis and critical illness can disrupt circadian organization through systemic inflammation, neuroendocrine dysregulation, sleep fragmentation, sedative exposure, mechanical ventilation, and altered light–dark cues in the ICU environment [38, 39]. Such circadian disruption may further aggravate immune dysregulation and impair physiological recovery. As a selective MT1/MT2 melatonin receptor agonist, ramelteon may help stabilize sleep–wake and circadian signaling pathways, thereby providing a plausible mechanistic link between melatonin signaling, circadian regulation, and inflammatory homeostasis [40, 41]. More broadly, circadian alignment has been increasingly recognized as an important regulator of metabolic and inflammatory homeostasis, providing additional mechanistic context for the potential relevance of chronobiological interventions in critical illness [42]. However, these mechanistic explanations remain speculative and should be validated in prospective and experimental studies. The absence of an additional significant association with ramelteon use beyond 7 days may be related to several factors. The initial inflammatory response, potentially modulated by concurrent therapeutic interventions, might subside after the first week, thereby diminishing any incremental advantage conferred by continued ramelteon administration [43, 44]. In addition, secondary infections, antibiotic resistance, cancer progression, or other late complications may become more dominant determinants of mortality beyond the initial phase of sepsis [45, 46]. Further research is warranted to clarify the relationship between ramelteon treatment duration and mortality outcomes.
This study has several key strengths. First, it evaluated the association between ramelteon use and mortality among critically ill patients with concurrent cancer and sepsis, a high‐risk population for whom evidence regarding ramelteon remains limited. Second, PSM reduced imbalance in measured baseline covariates between the ramelteon and non‐ramelteon groups, thereby improving group comparability. In addition, the exploratory evaluation of treatment duration and cumulative dose provided further insight into the exposure pattern of ramelteon in this population. However, several limitations should be acknowledged. First, because this study was based on a retrospective observational database, residual confounding and confounding by indication cannot be fully excluded despite the use of PSM and multivariable adjustment. PSM can reduce imbalance in measured covariates, but it cannot account for unmeasured or incompletely measured factors that may influence both the likelihood of receiving ramelteon and the risk of mortality. Patients who received ramelteon may have differed systematically from controls with respect to sleep disturbance, delirium risk, clinician prescribing preference, frailty, goals of care, evolving illness severity after ICU admission, infection source, cancer stage, active chemotherapy or immunotherapy status, vasopressor intensity, lactate levels, and other aspects of oncologic or sepsis severity. Several of these variables, including infection source, detailed cancer stage, active chemotherapy or immunotherapy status, and vasopressor dose, could not be reliably ascertained from the structured database fields used in this study. Serum lactate was also not incorporated into the primary analysis because of substantial missingness. Therefore, residual confounding related to sepsis severity and oncologic status cannot be fully excluded. Second, ramelteon exposure duration and cumulative dose were derived from prescription records, which may not perfectly reflect actual medication administration, patient adherence, temporary discontinuation, or missed doses. Therefore, exposure misclassification cannot be fully excluded. Moreover, the duration‐response findings may be affected by immortal time bias and survivor/time‐related bias because patients must survive long enough to initiate ramelteon and to accumulate longer treatment courses. Thus, longer treatment duration may partly reflect longer survival rather than a true duration‐dependent effect of ramelteon. As a result, the observed pattern of an apparent association within ≤ 7 days and no additional significant association beyond 7 days should not be interpreted as evidence of a definitive duration threshold. These analyses should therefore be considered exploratory and hypothesis‐generating rather than confirmatory. In addition, Schoenfeld residual diagnostics indicated that the proportional hazards assumption was not fully satisfied in some Cox models or covariates. Accordingly, the reported hazard ratios should be interpreted as average associations over the follow‐up period, or as weighted averages of potentially time‐varying hazard ratios, rather than as constant effects across the entire follow‐up period. Third, due to database limitations, we could not evaluate other clinically relevant outcomes, such as quality of life, delirium occurrence, sleep quality, or longitudinal organ function. Finally, the study population was derived from a North American critical care database, which may limit the generalizability of the findings to other healthcare systems and populations. Further validation in prospective and multicenter studies is warranted.
Despite these limitations, the results of this study provide preliminary evidence of an association between ramelteon use and lower mortality in patients with sepsis and cancer. Future research could focus on the following areas: conducting prospective randomized controlled trials to evaluate whether the observed association is causal in patients with sepsis and cancer, thereby validating our findings and clarifying the optimal dosage and timing of ramelteon; exploring the specific mechanisms by which ramelteon exerts its effects in patients with sepsis and cancer, such as how it influences the interaction between the immune system and the tumor microenvironment and how it modulates inflammation and oxidative stress; conducting multi‐center studies involving patients from different countries and regions to assess the generalizability of ramelteon efficacy; and examining the association between ramelteon use and other clinical outcomes such as quality of life, organ function, and healthcare resource utilization, in addition to survival outcomes.
In conclusion, ramelteon use was associated with lower 30‐day, 90‐day, ICU, and in‐hospital mortality among patients with sepsis and cancer after PSM and multivariable adjustment. However, because of the retrospective observational design, potential residual confounding, confounding by indication, and time‐related biases in duration analyses, these findings should be considered hypothesis‐generating. Prospective studies are needed to validate these associations and determine whether ramelteon has a causal effect on clinical outcomes in this population.
Author Contributions
Chendong Wang: conceptualization, writing – original draft, writing – review and editing. Jia Song: conceptualization, writing – review and editing. Mingjie Liu: conceptualization, formal analysis, data curation, writing – original draft, writing – review and editing.
Funding
The authors have nothing to report.
Ethics Statement
This study was a retrospective analysis based on the Medical Information Mart for Intensive Care IV (MIMIC‐IV) database (version 3.0). The establishment of the MIMIC‐IV database was approved by the Institutional Review Boards (IRBs) of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC). One author of this study obtained certified access to the database by completing the CITI (Collaborative Institutional Training Initiative) program's “Data or Specimens Only Research” course (Certificate Number: 66311890). This study did not involve a clinical trial; therefore, clinical trial registration was not applicable. All procedures were conducted in accordance with the Declaration of Helsinki.
Consent
As all patient data within the database is de‐identified, the requirement for individual patient informed consent was waived for this study.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Missing number for included variables in the datasets.
Table S2: Variance inflation factors of clinical variables.
Table S3: ICD code definitions used for malignancy ascertainment and subtype classification.
Table S4: Timing of ramelteon initiation relative to hospital admission.
Acknowledgments
We sincerely thank all contributors to the MIMIC‐IV cohort.
Data Availability Statement
The data for this study were sourced from the MIMIC‐IV database (version 3.0). Access to the original data is governed by a Data Use Agreement (DUA) that prohibits redistribution. Researchers can obtain access by completing the required certification and signing the DUA via the PhysioNet platform (https://physionet.org/content/mimiciv/). The authors confirm they had no special access privileges.
References
- 1. Echavarria I., Carrión Galindo J. R., Corral J., et al., “SEOM Clinical Guidelines for the Prophylaxis of Infectious Diseases in Cancer Patients (2021),” Clinical & Translational Oncology 24, no. 4 (2022): 724–732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Zhang W., Anyalebechi J. C., Ramonell K. M., et al., “TIGIT Modulates Sepsis‐Induced Immune Dysregulation in Mice With Preexisting Malignancy,” JCI Insight 6, no. 11 (2021): e139823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Li Y., Wu X., Song Y., et al., “The Short‐ and Long‐Term Readmission of Four Major Categories of Digestive System Cancers: Does Obesity or Metabolic Disorder Matter?,” Frontiers in Endocrinology 14 (2023): 1214651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Maruani J., Reynaud E., Chambe J., Palagini L., Bourgin P., and Geoffroy P. A., “Efficacy of Melatonin and Ramelteon for the Acute and Long‐Term Management of Insomnia Disorder in Adults: A Systematic Review and Meta‐Analysis,” Journal of Sleep Research 32, no. 6 (2023): e13939. [DOI] [PubMed] [Google Scholar]
- 5. Kast R. E., “Potential Benefits of Adding Alendronate, Celecoxib, Itraconazole, Ramelteon, and Simvastatin to Endometrial Cancer Treatment: The EC5 Regimen,” Current Issues in Molecular Biology 47, no. 3 (2025): 153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. de Oliveira H. M., Gallo Ruelas M., Zamora F. V., et al., “Efficacy of Ramelteon, Suvorexant, and Lemborexant for Delirium Prevention in Hospitalized Patients: A Systematic Review and Meta‐Analysis,” Critical Care Medicine 53 (2025): e1781–e1789. [DOI] [PubMed] [Google Scholar]
- 7. Osanai K., Kobayashi Y., Otsu M., Izawa T., Sakai K., and Iwashita M., “Ramelteon, a Selective MT1/MT2 Receptor Agonist, Suppresses the Proliferation and Invasiveness of Endometrial Cancer Cells,” Human Cell 30, no. 3 (2017): 209–215. [DOI] [PubMed] [Google Scholar]
- 8. Miyata R., Omasa M., Fujimoto R., Ishikawa H., and Aoki M., “Efficacy of Ramelteon for Delirium After Lung Cancer Surgery,” Interactive Cardiovascular and Thoracic Surgery 24, no. 1 (2017): 8–12. [DOI] [PubMed] [Google Scholar]
- 9. Mayanagi S., Haneda R., Inoue M., Ishii K., and Tsubosa Y., “Ramelteon and Suvorexant for Postoperative Delirium in Elderly Patients With Esophageal Cancer,” Esophagus 20, no. 4 (2023): 635–642. [DOI] [PubMed] [Google Scholar]
- 10. Ribeiro R. F. N., Santos M. R., Aquino M., de Almeida L. P., Cavadas C., and Silva M. M. C., “The Therapeutic Potential of Melatonin and Its Novel Synthetic Analogs in Circadian Rhythm Sleep Disorders, Inflammation‐Associated Pathologies, and Neurodegenerative Diseases,” Medicinal Research Reviews 45, no. 5 (2025): 1515–1539. [DOI] [PubMed] [Google Scholar]
- 11. Fink T., Glas M., Wolf A., et al., “Melatonin Receptors Mediate Improvements of Survival in a Model of Polymicrobial Sepsis,” Critical Care Medicine 42, no. 1 (2014): e22–e31. [DOI] [PubMed] [Google Scholar]
- 12. Cecon E., Izabelle C., Poder S. L., et al., “Therapeutic Potential of Melatonin and Melatonergic Drugs on K18‐hACE2 Mice Infected With SARS‐CoV‐2,” Journal of Pineal Research 72, no. 1 (2022): e12772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Han Y. Y., Tian Y., Zhao B. C., and Liu K. X., “Ramelteon Exposure and Survival of Critically Ill Sepsis Patients: A Retrospective Study From MIMIC‐IV,” BMC Anesthesiology 24, no. 1 (2024): 454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Sutton S. S., Magagnoli J., Cummings T. H., and Hardin J. W., “Melatonin Use and the Risk of 30‐Day Mortality Among US Veterans With Sepsis: A Retrospective Study,” Journal of Pineal Research 73, no. 2 (2022): e12811. [DOI] [PubMed] [Google Scholar]
- 15. Johnson A. E. W., Bulgarelli L., Shen L., et al., “MIMIC‐IV, a Freely Accessible Electronic Health Record Dataset,” Scientific Data 10, no. 1 (2023): 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Johnson A., Bulgarelli L., Pollard T., et al., “PhysioNet,” (2024), 10.13026/hxp0-hg59. [DOI]
- 17. Shankar‐Hari M., Phillips G. S., Levy M. L., et al., “Developing a New Definition and Assessing New Clinical Criteria for Septic Shock: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis‐3),” Journal of the American Medical Association 315, no. 8 (2016): 775–787. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Tian Y., Qin Z.‐S., and Han Y.‐Y., “Association Between Ramelteon and 30‐Day Mortality in ICU Patients With COVID‐19: A Retrospective Analysis of the MIMIC‐IV Database,” British Journal of Clinical Pharmacology 91 (2025): 2725–2735. [DOI] [PubMed] [Google Scholar]
- 19. Stekhoven D. J. and Bühlmann P., “MissForest—Non‐Parametric Missing Value Imputation for Mixed‐Type Data,” Bioinformatics 28, no. 1 (2012): 112–118. [DOI] [PubMed] [Google Scholar]
- 20. Shi Y., Tao T., Ye X., Ye B., Mi W., and Lou J., “Risk Factors for In‐Hospital Mortality in Surgical Patients With Abdominal Sepsis in China: A Nested Case–Control Study,” BMJ Open 15, no. 4 (2025): e092310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Fang Y., Dou A., Shen Y., et al., “Association of Triglyceride‐Glucose Index and Delirium in Patients With Sepsis: A Retrospective Study,” Lipids in Health and Disease 23, no. 1 (2024): 227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Stensrud M. J. and Hernán M. A., “Why Test for Proportional Hazards?,” Journal of the American Medical Association 323, no. 14 (2020): 1401–1402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Zhang L., Li F., Su X., et al., “Melatonin Prevents Lung Injury by Regulating Apelin 13 to Improve Mitochondrial Dysfunction,” Experimental & Molecular Medicine 51, no. 7 (2019): 1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. üstündağ h. and DoĞAnay S., “Melatonin‐Mediated Protection Against Sepsis‐Induced Organ Dysfunction,” in Septic Shock ‐ From Pathophysiology to Patient Care, ed. Shaikh N. (IntechOpen, 2025). [Google Scholar]
- 25. Aydin P., Aksakalli‐Magden Z. B., Civelek M. S., et al., “The Melatonin Agonist Ramelteon Attenuates Bleomycin‐Induced Lung Fibrosis by Suppressing the NLRP3/TGF‐Β1/HMGB1 Signaling Pathway,” Advances in Medical Sciences 68, no. 2 (2023): 322–331. [DOI] [PubMed] [Google Scholar]
- 26. Liu Y., Wang L., Du N., Yin X., Shao H., and Yang L., “Ramelteon Ameliorates LPS‐Induced Hyperpermeability of the Blood‐Brain Barrier (BBB) by Activating Nrf2,” Inflammation 44, no. 5 (2021): 1750–1761. [DOI] [PubMed] [Google Scholar]
- 27. Yang W., Zhang Y., Lu D., et al., “Ramelteon Protects Against Human Pulmonary Microvascular Endothelial Cell Injury Induced by Lipopolysaccharide (LPS) via Activating Nuclear Factor Erythroid 2‐Related Factor 2 (Nrf2)/Heme Oxygenase‐1 (HO‐1) Pathway,” Bioengineered 13, no. 1 (2022): 1518–1529. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yan H., Ge J., Gao H., Pan Y., Hao Y., and Li J., “Melatonin Attenuates AFB1‐Induced Cardiotoxicity via the NLRP3 Signalling Pathway,” Journal of International Medical Research 48, no. 10 (2020): 300060520952656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Frias‐Gomez J., Alemany L., Benavente Y., et al., “Night Shift Work, Sleep Duration and Endometrial Cancer Risk: A Pooled Analysis From the Epidemiology of Endometrial Cancer Consortium (E2C2),” Sleep Medicine Reviews 72 (2023): 101848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Menéndez‐Menéndez J. and Martínez‐Campa C., “Melatonin: An Anti‐Tumor Agent in Hormone‐Dependent Cancers,” International Journal of Endocrinology 2018 (2018): 3271948. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Mravec B., Horvathova L., and Hunakova L., “Neurobiology of Cancer: The Role of β‐Adrenergic Receptor Signaling in Various Tumor Environments,” International Journal of Molecular Sciences 21, no. 21 (2020): 7958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Ribeiro R. F. N., Pereira D., de Almeida L. P., Silva M. M. C., and Cavadas C., “SIRT1 Activation and Its Circadian Clock Control: A Promising Approach Against (Frailty in) Neurodegenerative Disorders,” Aging Clinical and Experimental Research 34, no. 12 (2022): 2963–2976. [DOI] [PubMed] [Google Scholar]
- 33. Wu M., Zhang J.‐B., Xiong Y.‐W., et al., “Promotion of Lung Cancer Metastasis by SIRT2‐Mediated Extracellular Protein Deacetylation,” Advanced Science 10, no. 3 (2023): e2205462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Margină D., Grădinaru D., Drăgoi C. M., Dumitrescu I.‐B., Nicolae A. C., and Ungurianu A., “Sirtuins and Melatonin: Linking Chronobiology to Inflammation and Aging,” in Advances in Geriatrics and Gerontology—Challenges of the New Millennium, ed. Palermo S. (IntechOpen, 2024). [Google Scholar]
- 35. Zhang J., Fang Y., Tang D., et al., “Activation of MT1/MT2 to Protect Testes and Leydig Cells Against Cisplatin‐Induced Oxidative Stress Through the SIRT1/Nrf2 Signaling Pathway,” Cells 11, no. 10 (2022): 1690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. He Y.‐J., Xu J.‐Q., Sun M.‐M., et al., “Glucocorticoid‐Induced Leucine Zipper: A Promising Marker for Monitoring and Treating Sepsis,” Frontiers in Immunology 11 (2020): 606649. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Wang T., Li Z., Xia S., Xu Z., Chen X., and Sun H., “The Protective Effects of Ramelteon Against Isoflurane‐Induced Insults and Inflammatory Response in Brain Microvascular Endothelial Cells,” Neurotoxicity Research 39, no. 3 (2021): 677–686. [DOI] [PubMed] [Google Scholar]
- 38. Mul Fedele M. L., Senna C. A., Aiello I., Golombek D. A., and Paladino N., “Circadian Rhythms in Bacterial Sepsis Pathology: What we Know and What we Should Know,” Frontiers in Cellular and Infection Microbiology 11 (2021): 773181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Ramsay S. D., Nenke M. A., Meyer E. J., Torpy D. J., and Young R. L., “Unveiling the Novel Role of Circadian Rhythms in Sepsis and Septic Shock: Unexplored Implications for Chronotherapy,” Frontiers in Endocrinology 16 (2025): 1508848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Hardeland R., “Redox Biology of Melatonin: Discriminating Between Circadian and Noncircadian Functions,” Antioxidants & Redox Signaling 37, no. 10–12 (2022): 704–725. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Cheng T.‐C., Lee Y.‐H., Chang Y.‐P., and Lee D.‐C., “Improved Urinary Melatonin Level as the Perspective Indicator That Leads to Better Sleeping Quality in Bladder Cancer Patients,” Therapeutic Advances in Urology 12 (2020): 1756287220940870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Zhang J., Chen Y., Zhong Y., et al., “Intermittent Fasting and Cardiovascular Health: A Circadian Rhythm‐Based Approach,” Science Bulletin 70, no. 14 (2025): 2377–2389. [DOI] [PubMed] [Google Scholar]
- 43. Gonzales G., Malka R., Bizios R., Dion G. R., and Guda T., “Burn Inhalation Injury and Intubation With Dexamethasone‐Eluting Endotracheal Tubes Modulate Local Microbiome and Alter Airway Inflammation,” Frontiers in Bioengineering and Biotechnology 13 (2025): 1524013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Xiong D., Geng H., Lv X., Wang S., and Jia L., “Inflammatory Response and Anti‐Inflammatory Treatment in Persistent Inflammation‐Immunosuppression‐Catabolism Syndrome (PICS),” Journal of Inflammation Research 18 (2025): 2267–2281. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Cao C., Yu M., and Chai Y., “Pathological Alteration and Therapeutic Implications of Sepsis‐Induced Immune Cell Apoptosis,” Cell Death & Disease 10, no. 10 (2019): 782. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Goldstein E., Olesen S. W., Karaca Z., Steiner C. A., Viboud C., and Lipsitch M., “Levels of Outpatient Prescribing for Four Major Antibiotic Classes and Rates of Septicemia Hospitalization in Adults in Different US States—A Statistical Analysis,” BMC Public Health 19, no. 1 (2019): 1138. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Missing number for included variables in the datasets.
Table S2: Variance inflation factors of clinical variables.
Table S3: ICD code definitions used for malignancy ascertainment and subtype classification.
Table S4: Timing of ramelteon initiation relative to hospital admission.
Data Availability Statement
The data for this study were sourced from the MIMIC‐IV database (version 3.0). Access to the original data is governed by a Data Use Agreement (DUA) that prohibits redistribution. Researchers can obtain access by completing the required certification and signing the DUA via the PhysioNet platform (https://physionet.org/content/mimiciv/). The authors confirm they had no special access privileges.