Preadmission Statin Use and 90-day Mortality in the Critically Ill: A Retrospective Association Study

Statins, which are 3-hydroxy-3-methylglutaryl-coenzyme-A inhibitors, are among the most commonly prescribed drugs worldwide.¹  They are known to lower cardiovascular risk by lowering serum cholesterol levels²  and also have antiinflammatory, antithrombotic, and immunomodulating effects.^3,4  Collectively, these effects are called the “pleiotropic effect,”⁵  which is known to improve outcomes of various diseases.⁶ 

Previous meta-analyses showed the mortality benefits of statin in patients with severe infection or sepsis.^7,8  However, some randomized clinical trials showed no benefit of statin therapy in patients with sepsis or septic shock,^9–11  and this was supported by a recent meta-analysis.¹²  Another meta-analysis showed that there was no adequate evidence on the mortality benefit of statin therapy in patients with acute respiratory distress syndrome (ARDS) or acute lung injury.¹³  Moreover, research on the mortality benefit of statin use, particularly in the critically ill, is lacking. Thus, whether statin use is associated with a lower mortality in critically ill patients requiring admission to the intensive care unit (ICU) remains unclear.

Therefore, this study aimed to examine the association between preadmission statin use and 90-day mortality in critically ill patients admitted to the ICU and whether this association differed according to statin type and dose.

This was a retrospective cohort study of ICU patients admitted to a tertiary-care academic medical center from 2012 to 2017. The study protocol was approved by the Institutional Review Board of Seoul National University Bundang Hospital, Seongnam, Korea (approval No.: B-1806/474-105). The need for informed consent was waived owing to the retrospective nature of the study.

The current study obtained data from the electronic health records stored in the Bundang Hospital Electronic System for Total Care at Seoul National University Bundang Hospital.¹⁴  Moreover, the dates of death of all patients as of May 16, 2018, including those lost to follow-up, were collected from the Statistics Korea database (Republic of Korea; http://kostat.go.kr/portal/eng, accessed November 30, 2018). As a central government organization for statistics, Statistics Korea provides services related to overall planning and coordination of national statistics. Causes of death were classified according to the Korean Standard Classification of Disease, Seventh Edition, a publication of Statistics Korea. To classify comorbidities of patients, we used the International Classification of Diseases, Tenth Revision, codes diagnostic system in the electronic health record system (Bundang Hospital Electronic System for Total Care) as shown in table S1 (Supplemental Digital Content 1, https://links.lww.com/ALN/B963).

The medical records of all adult patients (i.e., aged at least18 yr) admitted to all ICUs in a single tertiary academic hospital between January 2012 and December 2017 were analyzed. For patients who were admitted to the ICU more than once during the study period, only the last ICU admission was included, as this was considered to be the most severe. Patients with incomplete data on preadmission medication history were excluded from this study.

In Seoul National University Bundang Hospital, all hospitalized patients were required to report to their physician the type and amount of any medication that they were prescribed to take before admission. This preadmission medication history of all patients was recorded in an initial information record, which was also stored in an electronic health records system. Using the initial information record, we defined preadmission statin users as patients who had been prescribed statin and instructed to take it regularly for at least 1 month before admission. All other cases were classified as non–statin users. If the patient did not know the medication history from previous hospitalizations, it was confirmed by the pharmaceutical department. Statins were classified as atorvastatin, rosuvastatin, simvastatin, pitavastatin, or other statins (pravastatin, fluvastatin, and lovastatin) based on the drug component. The daily statin dose was classified as low-, moderate-, or high-intensity based on the guidelines of the American College of Cardiology/American Heart Association (table S2, Supplemental Digital Content 2, https://links.lww.com/ALN/B964).¹⁵  Statins were continued during hospitalization in almost all preadmission statin users who could take statin orally or by the enteral route via Levin tube. However, statin use was sometimes interrupted for patients who could not take statin via the oral or enteral route.

Data collected included (1) demographic characteristics (sex, age [yr], body mass index [kg/m²]); (2) socioeconomic status (type of insurance [medical aid program or national health insurance program], highest educational attainment, occupation, marital status); (3) Acute Physiology and Chronic Health Evaluation II score; (4) comorbidities at ICU admission (hypercholesterolemia [at least 200 mg/dl], hypertension, diabetes mellitus, history of coronary disease and cerebrovascular disease, liver disease, chronic obstructive lung disease, chronic kidney disease, anemia [less than 10 g/dl of hemoglobin], and cancer); (4) whether or not the patient was admitted via the emergency department; (5) the admitting department; and (6) preadmission cardiovascular medication (calcium channel blockers, β blockers, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, thiazides, furosemide, clopidogrel, and aspirin). Patients who are listed in the medical aid program are those who are classified to have low income, and most hospital charges of these patients are paid by the government. Meanwhile, for patients in the national health insurance program, the government covers approximately two thirds of their hospital charges.

The 90-day mortality rate was defined as the percentage of patients who died within 90 days of ICU admission. Considering that statin therapy is related to lower major adverse cardiovascular events,¹⁶  we classified the 90-day mortality rate into cardiovascular 90-day mortality and noncardiovascular 90-day mortality. Cardiovascular 90-day mortality was defined as death directly related to cardiovascular disease within 90 days of ICU admission. The Korean Standard Classification of Disease, Seventh Edition, codes for classifying cardiovascular disease related 90-day mortality are presented in table S3 (Supplemental Digital Content 3, https://links.lww.com/ALN/B965). Meanwhile, noncardiovascular 90-day mortality was defined as death due to any non–cardiac-related cause within 90 days of admission. The overall survival time was calculated from the date of ICU admission to the date of death or until May 16, 2018.

The primary endpoint of this study was 90-day mortality. The secondary endpoints were cardiovascular or noncardiovascular 90-day mortality.

Based on a previous retrospective study showing that preadmission statin therapy before sepsis development was associated with a 12% reduction in 90-day mortality,¹⁷  we considered a mortality reduction of at least 10% to be clinically useful. Before collecting data, we also assumed that at least 30% would be statin users and that 90-day mortality after ICU admission would be 40%; these assumptions were based on a recent national cohort study.¹⁸  To detect a 5% and 10% difference in the incidence of 90-day mortality between the statin users and non–statin users with a 0.05 chance of type 1 error and 80% power, 1,019 patients and 4,165 patients were needed, respectively, in this study. Multivariable adjustment using Cox regression models was initially performed, followed by a decision to include propensity score methods in response to peer review.

There were some patients whose height or weight could not be measured at ICU admission because they were severely ill, and as a result, some body mass index data (5,915 of 24,928, 23.7%) were missing. To lessen bias, we performed multiple imputation to replace the missing body mass index data using PROC MI in SAS (SAS Institute Inc., USA), before propensity score matching.¹⁹  The multiple imputations were performed five times to generate five datasets, and each missing value in the original dataset was replaced by the average of the five body mass index values in each of the five datasets. These five datasets were not included in the propensity score model and were not used to evaluate consistency of propensity score matching. Next, we performed propensity score matching without replacement.²⁰  All covariates were matched at a 1:2 ratio with caliper 0.1 via the nearest neighbor method. As statin was usually prescribed to prevent cardiovascular or cerebrovascular disease with various cardiovascular medications, statin users and non–statin users might have many different characteristics regarding comorbidities (e.g., hypertension) or a medication history that included another cardiovascular medication. Therefore, we decided to use a conservative caliper of 0.1 to match the two groups. To determine balance between the two groups before and after propensity score matching, absolute standardized difference was used. An absolute standardized difference of less than 0.1 for the covariates indicated that the two groups were sufficiently balanced.

After propensity score matching, Cox regression analysis was performed in the matched cohort to investigate hazard function in 90-day mortality between statin users and non–statin users. The three main independent variables (i.e., preadmission statin use, type of statin, and daily dose of preadmission statin) were analyzed using three separate Cox regression models to avoid multicollinearity in the Cox proportional hazard model. The results of the Cox regression analysis were presented as hazard ratio with 95% CIs.

The Grambsch and Therneau test was used to confirm that all central assumptions of the Cox proportional hazards were satisfied. In addition, we performed competing risk analysis for cardiovascular and noncardiovascular 90-day mortality in the matched cohort.²¹  The results of competing risk analysis were presented as cause-specific hazard ratios with 95% CIs.

For sensitivity analysis, we performed multivariable Cox regression analysis for 90-day mortality with all covariates in the entire cohort to demonstrate that the results in the matched cohort would be generalizable to all patients in our hospital. We also performed multivariable Cox regression analysis with all covariates in the surgical or medical ICU admission groups as a sensitivity analysis in the entire cohort. The 90-day survival in the propensity score-matched cohort for each group was shown using Kaplan-Meier curves, and a log-rank test was used for comparison between the groups. The packages of “MatchIt,” “Cmprsk,” and “Survival” of the R program (Version 3.4.4; www.r-project.org; accessed March 15, 2019) were used for propensity score matching, competing risk analysis, and Kaplan-Meier estimation, respectively. All other statistical analyses were performed with SAS software version 9.4 (SAS Institute Inc.), and P < 0.05 was considered significant.

Between January 2012 and December 2017, a total of 30,398 patients were admitted to the ICU for a total of 40,533 times. Of these, 10,135 ICU admissions were excluded because they were admitted more than once; only the last ICU admission of any repeatedly admitted patient was included in the analysis, as this was considered to be the most severe. Next, 5,440 pediatric patients aged under 18 yr and 30 patients (0.001%, 30 of 24,958) with incomplete or missing data on preadmission medication history in initial information records were excluded. There were no additional missing data, with the exception of body mass index data (5,915 of 24,958, 23.7%), at ICU admission in our electronic health record system. Thus, 24,928 patients were included in the analysis. Of these, 7,396 were preadmission statin users (29.7%), and 17,532 were non–statin users (70.3%). After propensity score matching, 5,354 statin users and 7,758 non–statin users were included in the final analysis (fig. 1). There was a difference between the intended matching ratio (1:2) and actual matching ratio (approximately 1:1.45) because the matching algorithm was set to the nearest neighbor method with a caliper of 0.1. After propensity score matching, all covariates were well balanced with an absolute standardized difference of less than 0.1 (table 1). Figure S1 (Supplemental Digital Content 4, https://links.lww.com/ALN/B966) shows the distribution of propensity scores before (A) and after (B) propensity score matching. Notably, the distribution of propensity scores between the two groups became similar after propensity score matching.

Table 2 shows the results of survival analysis after propensity score matching. The 90-day mortality rate was significantly higher in non–statin users (918 of 7,758; 11.8%) than that in statin users (455 of 5,354; 8.5%; P < 0.001). In Cox regression analysis, the risk of 90-day mortality among statin users was 30% lower than among non–statin users (hazard ratio: 0.70, 95% CI: 0.63 to 0.79; P < 0.001; model 1). The risks of 90-day mortality in atorvastatin users and rosuvastatin users were 28% (hazard ratio: 0.72, 95% CI: 0.63 to 0.82; P < 0.001; model 2) and 42% (hazard ratio: 0.58, 95% CI: 0.47 to 0.72; P < 0.001; model 2) lower, respectively, than the rate in non–statin users. Additionally, the risk of 90-day mortality among patients with daily low-, moderate-, and high-dose statin use was 26% (hazard ratio: 0.74, 95% CI: 0.65 to 0.84; P < 0.001; model 3), 36% (hazard ratio: 0.64, 95% CI: 0.53 to 0.78; P < 0.001; model 3), and 33% (hazard ratio: 0.67, 95% CI: 0.49 to 0.91; P = 0.010; model 3) lower, respectively, than that of non–statin users.

Table 3 shows the results of competing risk analysis for cardiovascular and noncardiovascular 90-day mortality after propensity score matching. The risk of cardiovascular 90-day mortality was not significantly different between statin and non–statin users (P = 0.253). However, the risk of noncardiovascular 90-day mortality in statin users was 32% lower than that in non–statin users (hazard ratio: 0.68, 95% CI: 0.60 to 0.78; P < 0.001; model 1). Cumulative hazard functions for cardiovascular and noncardiovascular 90-day mortality in non–statin users and statin users are presented in figure 2, A and B.

Figure 3 shows the Kaplan–Meier curve results of the cumulative percentage of surviving patients at 90 days according to statin use (fig. 3A), type of statin (fig. 3B), and daily dosage of statin (fig. 3C) after propensity score matching. The 90-day cumulative survival rate was lower in non–statin users than in statin users (fig. 3A; P < 0.001).

Table 4 shows the results of multivariable Cox regression analysis for sensitivity in the entire cohort. The risk of 90-day mortality was 26% lower in statin users than in non–statin users (hazard ratio: 0.74, 95% CI: 0.67 to 0.83; P < 0.001; model 1). Table 5 shows the results of multivariable Cox regression analysis with all covariates in the surgical or medical ICU admission groups. In the medical admission group, the risk of 90-day mortality was 38% lower in statin users than in non–statin users (hazard ratio: 0.62, 95% CI: 0.66 to 0.90; P < 0.001), while it was not significantly different in the surgical ICU admission group (P = 0.065).

This study showed that preadmission statin use was associated with a lower 90-day mortality in ICU patients. This association was particularly strong in the rosuvastatin use group, while there was no specific difference in associations between daily dose of statin and 90-day mortality. In the competing risk analysis, statin use was more strongly correlated with lower incidence of noncardiovascular 90-day mortality than cardiovascular 90-day mortality. In addition, the results of sensitivity analysis showed that a lower 90-day mortality among statin users was similar in both the propensity score–matched cohort and the entire cohort, and this association was more evident in the medical ICU population.

The results of our study differ from those of recently published meta-analyses that reported no mortality benefit of statin therapy in patients with sepsis or acute lung injury/ARDS.^12,13  However, our results are similar to those of a recently published observational study.¹⁷  The differences between our study and the meta-analyses might be explained by differences in the design of the included studies. First, most randomized clinical trials focused on the effect of statin administration on outcomes of sepsis or acute lung injury/ARDS patients after admission.^12,13  However, our study and the study by Lee et al.¹⁷  focused on the effect of statin administration before hospital admission on mortality in critically ill patients and patients with sepsis, respectively. Patients who were defined as preadmission statin users in observational studies are more likely to have taken long-term statin therapy, which could affect the results of our study. Second, while we focused on the effect of statin therapy in a mixed ICU population, previous randomized clinical trials focused specifically on patients with acute lung injury/ARDS or sepsis. The mixed ICU population in this study may have been healthier overall or experienced more self-limiting disease processes (i.e., postoperative admission) than other ICU populations, such as patients with sepsis or acute lung injury/ARDS; this could have resulted in the relatively lower 90-day mortality observed in this study compared to that reported in other studies.^12,13  Therefore, the differences in outcomes between this study and previous studies might be attributed to the characteristics of different ICU populations.

There are some differences between our study and that by Lee et al.¹⁷  First, we focused on a mixed ICU population admitted to a single tertiary care hospital from 2012 to 2017, whereas the study by Lee et al. focused on patients with sepsis in a national cohort study in Taiwan from 1999 to 2011. Second, while we defined statin users as patients who had been prescribed regular doses of statins for at least 1 month before ICU admission, Lee et al. defined statin users as patients who received statin therapy during hospitalization.¹⁷  Therefore, it is unclear whether the patients in the study by Lee et al. took statins before hospitalization. There have been other observational studies that focused on general ICU patients, as we have done in the current study. Al Harbi et al. reported a lower hospital mortality for ICU patients who received statin therapy while admitted to the ICU,²²  and Christensen et al. reported that preadmission statin use was associated with a lower one-year mortality.²³  However, Beed et al. reported that there was no association between previous statin use and survival outcome after ICU admission.²⁴  These studies did not consider the effects of different types of statins²⁵  or the daily dose.¹⁵  This is an important issue given that the antiinflammatory effects of statins differ according to their lipophilicity²⁶  and that clinical outcomes differ according to statin dosage.²⁷ 

Interestingly, rosuvastatin, which is a hydrophilic statin, showed the strongest association with 90-day mortality. One possible reason might be the relatively strong potency of rosuvastatin. Although there is still a debate about which statins are the most effective, rosuvastatin is known to have a highly potent cholesterol-lowering effect.²⁸  In one randomized clinical trial of patients with acute coronary syndrome, rosuvastatin had a more pronounced effect than atorvastatin and rosuvastatin.²⁹  Considering this result, it is possible that patients who had taken rosuvastatin before ICU admission received stronger antiinflammatory and cholesterol-lowering effects from the statin. Another possible point to consider is that this study was conducted in a Korean population. The oral clearance of rosuvastatin is 44% lower in Asians than in Caucasians.³⁰  A recent randomized clinical trial reported that polymorphism in the SLCO₁B₁ and ABCG₂ genes is associated with ethnic differences in the pharmacokinetics of rosuvastatin.³¹  Therefore, plasma exposure to rosuvastatin could be greater in Asians than in Caucasians, and this could have affected the results of this study in a Korean population. Nevertheless, the use of rosuvastatin, which has high potency and few adverse effects, remains a controversial issue, and there is still ongoing debate on whether it helps to lower mortality.

Another important issue is the relationship of mortality with daily dose. In our study, to investigate whether 90-day mortality in critically ill patients was affected by the daily dose of various types of statin, we used the classification suggested in the American College of Cardiology/American Heart Association guidelines.¹⁵  A recent cohort study reported that the intensity of statin therapy is associated with mortality in patients with atherosclerotic cardiovascular disease,³²  while another recent cohort study reported that there was no significant association between statin intensity and mortality in patients with chronic kidney disease.³³  The inconsistent findings between these studies might be explained by two reasons. First, the intensity of statin therapy was validated based on its cholesterol-lowering effect. The application of the concept of dose-dependency to the antiinflammatory and immunomodulative effects of statins is still debated. One previous study reported that the dose dependency of the antiinflammatory or immunomodulative effects of atorvastatin differed according to individual biologic effects.³⁴  Second, the preadmission statin dosage for most patients was the daily dosage prescribed by their individual physicians. Thus, most patients would have been prescribed the optimal dosage based on their personal characteristics and condition, and this may explain the absence of any effect of daily dosage on outcomes such as 90-day mortality.

The results of the competing risk analysis are also notable in this study. The association between lower 90-day mortality and statin use was more evident for noncardiovascular 90-day mortality than for cardiovascular 90-day mortality. This might be explained by the following mechanisms. First, the effect of statin therapy on reducing any cardiovascular event or mortality could differ according to age or comorbidities. A recently published observational study reported that statin therapy was not associated with a lower cardiovascular disease or all-cause mortality in people without diabetes older than 74 yr.³⁵  Second, statin therapy exerts its lowering effect on cardiovascular mortality through a preventive mechanism; in other words, it lowers blood cholesterol levels.³⁶  Therefore, because we evaluated a relatively short-term mortality (i.e., 90 days) after ICU admission in this study, the preventive effect of statin therapy on cardiovascular mortality might be limited.

The results of the sensitivity analysis also showed that the association between statin use and lower 90-day mortality in the ICU population was more evident in the medical ICU population than in the surgical ICU population. First, statin therapy might be discontinued in surgical ICU patients because they will be placed on nothing per orem status as preoperative preparation. Considering that early perioperative exposure to statin was associated with lower all-cause mortality,³⁷  its discontinuation in the surgical population might affect the results. Second, patients in the medical ICU have relatively more severe illness than patients in the surgical ICU. Thus, the pleiotropic effect of statin on 90-day mortality might be stronger in medical ICU patients.

Our study had several limitations. First, given its retrospective observational design, there were many confounders that needed to be controlled by propensity score matching or multivariable adjustment. Second, there could be issues with generalizability because this study was conducted at a single center. Third, because the duration of statin therapy during the preadmission period varied, it could not be evaluated. Fourth, although we performed propensity score matching to adjust for confounders in this study, there might be unmeasured confounders and selection bias. For example, severity of disease was not sufficiently measured or included in the propensity score model, and thus might affect statin use in this study. Moreover, the patients who received statins, despite having similar comorbidities after propensity score matching, might be receiving a different quality of outpatient care or other macro-level health care. Although we included socioeconomic status before hospital admission, the preadmission health care quality was not considered as a variable in this study, and this could be a limitation. Last, since we focused on preadmission statin usage among critically ill patients, inpatient statin administration was not evaluated in this study. Therefore, adherence of statin usage in statin users was also unclear during hospitalization among patients in this study. Similarly, the absence of any data regarding in-hospital care might be a limitation of this study.

In conclusion, preadmission statin use was associated with a lower 90-day mortality among critically ill patients. This association was stronger in rosuvastatin users and medical ICU patients. Furthermore, this association appeared to be stronger in noncardiovascular 90-day mortality. No differences according to daily dosage intensity were observed.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.