A Population-Based Cohort Study on the Drug-Specific Effect of Statins on Sepsis Outcome

Accepted Manuscript A population-based cohort study on the drug specific effect of statins on sepsis outcome Chien-Chang Lee, MD, ScD, Meng-tse Gabriel Lee, PhD, Tzu-Chun Hsu, BSc, Lorenzo Porta, MD, Shy-Shin Chang, MD, PhD, Chia-Hung Yo, MD, Kuang-Chau Tsai, MD, MSc, Matthew Lee, PhD, JD PII:

S0012-3692(17)32803-9

DOI:

10.1016/j.chest.2017.09.024

Reference:

CHEST 1353

To appear in:

CHEST

Received Date: 17 April 2017 Revised Date:

16 July 2017

Accepted Date: 6 September 2017

Please cite this article as: Lee CC, Lee MtG, Hsu TC, Porta L, Chang SS, Yo CH, Tsai KC, Lee M, A population-based cohort study on the drug specific effect of statins on sepsis outcome, CHEST (2017), doi: 10.1016/j.chest.2017.09.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

A population-based cohort study on the drug specific effect of statins on sepsis outcome

3 4 5

Chien-Chang Lee1 MD, ScD, Meng-tse Gabriel Lee1 PhD, Tzu-Chun Hsu1 BSc, Lorenzo Porta2 MD, Shy-Shin Chang3 MD, PhD, Chia-Hung Yo4 MD, Kuang-Chau Tsai4 MD, MSc , Matthew Lee5 PhD, JD.

6 7

1

Department of Emergency Medicine, National Taiwan University Hospital, Taipei, Taiwan 2

Dipartimento di scienze Biomediche e Cliniche, Ospedale "L. Sacco", Università degli Studi di Milano, Milan, Italy.

SC

8 9 10 11 12

RI PT

1 2

3

Department of Family Medicine, Taipei Medical University Hospital and School of Medicine, Taipei Medical University, Taipei, Taiwan

13 14

4

15 16 17 18 19 20

5

21 22 23 24 25

Funding: This study is supported by the Taiwan National Science Foundation Grant NSC 102-2314-B-002 -131 -MY3; Taiwan National Ministry of Science and Technology Grants MOST 104-2314-B-002 -039 -MY3, and MOST 105-2811-B-002-031. No funding bodies had any role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

26 27 28

Running title: Drug specific effect of statins

M AN U

Department of Emergency Medicine, Far Eastern Memorial Hospital, New Taipei City, Taiwan Medical Wisdom Consultants Inc. Houston, USA

AC C

EP

TE D

* Correspondence and address reprint request to: Dr. Chien-Chang Lee, Department of Emergency Medicine, National Taiwan University Hospital, No. 7, Chung-Shan South Road, Taipei 100, Taiwan. E-mail: [email protected]

Manuscript words: 3395

1

29

ABBREVIATION LIST

30

PS: Propensity score

31

CI: Confidence Interval

32 33

ICD-9: International Classification of Disease version 9 PS: Propensity Score

AC C

EP

TE D

M AN U

SC

34

RI PT

ACCEPTED MANUSCRIPT

2

ACCEPTED MANUSCRIPT

Abstract

RI PT

Background: Whether statin treatment, proved by recent experimental studies to have an antimicrobial activity, exerts a drug or a class specific effect in sepsis remains unknown.

SC

Methods: Short-term mortality in sepsis patients was analyzed using data from the National Health Insurance Research Database. Use of statins was defined as the cumulative use of a specific statin (atorvastatin, simvastatin or rosuvastatin) for more than 30 days prior to the index sepsis admission. We determined the association between statin and sepsis outcome by multivariate-adjusted Cox models and propensity score (PS) matched analysis, using a 1:1:1 PS matching technique.

M AN U

Results: A total of 52,737 sepsis patients fulfilled the inclusion criteria, of which 1855 were prescribed atorvastatin, 916 simvastatin, and 732 rosuvastatin. Compared with nonusers, simvastatin (Hazard Ratio [HR] 0.72, 95% CI 0.58-0.90) or atorvastatin (HR 0.78, 95% CI 0.68-0.90) were associated with an improved 30-day survival, while rosuvastatin was not (HR 0.87, 95% CI 0.73-1.04). Using rosuvastatin as the reference, atorvastatin (HR 0.79, 95% CI 0.64-0.99) and simvastatin (HR 0.77, 95% CI 0.59-0.99) had superior effectiveness in preventing mortality.

EP

TE D

Conclusions: Compatible with in vitro experimental findings, our results suggest that the drug specific effect of statins on sepsis, not correlated to their lipid-lowering potency.

AC C

35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57

3

ACCEPTED MANUSCRIPT

Introduction

60

Sepsis is defined as a life-threatening organ dysfunction caused by a dysregulated host

61

response to an infection. Despite major advancements in medical care, the mortality of

62

sepsis remains high, with a case-fatality rate around 17 to 26%.1-4 Empirical

63

antibiotic and organ supportive care have been the cornerstones of sepsis management.

64

However, the global escalation in antimicrobial-resistant bacteria is increasingly

65

compromising the effectiveness of antimicrobial therapy. Thus, therapies that can

66

attenuate the dysregulated immune response in sepsis have been intensively studied in

67

the past decade.

68

Due to the pleiotropic immunomodulatory effects of statins demonstrated in animal

69

models of sepsis, several clinical observational studies and randomized controlled trials

70

have investigated whether the use of statins could improve the outcome of sepsis in

71

humans.5-8 Most observational studies have found that preadmission use of statin

72

improves the outcome of sepsis and a meta-analysis of 27 observational studies has

73

concluded that statin treatment can decrease sepsis mortality by 35% (RR, 0.65; 95% CI,

74

0.57 to 0.75).9-11 In contrast to observational studies, randomized control trials, however,

75

did not observe beneficial effects of post-admission use of statins in sepsis patients.12-16

76

It has been controversially debated if the observed protective effect of preadmission use

77

of statin was due solely to the healthy user effect. Healthy-user effect describes the

78

phenomenon in which patients prescribed with statins tend to have a constellation of

79

healthier behaviors and, thus, improved outcomes in most systemic diseases.

80

Just when the enthusiasm in the potential role of statins in improving sepsis outcome

AC C

EP

TE D

M AN U

SC

RI PT

58 59

4

ACCEPTED MANUSCRIPT

has tempered, emerging new evidence demonstrates that statins may have direct

82

antibacterial effects and modulate the bacterial virulence.17-19 The antibacterial or anti-

83

virulence effects may be statin-specific, not directly correlating with their lipid-lowering

84

potency.17,20 For example, while simvastatin was associated with better antibacterial

85

effects than rosuvastatin, the latter was found to have a more potent lipid-lowering

86

capacity.21,22 Thus, the primary aim of this research is to determine if the protective

87

effects of statins are drug or class specific. In addition, a head to head comparison on

88

different types of statins may provide insights on whether the protective effect of statins

89

is solely due to the healthy user effect.

M AN U

SC

RI PT

81

90 91

Methods

Data source

93

Taiwan’s National Health Insurance (NHI) program is a single-payer government operated

94

compulsory health insurance program, which covers over 98.4% of the 23 million people

95

that reside in Taiwan. Specific data subsets were constructed for different research

96

purposes. We used the year 2000 version of the Longitudinal Health Insurance Database

97

(LHID) for this analysis, which used a systematic approach to randomly sample 1 million

98

beneficiaries from the Registry for Beneficiaries of the NHIRD at 2000 to ensure the

99

selected sample represents demographic and geographic region distribution of the

100

entire Taiwanese population. The selected participants were followed from 2000 to 2011

101

to form a longitudinal close cohort for research use. The longitudinal nature of LHID

102

permits to identify a cohort based on diagnoses, health services and drugs utilization, to

AC C

EP

TE D

92

5

ACCEPTED MANUSCRIPT

track medical history, to establish a prescription drug profile, and to determine the

104

endpoint of drug treatments. The LHID included detailed information on patient

105

demographics, inpatient and outpatient electronic claims records, individual diagnoses,

106

surgical and medical operations, dispositions, and detailed data on prescribed

107

medication, such as brand/generic name of the prescribed drugs, route, quantity and

108

number

109

National Taiwan University Hospital Research Ethics Committee (IRB No.201311044RINB)

110

and patients’ consent was waived for this anonymized electronic research database.

111

Study cohort and identification of sepsis patient

112

The study cohort consisted of all the LHID participants hospitalized for sepsis between

113

year 2001 and 2011. Year 2000 was used for assessment of covariates and statin use of

114

sepsis patients identified in 2001. Compatible with the sepsis-3 definition, sepsis was

115

defined as a life-threatening organ dysfunction caused by a dysregulated host response

116

to an infection. However, as the laboratory test results for Sequential [Sepsis-related]

117

Organ Failure Assessment (SOFA) score were not available in a health claims database,

118

we identified sepsis cases using a validated International Classification of Diseases – 9th

119

revision- Clinical Modification (ICD-9-CM) coding system proposed by Angus DC,3 in

120

which at least one acute organ dysfunction and a diagnosis of bacterial or fungal

121

infections is required to define an episode of sepsis. Acute organ dysfunctions used for

122

this study were cardiovascular/shock, respiratory, central nervous system, hematologic,

123

hepatic, renal and metabolic system dysfunctions. Supplementary eAppendix 1 lists the

124

ICD-9-CM codes used to identify patients with bacterial and fungal infections and acute

days

of

administration.

This

study

was

approved

by

AC C

EP

TE D

M AN U

SC

of

RI PT

103

6

ACCEPTED MANUSCRIPT

organ dysfunctions. We defined index date as the first day of an emergency department

126

or hospital visit due to sepsis within one given year, and recurrent sepsis admissions in

127

the same year were not considered in our analysis. Patients were followed from the

128

index admission date to the occurrence of death, termination of health insurance

129

coverage or the end of the study period, whichever came first.

130

Outcomes and covariates

131

The primary endpoint for the analysis is 30-day all-cause mortality; the secondary

132

endpoints are 90-day all-cause mortality and acute respiratory failure. Acute respiratory

133

failure was defined by the requirement of ventilation during the index sepsis admission

134

episode. Based on literature review, we collected a total of 61 covariates in the following

135

dimensions: demographics, sources of infection, pre-existing comorbidities, proxies for

136

lifestyle factors, health care facilities utilization and use of specific medications. Chronic

137

comorbidities and risk factors for sepsis were collected from 1999 to the index sepsis

138

event, and exposure to specific medications including statin was assessed one year prior

139

to the index sepsis event

140

Medication exposure

141

In this study, we took three statins into account: atorvastatin, rosuvastatin, and

142

simvastatin. We defined statin use by having a prescription record of statin ≥ 30 days

143

duration in the 1-year period before the index sepsis event.

144

Data Analysis

145

Patients’ baseline characteristics were presented as percentages for categorical

146

variables, and means with standard deviations for continuous variables. To examine the

AC C

EP

TE D

M AN U

SC

RI PT

125

7

ACCEPTED MANUSCRIPT

differences in patient characteristics, we used Mann-Whitney U tests for continuous

148

variables and Pearson chi-square tests for dichotomous variables. Cox proportional

149

hazard regression models were used to assess the association between statins and

150

mortality. We tested the proportional hazard assumption by introducing an interaction

151

term of exposure and follow-up time in the model. In addition, we confirmed the

152

assumption of proportional hazards by an examination of the log (minus log) curves.

153

To consolidate the strength of our findings, three methods were used to calculate hazard

154

ratios. The first method obtains an unadjusted estimate, and the second adjusted for all

155

individual covariates in the Cox models. In the third methods, we conducted a

156

propensity score (PS) matched analysis, using the greedy matching algorithm.23 We built

157

a PS to predict statin use in a full cohort and then applied the PS to different subgroup

158

analysis. The validity of this approach has been proved by an empirical and simulation

159

study.24 PS was defined as the conditional probability of any statin use derived from the

160

logistic regression model that includes all potential predictors of statin prescription.

161

Supplementary eAppendix 2 lists the component variables of the PS model and their

162

corresponding weights. To verify baseline covariates were balanced after PS matching,

163

we calculated standardized differences of baseline covariates for the simvastatin-

164

atorvastatin, simvastatin-rosuvastatin, and atorvastatin-rosuvastatin pairs before and

165

after PS matching. Values less than 10% were considered to be sufficiently balanced.

166

Results were presented with a standardized difference plots. Finally, we conducted a

167

1:1:1 propensity score (PS)-matched analysis using the triad optimized nearest-neighbor

168

matching algorithm developed by Rassen et al.25 Each triad was selected using a distance

AC C

EP

TE D

M AN U

SC

RI PT

147

8

ACCEPTED MANUSCRIPT

function defined by the perimeter of a triangle with the maximum allowable propensity-

170

score distance between patients set at 0.05. We plotted the cumulative hazard plots for

171

mortality in the original and PS-matched cohort according to three different statin users.

172

The survival difference between of three types of statin users were tested by a log rank

173

test. All statistical analysis was performed using SAS 9.4 for Windows (SAS Institute Inc,

174

Cary, NC), except for the cumulative hazard graphs, which were plotted with

175

the survminer package in R-statistical software. A two-sided P value of 0.05 was deemed

176

as significant for all analysis.

M AN U

SC

RI PT

169

177

Result

Cohort Enrollment and Baseline Characteristics

179

We identified 52,737 individuals with sepsis, who met the inclusion criteria. The process

180

of cohort assembling was demonstrated by a flow chart in Figure 1. In this cohort, 9188

181

patients died at 30 days (17.4 %) and 11953 patients died at 90 days (22.7%) after

182

hospital admission. 3598 patients (6.82%) received statin treatment before the index

183

sepsis episodes, of which 916 received simvastatin, 1855 received atorvastatin, and 732

184

received rosuvastatin. 95 patients used more than 1 types of statins and were excluded

185

from analysis.

186

Characteristics of the three types of statin users are presented in Table 1. There were no

187

significant differences in the demographics, living areas, and major sources of infection

188

among the three types of statin users. Rosuvastatin users are more likely to pay higher

189

insurance premium and to have higher comorbidity score, history of chronic pulmonary

190

diseases, fluid and electrolytes disorders and liver diseases than simvastatin or

AC C

EP

TE D

178

9

ACCEPTED MANUSCRIPT

atorvastatin users. While simvastatin users are more likely to have lower combined

192

comorbidity score, congestive heart failure, renal failure, cardiac arrhythmias and

193

chronic pulmonary disease than the other two types of statins, atorvastatin users are

194

more likely to have a history of alcohol-related diseases and complicated diabetes.

195

Association between preadmission statin treatment and sepsis outcome

196

Compared to patients who have never received statin treatment prior to developing

197

sepsis, the use of statins before sepsis development was associated with a decreased

198

risk of 30-day mortality in crude (HR 0.84, 95% CI: 0.77-0.91), confounder-adjusted (HR

199

0.86, 95% CI:0.78-0.94), and propensity score matched (HR 0.88, 95% CI:0.78-0.99)

200

analysis (Table 2). The beneficial effect of statins was attenuated, but remains significant

201

for 90 day-mortality (PS-matched HR: 0.93, 95% CI: 0.88-0.99).

202

Next, we compared use of individual type of statins to non-use (Table 2). Simvastatin

203

tends to associate with the best outcome, followed by atorvastatin and rosuvastatin.

204

Simvastatin and atorvastatin were associated with a decreased 30-day mortality, while

205

rosuvastatin were not. For 90-day mortality, all three types of statins were associated

206

with a reduced mortality in the PS-matched analysis, but the differences in the three

207

types of statins were attenuated.

208

Head to head comparison of individual class of statins

209

To further analyze the comparative beneficial effect among the three types of statins, we

210

performed confounder adjusted and PS-matched pairwise comparison on mortality and

211

acute respiratory failure. Results of pairwise comparison were summarized in table 3.

212

Using 1:1:1 PS matching, we matched 536 simvastatin users with 536 atorvastatin and

AC C

EP

TE D

M AN U

SC

RI PT

191

10

ACCEPTED MANUSCRIPT

536 rosuvastatin users. The standardized difference plot showed minimal standardized

214

differences between the three groups after matching (Supplementary efigure 1).

215

The effect estimates on the 30-day and 90-day mortality endpoints were very similar.

216

However, due to more deceased outcomes, there were more significant differences

217

between various types of statins users at the 90-day mortality endpoint. Even though

218

there was no significant discrepancy between simvastatin and atorvastatin, these two

219

statins consistently showed a superior beneficial effect over rosuvastatin in both

220

confounder-adjusted and PS-matched analysis. In the PS-matched comparison cohort,

221

use of simvastatin was associated with a 23% reduction (HR 0.77; 95%CI: 0.59-0.99), and

222

use of atorvastatin was associated with a 21% reduction (HR 0.79; 95% CI: 0.64-0.99) of

223

90-day mortality, compared with rosuvastatin. For acute respiratory failure, no

224

significant difference was found for any pair of comparison.

225

We created cumulative hazard plots according to three types of statin users in the

226

original and PS-matched cohorts. In the original cohort, rosuvastatin users consistently

227

had a higher cumulative probability of death at 30 (Figure 2 A, log-rank test, P= 0.013) or

228

90 days (Figure 2 B, log-rank test, P= 0.0025) than atorvastatin or simvastatin users. In

229

the PS-matched cohort, rosuvastatin users remained to have a significantly higher

230

cumulative mortality at 30 (Figure 2 C, log-rank test, P= 0.0072) or 90 days (Figure 2 D,

231

log-rank test, P= 0.0048) when compared to atorvastatin or simvastatin users.

AC C

EP

TE D

M AN U

SC

RI PT

213

232 233

Discussion

234

In this population based study, chronic statin treatment prior to sepsis development was

11

ACCEPTED MANUSCRIPT

associated with a 12% reduction in 30-day mortality in comparison to no prior statin

236

treatment. When compared to non statin users, Simvastatin was associated with the

237

most pronounced reduction in the 30-day mortality rate (28%), followed by atorvastatin

238

(22%) and lastly by rosuvastatin (13%) of which the association was not significant. The

239

beneficial effect of statins was attenuated at 90 days, but remains significant for the

240

three types of statins. By applying 1:1:1 PS matching, we were able to make a head to

241

head comparison among the three types of statins. In comparison to rosuvastatin,

242

simvastatin and atorvastatin were associated with a significantly lower 30-day mortality.

243

Even though in a previous study Ou et al. demonstrated that high potency statins

244

(rosuvastatin, atorvastatin, and simvastatin) had better survival than low potency statins

245

(fluvastatin, lovastatin, pitavastatin, pravastatin), they did not perform a head to head

246

comparison of the individual type of statins. While Ou et al. study used one-year

247

mortality as the main outcome measure to investigate the chronic cardiovascular

248

complications of sepsis, we used 30-day mortality, as we were more interested in the

249

potential pleiotropic effects of individual statins on the acute phase of sepsis.

250

Furthermore, Ou et al. might have missed up to 70% of severe sepsis patients since they

251

identified septicemia patients (ICD-9 CM code: 038.9) without including acute organ

252

dysfunction.26,27 In fact, this approach lacks sensitivity, as the clinical diagnosis of

253

septicemia requires a positive microbiological culture, which is only present in 20-40% of

254

sepsis patients.28,29 To avoid this problem, we used the modified comprehensive coding

255

system proposed by Angus DC, in which ICD-9CM codes for both infectious diseases and

256

acute organ dysfunctions are required to identify sepsis. The approach mimics the

AC C

EP

TE D

M AN U

SC

RI PT

235

12

ACCEPTED MANUSCRIPT

clinical diagnosis and had been shown to have nearly three folds higher sensitivity than

258

the septicemia approach in a validation study using chart review.30

259

Previous experimental studies using animal model of sepsis, found that pre-treatment

260

with statins can improve sepsis survival. Specifically, pre-treatment with simvastatin has

261

been found to improve sepsis survival through preservation of cardiac function,

262

attenuation of inflammatory cytokines, attenuation of neutrophil infiltration in the lung,

263

and inhibiting T-cell dysfunction.31-33 Recently, specific statins were also associated with

264

direct anti-microbial and anti-virulence effects.17-19 Of note, simvastatin was shown by

265

several reports to have the most potent antibacterial activity. The antimicrobial

266

spectrum includes methicillin-sensitive Staphylococcus aureus (MSSA), methicillin-

267

resistant Staphylococcus aureus (MRSA), and several Gram positive and Gram negative

268

bacteria.22,34-37 Atorvastatin, pravastatin, and rosuvastatin were also associated with

269

antimicrobial potential, but had a higher minimal inhibitory concentration (MIC) than

270

simvastatin.17,22,38 In a head to head comparison of MICs among the three statins of

271

interest in this study, Masadeh M. et al. found that simvastatin and atorvastatin have a

272

more potent antibacterial activity than rosuvastatin.22 Interestingly, our results correlate

273

well with the previous findings, and provide support for the notion that antibacterial

274

activity is playing a role in the observed mortality association.

275

The exact mechanism of the differential antibacterial activity between different types of

276

statins has not been elucidated, but several mechanisms have been proposed. Firstly,

277

statins’ ability to inactivate the 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA)

278

reductase (HMGR) is not only vital for cholesterol biosynthesis, but also contributes to

AC C

EP

TE D

M AN U

SC

RI PT

257

13

ACCEPTED MANUSCRIPT

the production of isoprenoids and lipid compounds that are essential for cell signaling

280

and structure in the pathogen.39 Secondly, the chemical property of different types of

281

statins may affect their targeting to bacteria. The lipophilic properties of simvastatin or

282

atorvastatin may allow better binding to bacteria cell wall than the hydrophilic

283

properties of rosuvastatin.40

284

Clinical evidence on the benefits of statins in sepsis patients have been inconsistent.

285

While early observational studies showed favorable effects of statins on sepsis patients,

286

recent randomized trial using de novo statin treatment in ICU patients with severe sepsis

287

failed to demonstrate clinical benefits.10,14 Healthy user bias, connected to the higher

288

health awareness and healthier lifestyle of patients under statin treatment, was thought

289

to account for the observed benefits in early observational studies.12 In fact, for most

290

clinical studies the lifestyle factors and preventive service seeking behavior were difficult

291

to measure and could not be adjusted for in the analysis. Even though our study could

292

not obtain lifestyle information as well, we took into account several factors that may

293

minimize the effect of healthy user bias. Firstly, in Taiwan statins are not available over-

294

the-counter and are covered by the government health insurance; the decision on the

295

initiation of statin treatment is based on a set of objective laboratory and clinical criteria,

296

regardless of patients’ income status or education level. Secondly, we used insurance

297

premium categories as a proxy to adjust for the socioeconomic status and frequency of

298

healthcare utilization as a proxy to adjust for the preventive service seeking behavior.

299

Lastly and most importantly, the unmeasured health behaviors associated with statin

300

users would not be different in patients receiving different types of statins, thus the

AC C

EP

TE D

M AN U

SC

RI PT

279

14

ACCEPTED MANUSCRIPT

presence of a significant difference in the outcome of sepsis in patients receiving

302

different types of statins would be an evidence against a strong influence of the healthy

303

user bias in this observational study.

304

We do not think the contradictory results between RCTs and observational studies could

305

be explained by healthy user bias alone, instead the difference in results might be

306

explained by three other factors: the specific type of statin, timing of statin treatment,

307

and severity of patient population. Firstly, as described in this manuscript, the benefit of

308

statin on sepsis patients is not equivalent across the entire class. Secondly, results of our

309

previous research and the majority of observational studies found that pre-admission

310

use of statin was associated with improved survival.9-11 Most of RCTs, however,

311

examined the de novo use of statin after hospital admission.12-16 There is biological and

312

pharmacokinetic plausibility that the anti-inflammatory properties of statins require

313

appropriate treatment duration before taking effect.41,42 Finally, the majority of RCTs

314

were performed on ICU patients who were severely ill. In our previous analysis, we

315

found that the beneficial effects of statins disappeared in patients requiring ICU

316

admission or patients with acute organ dysfunctions such as respiratory failure

317

(Supplementary eTable1).

318

Results of this study have to be interpreted in light of several limitations. Firstly, due to

319

the administrative nature of the dataset, we do not have laboratory data to verify the

320

distribution of infectious pathogens in patients receiving the three types of statins.

321

Simvastatin has been shown to be more effective against Staphylococcus aureus.

322

However, we found that the sources of infection were balanced in three types of statin

AC C

EP

TE D

M AN U

SC

RI PT

301

15

ACCEPTED MANUSCRIPT

users. In addition, we also did not have data on the lipid profile and could not analyze

324

the effect of lipid control on sepsis outcome. Secondly, as encountered in all

325

observational studies, we could not totally exclude the possibility of residual

326

confounding. By applying the newly developed 1:1:1 PS matching technique, we were

327

able to construct a comparison cohort with balanced covariates between the three

328

groups. Finally, the observational nature of the study is an intrinsic limitation because

329

the lack of randomization precludes a definite analysis of statin benefits. However, it

330

may be difficult to enroll a sufficient number of chronic statin users for randomization in

331

a short period. In this cohort of 52,737 patients with sepsis, only 6.8% of patients were

332

chronic user of statins. Thus, a randomized control trial on the effect of long-term statin

333

treatment on the outcome of sepsis would require tens of thousands of participants,

334

which is highly unlikely in the near future. Observational data may currently remain the

335

best available evidence for a class effect of statins for this indication.

336

In conclusion, results of this population-based study suggest that the benefit of statin on

337

sepsis patients is not equivalent across the entire class. Our results suggest superiority of

338

simvastatin and atorvastatin to rosuvastatin in reducing short-term mortality of sepsis.

339

The favorable effects of statins correlate with statin’s in vitro antibacterial activity, but

340

not with the lipid-lowering potency. The evidence also supports that the benefits of

341

statin in sepsis patients is not entirely due to healthy user bias. To conclude, our results

342

suggest the possibility for physicians to selectively prescribe simvastatin to patients who

343

are indicated for dyslipidemia treatment and are also at high risk of developing sepsis.

344

These high sepsis risk patients include those with cancer, diabetes, old age,

AC C

EP

TE D

M AN U

SC

RI PT

323

16

ACCEPTED MANUSCRIPT

immunocompromised conditions, and a history of sepsis. Further studies in different

346

populations or using different study designs are warranted to validate our findings.

AC C

EP

TE D

M AN U

SC

RI PT

345

17

ACCEPTED MANUSCRIPT

347 348

Figure legends Figure 1. Flowchart for cohort assembling process.

RI PT

349 Figure 2. Cumulative hazard plots for mortality before and after PS matching. (A) 30-day

351

mortality before matching, (B) 90-day mortality before matching, (C) 30-day mortality

352

after PS matching, (D) 90-day mortality after PS matching.

SC

350

AC C

EP

TE D

M AN U

353

18

ACCEPTED MANUSCRIPT

388

RI PT

Ethical approval This study was approved by the institutional review board of National Taiwan University Hospital.

SC

Conflict of Interest The authors declare that they had no competing interests when conducting the research.

M AN U

Acknowledgement We thank the staff of the Core Labs, the Department of Medical Research, and National Taiwan University Hospital for technical support, and medical wisdom consulting group for technical assistance in statistical analysis.

TE D

Contributors C-CLee designed the study, obtained funding, drafted the analytical plan, guided the statistical analysis, interpreted the data, wrote the final draft, and critically revised the manuscript. M-TL interpreted the results, wrote the first and final draft and point-topoint response. T-CH performed most of the statistical analysis. L P helped with language editing and provided critical comments C-HY, K-CT, LP and S-CC reviewed the manuscript and provided critical insights in the medical contents. M-L, analyzed the data, provided critical feedback, and authorized the final manuscript.

Transparency The lead author (the manuscript’s guarantor) affirms that the manuscript is an honest, accurate, and transparent account of the study; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

EP

382 383 384 385 386 387

Declarations Disclaimer The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health or National Health Research Institutes.

AC C

354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381

389

19

ACCEPTED MANUSCRIPT

References

391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433

1 Rhodes A, Evans LE, Alhazzani W, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock: 2016. Intensive Care Med 2017; 43:304-377 2 Rhodes A, Evans LE, Alhazzani W, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock: 2016. Crit Care Med 2017; 45:486-552 3 Angus DC, Linde-Zwirble WT, Lidicker J, et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001; 29:1303-1310 4 Fleischmann C, Scherag A, Adhikari NK, et al. Assessment of Global Incidence and Mortality of Hospital-treated Sepsis. Current Estimates and Limitations. Am J Respir Crit Care Med 2016; 193:259-272 5 Gao F, Linhartova L, Johnston AM, et al. Statins and sepsis. British Journal of Anaesthesia 2008; 100:288-298 6 Crisby M. Modulation of the inflammatory process by statins. Timely Top Med Cardiovasc Dis 2005; 9:E3 7 Crisby M. Modulation of the inflammatory process by statins. Drugs Today (Barc) 2003; 39:137-143 8 Hackam DG, Mamdani M, Li P, et al. Statins and sepsis in patients with cardiovascular disease: a population-based cohort analysis. The Lancet 2006; 367:413-418 9 Al Harbi SA, Tamim HM, Arabi YM. Association between statin therapy and outcomes in critically ill patients: a nested cohort study. BMC Pharmacology and Toxicology 2011; 11:12 10 Wan YD, Sun TW, Kan QC, et al. Effect of statin therapy on mortality from infection and sepsis: a meta-analysis of randomized and observational studies. Crit Care 2014; 18:R71 11 Mansur A, Steinau M, Popov AF, et al. Impact of statin therapy on mortality in patients with sepsis-associated acute respiratory distress syndrome (ARDS) depends on ARDS severity: a prospective observational cohort study. BMC Med 2015; 13:128 12 Brookhart MA, Patrick AR, Dormuth C, et al. Adherence to lipid-lowering therapy and the use of preventive health services: an investigation of the healthy user effect. American journal of epidemiology 2007; 166:348-354 13 Dinglas VD, Hopkins RO, Wozniak AW, et al. One-year outcomes of rosuvastatin versus placebo in sepsis-associated acute respiratory distress syndrome: prospective follow-up of SAILS randomised trial. Thorax 2016; 5:401-10 14 Deshpande A, Pasupuleti V, Rothberg MB. Statin therapy and mortality from sepsis: a meta-analysis of randomized trials. Am J Med 2015; 128:410-417 e411 15 McAuley DF, Laffey JG, O'Kane CM, et al. Simvastatin in the acute respiratory distress syndrome. New England Journal of Medicine 2014; 371:1695-1703 16 Heart TN. Rosuvastatin for sepsis-associated acute respiratory distress

AC C

EP

TE D

M AN U

SC

RI PT

390

20

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

syndrome. The New England journal of medicine 2014; 370:2191 17 Hennessy E, Adams C, Reen FJ, et al. Is there potential for repurposing statins as novel antimicrobials? Antimicrobial Agents and Chemotherapy 2016; 60:5111-5121 18 Ting M, Whitaker EJ, Albandar JM. Systematic review of the in vitro effects of statins on oral and perioral microorganisms. Eur J Oral Sci 2016; 124:4-10 19 Jerwood S, Cohen J. Unexpected antimicrobial effect of statins. J Antimicrob Chemother 2008; 61:362-364 20 Shepherd J, Hunninghake DB, Barter P, et al. Guidelines for lowering lipids to reduce coronary artery disease risk: a comparison of rosuvastatin with atorvastatin, pravastatin, and simvastatin for achieving lipid-lowering goals. Am J Cardiol 2003; 91:11C-17C; discussion 17C-19C 21 Kamat SA, Gandhi SK, Davidson M. Comparative effectiveness of rosuvastatin versus other statin therapies in patients at increased risk of failure to achieve low-density lipoprotein goals. Current medical research and opinion 2007; 23:1121-1130 22 Masadeh M, Mhaidat N, Alzoubi K, et al. Antibacterial activity of statins: a comparative study of atorvastatin, simvastatin, and rosuvastatin. Ann Clin Microbiol Antimicrob 2012; 11:13 23 Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the Twenty-sixth Annual SAS Users group international conference: SAS Institute, Cary, NC, 2001; 214-226 24 Rassen JA, Glynn RJ, Rothman KJ, et al. Applying propensity scores estimated in a full cohort to adjust for confounding in subgroup analyses. Pharmacoepidemiol Drug Saf 2012; 21:697-709 25 Rassen JA, Shelat AA, Franklin JM, et al. Matching by propensity score in cohort studies with three treatment groups. Epidemiology 2013; 24:401-409 26 Wilhelms SB, Huss FR, Granath G, et al. Assessment of incidence of severe sepsis in Sweden using different ways of abstracting International Classification of Diseases codes: difficulties with methods and interpretation of results. Critical care medicine 2010; 38:1442-1449 27 Gaieski DF, Edwards JM, Kallan MJ, et al. Benchmarking the incidence and mortality of severe sepsis in the United States. Critical care medicine 2013; 41:1167-1174 28 Serody JS, Berrey MM, Albritton K, et al. Utility of obtaining blood cultures in febrile neutropenic patients undergoing bone marrow transplantation. Bone Marrow Transplant 2000; 26:533-538 29 Glerant JC, Hellmuth D, Schmit JL, et al. Utility of blood cultures in communityacquired pneumonia requiring hospitalization: influence of antibiotic treatment before admission. Respir Med 1999; 93:208-212 30 Iwashyna TJ, Angus DC. Declining case fatality rates for severe sepsis: good data bring good news with ambiguous implications. JAMA 2014; 311:1295-1297 31 Merx MW, Liehn EA, Janssens U, et al. HMG-CoA reductase inhibitor simvastatin profoundly improves survival in a murine model of sepsis. Circulation 2004; 109:2560-2565 32 Zhang S, Luo L, Wang Y, et al. Simvastatin protects against T cell immune

AC C

434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479

21

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

dysfunction in abdominal sepsis. Shock 2012; 38:524-531 33 Zhang S, Rahman M, Zhang S, et al. Simvastatin antagonizes CD40L secretion, CXC chemokine formation, and pulmonary infiltration of neutrophils in abdominal sepsis. J Leukoc Biol 2011; 89:735-742 34 Wang CC, Yang PW, Yang SF, et al. Topical simvastatin promotes healing of Staphylococcus aureus-contaminated cutaneous wounds. Int Wound J 2016; 13:1150-1157 35 Graziano TS, Cuzzullin MC, Franco GC, et al. Statins and Antimicrobial Effects: Simvastatin as a Potential Drug against Staphylococcus aureus Biofilm. PloS one 2015; 10:e0128098 36 Thangamani S, Mohammad H, Abushahba MFN, et al. Exploring simvastatin, an antihyperlipidemic drug, as a potential topical antibacterial agent. Scientific reports 2015; 5:16407 37 Emani S, Gunjiganur GV, Mehta DS. Determination of the antibacterial activity of simvastatin against periodontal pathogens, Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans: An in vitro study. Contemp Clin Dent 2014; 5:377-382 38 Welsh A-M, Kruger P, Faoagali J. Antimicrobial action of atorvastatin and rosuvastatin. Pathology 2009; 41:689-691 39 Haeri MR, White K, Qharebeglou M, et al. Cholesterol suppresses antimicrobial effect of statins. Iranian journal of basic medical sciences 2015; 18:1253 40 Schachter M. Chemical, pharmacokinetic and pharmacodynamic properties of statins: an update. Fundamental & clinical pharmacology 2005; 19:117-125 41 Plenge JK, Hernandez TL, Weil KM, et al. Simvastatin lowers C-Reactive protein within 14 days an effect independent of low-density lipoprotein cholesterol reduction. Circulation 2002; 106:1447-1452 42 Chen WT, Krishnan GM, Sood N, et al. Effect of statins on atrial fibrillation after cardiac surgery: a duration- and dose-response meta-analysis. J Thorac Cardiovasc Surg 2010; 140:364-372

AC C

480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509

22

ACCEPTED MANUSCRIPT Table 1. Baseline characteristics of the three types of statin users. Simvastatin, N= 916

Atorvastatin, N=1855

Rosuvastatin, N=732

P-value

436 (47.6%) 69.47±12.78

894 (48.19%) 70.57±12.57

379 (51.78%) 69.79±12.67

0.18

Urban Area

378(41.27)

839(45.23)

329(44.88)

Metro Area

251(27.4)

485(26.15)

190(25.92)

Suburban Area

198(21.62)

357(19.25)

142(19.37)

89(9.72)

174(9.38)

72(9.82)

Demographics Gender male (%) Age (mean ± SD)

0.072

Insurance premium level Dependent

137(14.96)

$1-$19,999

71(7.75)

SC

Countryside Area

RI PT

Areas of residence

0.58

268(14.45)

82(11.19)

184(9.92)

76(10.37)

767(41.35)

235(32.06)

342(37.34)

636(34.29)

340(46.38)

424(46.29)

918(49.49)

363(49.52)

0.25

55(6)

100(5.39)

38(5.18)

0.73

15(1.64)

33(1.78)

9(1.23)

0.61

328(35.81)

656(35.36)

242(33.02)

0.44

43(4.69)

105(5.66)

43(5.87)

0.49

52(5.68)

107(5.77)

53(7.23)

0.32

3.42±2.7

4.11±2.9

4.1±2.8

<0.0001*

27(2.95)

71(3.83)

16(2.18)

0.087

315(34.39)

807(43.50)

332(45.29)

<0.0001*

102(11.14)

254(13.69)

92(12.55)

0.16

202(22.05)

561(30.24)

224(30.56)

<0.0001*

16(1.75)

38(2.05)

13(1.77)

0.82

113(12.34)

238(12.83)

96(13.10)

0.89

15(1.64)

16(0.86)

16(2.18)

0.021*

163(17.79)

365(19.68)

142(19.37)

0.486

Cardiac arrhythmias

247(26.97)

597(32.18)

222(30.29)

0.019*

Chronic pulmonary disease

467(50.98)

1074(57.9)

434(59.21)

0.0006*

26(2.84)

55(2.96)

17(2.32)

0.67

Complicated diabetes

442(48.25)

982(52.94)

362(49.39)

0.043*

Deficiency anemia

235 (25.66)

502(27.06)

226(30.83)

0.054

366(39.96)

M AN U

$20,000-$39,999 >=$40,000 Sources of infection Lower respiratory infection Intra-abdominal infection Biliary tract infection

Orthopedic infection Comorbidity Combined comorbidity score Metastatic cancer Congestive heart failure

Renal failure Weight loss Hemiplegia Alcohol abuse Any tumor

AC C

Dementia

EP

Skin infection

TE D

Genitourinary tract infection

Coagulopathy

<0.0001*

23

ACCEPTED MANUSCRIPT 195(21.29)

467(25.18)

203(27.69)

0.0087*

Liver disease

321(35.04)

680(36.66)

300 (40.93)

0.0405*

Peripheral vascular disorder

148(16.16)

356(19.19)

133(18.14)

0.15

Psychosis Pulmonary circulation disorders

85(9.28)

178(9.6)

61(8.32)

25(2.73)

67(3.61)

28(3.82)

0.60 0.39

HIV/AIDS

1(0.11)

2(0.11)

1(0.14)

0.98

797(87.01)

1675(90.3)

660(90.04)

0.025*

Number of OPD visit

45±28.95

42.68±26.42

43.59±26.86

0.016

Number of emergency department visit

0.69±1.77

1.05±3.022

1.17±2.22

0.0002*

Number of hospitalization

1.1±1.784

1.21±1.81

1.28±2.03

0.12

775(41.78)

323(44.07)

0.0002*

925(49.87)

359(48.98)

0.0025*

24(1.29)

21(2.86)

Healthcare utilization

Comedications use 458(50.00)

NSAIDs

394(43.01)

Aspirin Systemic immunosuppressive agents and biologics Systemic corticosteroids

0.022*

220(24.02)

402(21.67)

185(25.24)

0.108

17(1.86)

52(2.82)

21(2.86)

0.28

AC C

EP

TE D

DMARDs * means p-value <0.05

M AN U

19(2.07)

SC

Hypertension

RI PT

Fluid and electrolyte disorder

24

ACCEPTED MANUSCRIPT Table 2. Crude and adjusted effect measures for the association between use of statins and risk of sepsis in-hospital mortality. Crude effect estimate Confounder adjusted Propensity score (HR, 95% confidence effect estimate matched effect estimate interval) (HR, 95% confidence (HR, 95% confidence interval) interval)

0.84(0.77 – 0.91)***

0.86(0.78 – 0.94)***

0.88(0.78 – 0.99)*

Simvastatin

0.82(0.70-0.98)*

0.88(0.74-1.04)

0.72(0.58-0.90)**

Atorvastatin

0.86(0.77-0.97)*

0.86(0.76-0.96)*

0.78(0.68-0.90)***

Rosuvastatin

1.12(0.95-1.32)

1.12(0.95-1.32)

0.87(0.73-1.04)

Statin use

0.91(0.87-0.95)***

0.94(0.90-0.98)**

0.93(0.88-0.99)**

Simvastatin

0.93(0.86-1.01)**

0.99(0.91-1.08)

0.87(0.78-0.98)*

Atorvastatin

0.91(0.86-0.97)

0.93(0.88-0.99)*

Rosuvastatin

0.94(0.85-1.03)

0.98(0.90-1.08)

90-day mortality

0.90(0.83-0.97)**

0.90(0.82-0.99)**

AC C

EP

TE D

M AN U

*** means p-value <0.001, ** means p-value <0.01, * means p-value <0.05

RI PT

Statin use

SC

30-day mortality

25

ACCEPTED MANUSCRIPT Table 3. Head to head comparison of the three types of statin users. Simvastatin vs. Simvastatin vs. Atorvastatin Rosuvastatin (HR, 95% confidence interval) (HR, 95% confidence interval) 30 days mortality

Atorvastatin vs. Rosuvastatin (HR, 95% confidence interval)

Confounder adjusted analysis

1.03(0.83-1.27)

0.78(0.61-0.99)*

0.77(0.62-0.94)*

Propensity score matched analysis

1.05(0.82-1.36)

0.80(0.60-1.06)

0.82(0.64-1.06)

Confounder adjusted analysis

1.01(0.84-1.22)

0.78(0.63-0.97)*

Propensity score matched analysis

1.03(0.82-1.29)

0.77(0.59-0.99)*

Confounder adjusted analysis

1.07(0.97-1.19)

1.04 (0.92-1.19)

Propensity score matched analysis

1.04(0.91-1.18)

1.00 (0.85-1.18)

SC

Acute respiratory failure

RI PT

90 days mortality

0.79(0.64-0.99)*

0.96 (0.86-1.07)

0.93 (0.80-1.08)

AC C

EP

TE D

M AN U

* means p-value <0.05

0.76(0.63-0.90)*

26

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

 

SC

M AN U

Codes associated with infection Gastrointestinal infection 001, Cholera; 002, Typhoid/paratyphoid fever; 003, Other salmonella infection; 004, Shigellosis; 005 Other food poisoning; 008, Intestinal infection not otherwise classified; 009, Ill-defined intestinal infection; 562.01, Diverticulitis of small intestine without hemorrhage; 562.03, Diverticulitis of small intestine with hemorrhage; 562.11, Diverticulitis of colon without hemorrhage; 562.13, Diverticulitis of colon with hemorrhage;

RI PT

e-Appendix 1: Codes associated with infection and organ dysfunction

AC C

EP

Zoonosis 020, Plague; 021, Tularemia; 022, Anthrax; 023, Brucellosis; 024, Glanders; 025, Melioidosis; 026, Rat-bite fever; 027, other bacterial zoonoses;

TE D

Tuberculosis 010, Primary tuberculosis infection; 011 Pulmonary tuberculosis; 012, Other respiratory tuberculosis; 013, Central nervous system tuberculosis; 014, Intestinal tuberculosis; 015, Tuberculosis of bone and joint; 016, Genitourinary tuberculosis; 017, Tuberculosis not otherwise classified; 018, Miliary tuberculosis;

Other mycobacterial disease 030, Leprosy; 031, Other mycobacterial disease; Syphilis 090, Congenital syphilis; 091, Early symptomatic syphilis; 092, Early syphilis latent; 093, Cardiovascular syphilis; 094, Neurosyphilis; 095, Other late symptomatic syphilis; 096, Late syphilis latent; Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

  Other and unspecified syphilis; Yaws; Pinta; Other spirochetal infection;

037, Tetanus;

infection Bacterial meningitis; Meningitis, unspecified; Central nervous system abscess; Phlebitis of intracranial sinus; Meningococcal infection;

EP

CNS 320, 322, 324, 325, 036,

TE D

M AN U

Systematic fungal infection 110, Dermatophytosis; 111, Dermatomycosis not otherwise classified or specified; 112, Candidiasis; 114, Coccidioidomycosis; 115, Histoplasmosis; 116, Blastomycotic infection; 117, Other mycoses; 118, Opportunistic mycoses; 117.9 Disseminated fungal infection 112.5 Dissemintaed candidal infection 112.81 Disseminated fungal endocarditis 039, Actinomycotic infections;

SC

Other bacterial diseases 040, Other bacterial diseases; 041, Bacterial infection in other diseases not otherwise specified;

RI PT

097, 102, 103, 104,

AC C

Cardiovascular infection 420, Acute pericarditis; 421, Acute or subacute endocarditis; 451, Thrombophlebitis;

Upper respiratory tract infection 461, Acute sinusitis; 462, Acute pharyngitis; 463, Acute tonsillitis; 464, Acute laryngitis/ tracheitis; 465, Acute upper respiratory infection of multiple sites/not otherwise specified; 101, Vincent’s angina; 034, Streptococcal throat/scarlet fever; 032, Diphtheria; Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

Biliary tract infection 572.1, Portal pyremia; 575.0, Acute cholecystitis;

M AN U

TE D

Intra-abdominal infection 540, Acute appendicitis; 541, Appendicitis not otherwise specified; 542, Other appendicitis; 566, Anal and rectal abscess; 567, Peritonitis; 569.5, Intestinal abscess; 569.83, Perforation of intestine; 572.0, Abscess of liver;

SC

Lower respiratory tract infection 481, Pneumococcal pneumonia; 482, Other bacterial pneumonia; 485, Bronchopneumonia with organism not otherwise specified; 486, Pneumonia, organism not otherwise specified; 491.21, Acute exacerbation of obstructive chronic bronchitis; 494 Bronchiectasis; 510, Empyema; 513, Lung/mediastinum abscess; 033, Whooping cough; 484 Pneumonia classified in elsewhere 483 Pneumonia by other pathogens

RI PT

 

AC C

EP

Genitourinary tract infection 590, Kidney infection; 597, Urethritis/urethral syndrome; 599.0, Urinary tract infection not otherwise specified; 601, Prostatic inflammation; 098, Gonococcal infections; Gynecological infection 614, Female pelvic inflammation disease; 615, Uterine inflammatory disease; 616, Other female genital inflammation; 681, Cellulitis, finger/ toe; 098, Gonococcal infections; Skin and appendix structure infection 682, Other cellulitis or abscess; 683, Acute lymphadenitis; 686, Other local skin infection; 035, Erysipelas; Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

 

Post-operative complication 998.5, Postoperative infection;

Septicemia o o o o

SC

Nosocomial infection 999.3, Infectious complication of medical care not otherwise classified.

RI PT

Musculoskeletal infection 711.0, Pyogenic arthritis; 730, Osteomyelitis; 790.7, Bacteremia; 996.6, Infection or inflammation of device/graft;

Streptococcal septicemia Staphylococcal septicemia Pneumococcal septicemia [Streptococcus pneumoniae septicemia] Septicemia due to anaerobes Septicemia due to bacteroides Excludes: gas gangrene (040.0) § that due to anaerobic streptococci (038.0) o 038.4 Septicemia due to other gram-negative organisms § 038.40 Gram-negative organism, unspecified § Gram-negative septicemia NOS § 038.41 Hemophilus influenzae [H. influenzae] § 038.42 Escherichia coli [E. coli] § 038.43 Pseudomonas § 038.44 Serratia § 038.49 Other o 038.8 Other specified septicemias § Excludes: septicemia (due to): § anthrax (022.3) § gonococcal (098.89) § herpetic (054.5) § meningococcal (036.2) § septicemic plague (020.2) o 038.9 Unspecified septicemia § Septicemia NOS § Excludes: bacteremia NOS (790.7) 995.92 Severe sepsis 790.7 Bacteremia

o o

AC C

EP

TE D

M AN U

038.0 038.1 038.2 038.3 § §

Codes associated with organ dysfunction Cardiovascular dysfunction/Shock Shock 785.5 or hypotension 458

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

 

Hematologic system dysfunction Hematologic Secondary thrombocytopenia 287.4 Thrombocytopenia, unspecified 287.5 Other/unspecified coagulation defect 286.9 Defibrination syndrome 286.6

M AN U

Hepatic system dysfunction Hepatic Acute and subacute necrosis of liver 570 Hepatic encephalopathy 572.2 Hepatorenal syndrome 572,4 Other squeal of chronic liver disease 572.8 Hepatic infarction 573.4, 573.8 Liver replaced by transplant V42.7

SC

Central nervous system dysfunction Neurologic Encephalopathy 348.3 Transient organic psychosis 293 Anoxic brain damage 348.1

RI PT

Acute respiratory failure Mechanical ventilation 96.7 or use of ventilator 57001B,57002B, 57023B, 57029C

TE D

Renal system dysfunction Acute renal failure 584 or Use of CVVH (58014C)

AC C

EP

Metabolic system dysfunction Diabetic ketoscidosis 250.1 Hyperosmolar hyperglycemic state 250.2

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

  e-Table 1. Empirical predictors for prescription of statin and associated odds ratios of the propensity score model. The c-statistic for the propensity score model is 0.82. Adjusted odds ratio (95%CI)

Chronic pulmonary disease

EP

<.0001 <.0001

1.49 (1.31 – 1.71) 1.29 (1.12 – 1.48) 1.09 (0.94 – 1.25)

<.0001

0.99 (0.89 – 1.12) 1.37 (1.21 – 1.55) 1.50 (1.27 – 1.78)

<.0001

0.91 1.46 1.17 1.26 1.02 1.02

<.0001 <.0001 0.01 .0008 0.66 0.74

SC

Rheumatologic disease Peptic ulcer disease Mild liver disease Diabetes without chronic complications Diabetes with chronic complications Hemiplegia or paraplegia Renal disease Any malignancy, including leukemia and lymphoma Moderate or severe liver disease Metastatic solid tumor AIDS/HIV cardiovascular comorbidities Stroke or transient ischemic attack Peripheral arterial disease Angina Other ischemic heart disease Cerebral atherosclerosis Percutaneous coronary/coronary artery bypass graft intervention Additional comorbidities Alcohol/drug use Psychiatric disorder Neurologic disorder

AC C

0.68 (0.62 – 0.73) 1.11 (1.08 – 1.13)

M AN U

TE D

Demographics Gender male Age Area (countryside area as reference) Urban Area Metro Area Suburban Area Insurance type (dependent as reference) $1-$19,999 $20,000-$39,999 >=$40,000 Comorbidity Combined comorbidity score Myocardial infarction Congestive heart failure Peripheral vascular disease Cerebrovascular disease Dementia

P-value

RI PT

Characteristics

(0.89 (1.29 (1.05 (1.10 (0.92 (0.89

– – – – – –

0.94) 1.64) 1.31) 1.44) 1.13) 1.17)

1.03 (0.92 – 1.17)

0.59

0.98 1.02 1.08 2.18 1.59 1.12 1.38 0.85 0.35 1.50 0.68

(0.83 (0.94 (1.00 (1.94 (1.45 (0.98 (1.25 (0.77 (0.26 (1.15 (0.23

– – – – – – – – – – –

1.15) 1.11) 1.18) 2.39) 1.75) 1.27) 1.52) 0.98) 0.47) 1.96) 2.65)

0.79 0.70 0.05 <.0001 <.0001 0.09 <.0001 0.01 <.0001 0.003 0.68

1.22 0.91 1.22 1.31 1.02 1.50

(1.10 (0.77 (1.11 (1.20 (0.88 (1.32

– – – – – –

1.36) 1.07) 1.34) 1.44) 1.18) 1.72)

0.0002 0.27 <.0001 <.0001 0.82 <.0001

0.91 (0.77 – 1.07) 1.11 (1.02 – 1.20) 0.75 (0.67 – 0.84)

0.26 0.01 <.0001

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

  (1.09 (1.09 (0.85 (0.67 (0.77 (0.44

1.82) 1.31) 1.05) 1.04) 3.75) 0.89)

1.00 (1.00 – 1.01) 1.03 (1.02 – 1.04) 0.89 (0.86 – 0.91)

M AN U

0.98 1.48 1.18 1.00 1.16 1.25 1.21 1.18 2.10 0.90 1.14 1.00 1.03 1.37 0.99

(0.90 (1.37 (0.78 (0.91 (0.89 (1.14 (1.11 (1.08 (1.93 (0.78 (1.03 (0.76 (0.91 (1.26 (0.91

– – – – – – – – – – – – – – –

1.06) 1.61) 1.80) 1.10) 1.52) 1.36) 1.31) 1.29) 2.28) 1.04) 1.26) 1.33) 1.17) 1.48) 1.07)

0.008 .0002 0.27 0.11 0.18 0.009 <.0001 <.0001 <.0001 0.64 <.0001 0.43 0.98 0.28 <.0001 <.0001 0.0002 <.0001 0.15 0.01 0.98 0.60 <.0001 0.73

AC C

EP

TE D

– – – – – –

RI PT

1.41 1.19 0.94 0.84 1.70 0.63

SC

Obesity Other Cancer except Metastatic solid tumor Gastrointestinal or esophageal hemorrhage Bed-ridden status Solid organ transplantation such as renal or heart transplantation Malnutrition Healthcare utilization Number of OPD visit Number of emergency department visit Number of hospitalization Medication NSAID (Nonsteroidal anti-inflammatory drugs) Aspirin Systemic immunosuppressive agents and biologics Systemic corticosteroids DMARDs (disease modifying anti-rheumatic drugs) ACE inhibitors Beta blockers Loop diuretics Angiotensin II antagonists Digoxin Nitrates Antipsychotics Proton-pump inhibitors(PPI) Calcium channel blocker Acetaminophen

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

 

Propensity score adjusted HR (95% Confidence interval)

P-value

>75 years of age

0.92 (0.77 - 1.11)

0.38

<75 years of age

0.71 (0.59 - 0.87)

0.0006

Male

0.82 (0.68 - 0.98)

0.032

Female

0.80 (0.66 - 0.98)

0.027

Diabetes

0.76 (0.65 - 0.89)

0.0007

Non diabetes

0.83 (0.70 - 0.99)

0.039

With congestive heart failure

0.83 (0.68 - 1.00)

0.055

M AN U

e-Table 2. Risk of mortality in different patient subgroups

0.80 (0.67 - 0.96)

0.017

0.88 (0.64 - 1.21)

0.430

0.80 (0.69 - 0.92)

0.003

0.71 (0.53 - 0.96)

0.024

0.84 (0.73 - 0.98)

0.025

0.94 (0.85,1.04)

0.204

0.79 (0.64,0.98)

0.033

1.17 (0.72,1.91)

0.536

No Hematologic system dysfunction

0.87 (0.80,0.96)

0.004

Use of vasopressor

0.94 (0.85,1.03)

0.177

0.95 (0.73,1.23)

0.689

0.91 (0.64,1.30)

0.606

0.87(0.79,0.95)

0.003

1 or 2 organ dysfunction

0.87(0.78,0.96)

0.004

>= 3 organ dysfunction

1.00(0.80,1.25)

0.999

0.90 (0.78 - 1.03)

0.194

0.88 (0.81 – 0.97)

<0.0001

SC

RI PT

Patient subgroups

Without congestive heart failure Myocardial infarction Non myocardial infarction Cancer Non cancer

Non Ventilator Support

Dialysis No dialysis

ICU Non-ICU

AC C

No use of vasopressor

EP

Hematologic system dysfunction

TE D

Ventilator Support

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.

ACCEPTED MANUSCRIPT

 

AC C

EP

TE D

M AN U

SC

RI PT

e-Figure1. Standardized differences in the baseline covariates after propensity score matching. Simvastatin vs. atorvastatin (A), and simvastatin vs. rosuvastatin (B).

170913

Online supplements are not copyedited prior to posting and the author(s) take full responsibility for the accuracy of all data.