Postoperative Polymyxin B Hemoperfusion and Mortality in Patients With Abdominal Septic Shock: A Propensity-Matched Analysis* Masao Iwagami, MD, MPH1; Hideo Yasunaga, MD, PhD2; Kent Doi, MD, PhD3; Hiromasa Horiguchi, PhD4; Kiyohide Fushimi, MD, PhD5; Takehiro Matsubara, MD, PhD3; Naoki Yahagi, MD, PhD3; Eisei Noiri, MD, PhD1
Objectives: To examine the effect of postoperative polymyxin B hemoperfusion on mortality in patients with abdominal septic shock triggered by lower gastrointestinal tract perforation, identifying subpopulations of patients who may benefit from this treatment. Design: Propensity-matched analysis. Setting: We used a nationwide inpatient database in Japan. Patients: We included patients who are 18 years old or older hospitalized during a period of 34 months between July 2007 and October 2011, who had open abdominal surgery on the day of admission (day 0) for perforation of lower gastrointestinal tract, *See also p. 1309. 1 Department of Hemodialysis and Apheresis, The University of Tokyo Hospital, Tokyo, Japan. 2 Department of Health Economics and Epidemiology Research, School of Public Health, The University of Tokyo, Tokyo, Japan. 3 Department of Emergency and Critical Care Medicine, The University of Tokyo Hospital, Tokyo, Japan. 4 Department of Clinical Data Management and Research, Clinical Research Center, National Hospital Organization Headquarters, Tokyo, Japan. 5 Department of Health Informatics and Policy, Tokyo Medical and Dental University Graduate School of Medicine, Tokyo, Japan. Drs. Iwagami and Yasunaga had full access to all the study data and take responsibility for the integrity of the data and accuracy of the data analysis. All the authors directly participated in the planning, execution, or analysis of the study and read and approved the final submitted version of this article. Supported, in part, by grants H22-Policy-031 (Research on Policy Planning and Evaluation) from the Ministry of Health, Labour and Welfare, Japan; 22390131 (Scientific Research B) from Ministry of Education, Culture, Sports, Science and Technology, Japan; and 0301002001001 (Funding Program for World-Leading Innovative R&D on Science and Technology, FIRST program) from the Council for Science and Technology Policy, Japan. The funders had no role in the execution of this study or interpretation of results. Drs. Yasunaga and Fushimi’s institutions received grant support from the Japanese government. The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2013 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0000000000000150
Critical Care Medicine
and who required noradrenaline and/or dopamine. We excluded patients who died on day 0 or 1 and patients starting polymyxin B hemoperfusion on day 2 or later. Measurements and Main Results: The main outcome was 28-day mortality. Of 2,925 eligible patients, 642 received one or two polymyxin B hemoperfusion sessions, starting the first one on day 0 or 1. Propensity score matching created a matched cohort of 1,180 patients (590 pairs with and without polymyxin B hemoperfusion). The 28-day mortality was 17.1% (101 of 590) in the polymyxin B hemoperfusion group and 16.3% (96 of 590) in the control group (p = 0.696). Subgroup analyses by number of polymyxin B hemoperfusion sessions (one or two), timing of polymyxin B hemoperfusion initiation (day 0 or 1), the use of noradrenaline, and number of dysfunctional organs (one to six) did not show any significant difference in 28-day mortality between the groups. Multiple logistic did not show a significant association between the use of polymyxin B hemoperfusion and 28-day mortality (adjusted odds ratio, 1.10; 95% CI, 0.80–1.51; p = 0.569). Age, end-stage renal disease requiring maintenance hemodialysis, the use of noradrenaline, and number of dysfunctional organs were positively associated with 28-day mortality. Conclusions: In this retrospective study, postoperative polymyxin B hemoperfusion did not show any survival benefit for the overall study population or any of the studied subgroups of patients with abdominal septic shock. A large multicentered prospective randomized trial is warranted to identify the true role of polymyxin B hemoperfusion in sepsis caused by Gram-negative bacteria. (Crit Care Med 2014; 42:1187–1193) Key Words: abdominal septic shock; endotoxin; polymyxin b hemoperfusion; propensity-matched analysis
S
epsis, also known as “systemic inflammatory response syndrome” associated with infection, is a serious medical condition with high mortality (1). Endotoxin, a lipopolysaccharide derived from the outer membrane of Gram-negative bacteria, is recognized as a key factor in the sepsis cascade www.ccmjournal.org
1187
Iwagami et al
because it triggers a series of inflammatory reactions that lead to organ dysfunction (2, 3); its high serum level is linked to increased risk of multiple organ failure and death (4, 5). Thus, endotoxin has been regarded as a therapeutic target in treating sepsis (6). Endotoxin adsorption by direct hemoperfusion with polymyxin B-immobilized fiber column has been developed as an endotoxin-targeting therapy because polymyxin B has a high affinity for endotoxin through hydrophobic and ionic forces (7). A meta-analysis suggested that it can improve mortality in addition to mean arterial pressure, dopamine use, and mean Pao2/Fio2 ratio; however, the studies included in that analysis were of suboptimal quality (8). Abdominal infection is a major cause of sepsis (1), and it can be severe, particularly when caused by lower gastrointestinal tract perforation; this is because dissemination of feces containing intestinal Gram-negative bacteria into the abdominal cavity easily leads to bacteremia, with high levels of endotoxin. Many studies have investigated the efficacy of postoperative polymyxin B hemoperfusion (PMX) in patients with abdominal sepsis to identify improvements, mainly in cardiovascular function (9–13). However, controversy remains as to whether postoperative PMX can reduce mortality in such patients, partly because of the paucity of studies with adequate sample size. In Japan, PMX for treating septic shock has been covered by the Japanese health insurance system since 1994 and is therefore widely used in clinical practice. A postmarketing evaluation with a large sample size may help clarify the effectiveness of PMX, especially in reducing mortality. The purpose of this study was to investigate the survival benefit of postsurgical PMX in a homogenous target population of patients with lower gastrointestinal tract perforation requiring open abdominal surgery by means of the Japanese Diagnosis Procedure Combination (DPC) inpatient database.
MATERIALS AND METHODS Data Source The details of the DPC database are described elsewhere (14). Briefly, the DPC is a case-mix patient classification system, whereby administrative claims data and detailed medical data are collected for all inpatients discharged from participating hospitals. Until 2010, data were collected over a 6-month period every year between July 1 and December 31. Since 2011, this data collection has been conducted all year round. The DPC database includes the following information: patients’ age and sex; main diagnosis, comorbidities at admission and postadmission complications classified according to the International Classification of Diseases, 10th Revision (ICD-10) and text data in Japanese; medical procedures, including types of surgery, coded with original Japanese codes; daily records of drug administration and devices used; length of stay; and discharge status. All clinical data for each patient are recorded at discharge by attending physicians. To optimize the accuracy of recorded diagnoses, responsible physicians are obliged to record diagnoses with reference to medical charts. The dates 1188
www.ccmjournal.org
of hospital admission, surgery, discharge, bedside procedures performed, and drugs administered are recorded using a uniform data-submission format. In this study, hospital volume was defined as the average annual number of surgical procedures performed on eligible patients; a unique identifier was used for each hospital. All patient identifiers were removed from this database. Because of the anonymous nature of the data, the need for informed consent was waived. Study approval was obtained from the Institutional Review Board of the University of Tokyo. PMX Since 1994, PMX has been covered by the Japanese health insurance system for patients with sepsis who require vasoactive support with endotoxemia or in whom Gram-negative infection is suspected. A maximum of two hemoperfusion sessions are covered, provided they are not performed on the same day. The only commercially available polymyxinB-immobilized column for adults is Toraymyxin-20R (Toray Industries, Tokyo, Japan). Patient Selection Among 17 million inpatients in the DPC database over a total of 34 months between July 1, 2007, and October 31, 2011 (for 2007–2010, 6 mo annually from July 1 to December 31; for 2011, 10 mo from January 1 to October 31), patients who are 18 years old or older satisfying the following inclusion criteria were selected: 1) diagnosed with nontraumatic perforation of the lower gastrointestinal tract (including the small bowel, colon, and rectum, but not the appendix); 2) underwent open abdominal surgery except exploratory laparotomy on the day of admission (day 0); and 3) started noradrenaline and/or dopamine on day 0. The following patients were excluded: 1) those who died on day 0 or 1 and 2) those who underwent PMX starting on day 2 or later. Among the eligible patients, those whose first PMX session was on day 0 or 1 were identified. For the PMX group, the number of hemoperfusion sessions performed (one or two on consecutive days) was calculated. Variables and Endpoint Regarding underlining conditions, end-stage renal disease that required maintenance hemodialysis was assessed based on the records of the devices used. Liver cirrhosis and malignancy were determined according to ICD-10. The findings associated with surgery were assessed as follows. Perforation sites were categorized into small bowel, colon, rectum, or unknown when described only as “lower gastrointestinal tract perforation,” using ICD-10 in conjunction with the Japanese text data. Red cell concentrate transfusion and fresh-frozen plasma transfusion on day 0 were also taken into account. The presence of organ dysfunction for six organs at admission was defined as follows: 1) cardiovascular dysfunction that required initiation of noradrenaline and/or dopamine on day 0; 2) respiratory dysfunction that required postsurgical continuous mechanical ventilation; 3) renal dysfunction May 2014 • Volume 42 • Number 5
Clinical Investigations
that necessitated initiation of intermittent acute hemodialysis or continuous renal replacement therapy on day 0—except for maintenance hemodialysis patients; 4) hepatic dysfunction as a comorbidity at admission for recorded “liver failure”—except for liver cirrhosis patients; 5) hematologic dysfunction that required platelet concentrate transfusion on day 0; and 6) neurological dysfunction of 100 on the Japan Coma Scale score, which is equivalent to scores of 6–9 on the Glasgow Coma Scale, or more severe. The Japan Coma Scale was recorded in all patients at admission to assess the consciousness level, and it correlated well with the Glasgow Coma Scale (15). The only endpoint used in this study was 28-day mortality. Statistical Analysis Propensity score matching was used to adjust for differences in baseline characteristics and severity of condition at admission between patients with and without PMX (16). This was because previous retrospective observational studies have demonstrated that propensity score methods are a powerful tool for comparing groups with similar observed characteristics without specifying the relationship between confounders and outcomes (17, 18). First, the propensity score was estimated. The log odds of the probability that a patient received PMX treatment was modeled for potential confounders: age; sex; underlining conditions; findings related to surgery; use of noradrenaline, dopamine, and dobutamine; the six organ dysfunctions defined above; hospital volume; and type of hospital (university or nonuniversity hospital). C-statistic was calculated to evaluate the goodness of fit. A one-to-one matched analysis using nearest-neighbor matching was performed based on the estimated propensity scores of each patient. A match occurred when one patient in the PMX group had an estimated score within 0.25 sds of another in the control group (19). Descriptive statistics were presented for all patients and propensity-matched patients. Continuous variables were compared using t test, and categorical variables were compared using chi-square test. For a comparison of 28-day mortality, Fisher exact test or chi-square test was used among the propensity-score-matched groups. Subgroup analyses were also performed among the propensity-score-matched groups according to the number of sessions, day of the first hemoperfusion session, the use of noradrenaline, and number of organ dysfunctions. Patients discharged alive before day 28 were regarded as survivors: lacking hospice facilities or longterm acute-care facilities, the Japanese healthcare system allows patients to stay in the same hospital until recovery is complete. Univariate and multivariate logistic analyses were also performed for the propensity-matched patients to identify independent predictors of 28-day mortality. In a nonstepwise manner, we selected variables showing statistically significant differences between survivors and nonsurvivors with chisquare test. A p value of less than 0.05 was considered statistically significant. All statistical analyses were conducted using IBM SPSS version 20 (IBM SPSS, Armonk, NY). Critical Care Medicine
RESULTS Among 3,139 patients who satisfied the inclusion criteria, we excluded the following: 1) 193 patients who died on day 0 or 1 and 2) 21 patients who underwent the first PMX session on day 2 or later. The remaining 2,925 eligible patients were categorized into the PMX group (n = 642) or control group (n = 2,283), from which 590 propensity score-matched pairs were generated (Fig. 1). C-statistic indicated that the goodness of fit was 0.735 in the propensity score model. Table 1 shows the baseline characteristics of the unmatched PMX and control groups (n = 2,925) and of the propensity score-matched groups (n = 1,180). When the unmatched groups were compared, patients on maintenance hemodialysis, those who underwent colorectal perforation, those who underwent red cell concentrate and fresh-frozen plasma transfusion, those receiving noradrenaline and dobutamine, and those with respiratory, renal, and/or hematologic dysfunction were more likely to receive PMX treatment. After propensity score matching, the baseline patient characteristics were well balanced between the groups. Overall 28-day mortality was 13.4% (393 of 2,925). Table 2 shows the 28-day mortality for the propensity-matched groups: it was 17.1% (101 of 590) for the PMX group and 16.3% (96 of 590) for the control group (p = 0.696). Subgroup analyses according to the number of PMX sessions and day of the first hemoperfusion session revealed no statistically significant difference. Table 3 presents a comparison of 28-day mortality among the propensity-matched groups according to the use of noradrenaline and number of organ dysfunctions; there was no significant difference. Table 4 shows the results of the univariate and multivariate logistic regression analyses for 28-day mortality. No significant association was found between PMX treatment and 28-day mortality (adjusted odds ratio, 1.10; 95% CI, 0.80–1.51; p = 0.569), whereas age, end-stage renal disease requiring maintenance hemodialysis, the use of noradrenaline, and number of organ dysfunctions were positively associated with 28-day mortality.
Figure 1. Outline of patient selection. PMX = polymyxin B hemoperfusion. www.ccmjournal.org
1189
Iwagami et al
Table 1.
Baseline Patient Characteristics in Unmatched and Propensity-Matched Groups Unmatched Groups
Variable
Age (yr, mean ± sd) Sex (male), %
Polymyxin B Hemoperfusion (n = 642)
Control (n = 2,283)
74.5 ± 10.9
74.7 ± 11.9
Matched Groups Polymyxin B Hemoperfusion (n = 590)
Control (n = 590)
p
0.719
74.5 ± 10.8
74.9 ± 11.2
0.565
p
45.6
47.4
0.431
45.8
44.7
0.726
Maintenance hemodialysis
8.1
5.5
0.016
8.5
8.5
1.000
Liver cirrhosis
0.8
0.5
0.367
0.8
0.5
0.726
24.8
26.5
0.389
24.7
25.3
0.840
6.9
17.6
< 0.001
7.5
7.5
0.999
73.2
66.8
72.7
72.4
18.2
14.1
18.0
18.3
1.7
1.5
1.9
1.9
Red cell concentrate transfusion, %
33.5
28.7
0.019
32.5
33.2
0.804
Fresh-frozen plasma transfusion, %
34.6
23.6
< 0.001
32.5
31.9
0.803
Noradrenaline
53.1
36.6
< 0.001
50.3
53.2
0.322
Dopamine
82.4
85.9
0.030
83.2
83.9
0.753
Dobutamine
10.1
7.1
0.010
10.3
11.4
0.574
Underlining conditions, %
Malignancy Findings related to surgery Perforation site, % Small bowel Colon Rectum Unknown
a
Type of vasoactive drug, %
Organ dysfunctions at admission, %b Cardiovascular
100
100
1.000
100
100
1.000
Respiratory
86.1
57.9
< 0.001
84.9
86.6
0.405
Renal
20.4
5.1
< 0.001
13.7
11.9
0.338
Hematologic
4.2
2.5
0.022
4.1
3.9
0.882
Hepatic
0.8
0.4
0.175
0.7
0.5
1.000
Neurological
6.5
4.7
0.066
6.3
4.9
0.311
Described only as “lower gastrointestinal tract perforation,” using the International Classification of Diseases, 10th Revision, in conjunction with the Japanese text data. b Organ dysfunctions at admission were defined as follows: 1) cardiovascular dysfunction that required the use of at least one vasopressor on day 0; 2) respiratory dysfunction that necessitated postsurgical continuous mechanical ventilation; 3) renal dysfunction that required the initiation of intermittent acute hemodialysis or continuous renal replacement therapy on day 0—except for maintenance hemodialysis patients; 4) hepatic dysfunction recorded as “liver failure” at admission—except for liver cirrhosis patients; 5) hematologic dysfunction that necessitated platelet concentrate transfusion on day 0; and 6) neurological dysfunction at admission of 100 on the Japan Coma Scale (scores of 6–9 on the Glasgow Coma Scale) or more severe. a
DISCUSSION In this study, a propensity score analysis was performed using a nationwide database to examine the effect of PMX on mortality in patients with abdominal septic shock. Although the analysis of baseline patient characteristics in the unmatched groups indicated more active utilization of PMX among patients with more severe conditions, the one-to-one propensity score matching successfully balanced patient characteristics between the treated and untreated groups. In particular, adjusting for the proportion of organ dysfunctions was essential because 1190
www.ccmjournal.org
the number of dysfunctional organs and mortality have been shown to be strongly correlated (20, 21). The results showed that PMX treatment did not present any survival benefit for the overall study population—regardless of the number of sessions and the day of initiation. Likewise, none of the subgroup analyses performed in this study indicated a significant association of mortality with PMX treatment. Many studies have examined the effect of postoperative PMX on abdominal sepsis (9–13). A pilot randomized study in 36 postsurgical patients with abdominal sepsis indicated that May 2014 • Volume 42 • Number 5
Clinical Investigations
Table 2.
The 28-Day Mortality in the Propensity-Matched Groups With Subgroup Analyses No. of Patients
28-Day Mortality (%)
p
Polymyxin B hemoperfusion
590
101 (17.1)
0.696
Control
590
96 (16.3)
2 sessions
277
48 (17.3)
1 session
313
53 (16.9)
Control
590
96 (16.3)
Variables
Number of polymyxin B hemoperfusion sessions 0.919
The day of the first session of polymyxin B hemoperfusion Day 0
428
67 (15.7)
Day 1
162
34 (21.0)
Control
590
96 (16.3)
one session of PMX within 24 hours after emergency surgery (within 48 hr after elective surgery) improved cardiac function and reduced the need for continuous renal replacement therapy—but not 28-day mortality (9). A subsequent randomized controlled trial in 64 postsurgical patients with abdominal septic shock demonstrated that two sessions of PMX with a 24-hour interval (the first starting within 24 hr after emergency surgery) improved hemodynamics and Sequential Organ Failure Assessment score (11). However, there was a lack of statistical significance in the odds ratio of the crude 28-day mortality (22). This study had similar findings to the above studies in that the total number of PMX sessions was one or two, and the first session was performed on postoperative day 0 (up to 24 hr after surgery) or day 1 (up to 48 hr after surgery). However, this study included a more homogenous patient population, consisting only of patients with lower gastrointestinal tract perforation and excluding other abdominal conditions, such as cholecystitis, intestinal obstruction, and gastroduodenal perforation. Another difference is that the overall 28-day mortality in this study was lower than that in the studies cited above,
0.279
suggesting that less severe patients may have been included in this study. In this study, patients were stratified into subgroups according to the use of noradrenaline and number of organ dysfunctions as indicators of severity. As a result, none of the studied subgroups benefited from PMX treatment. However, the number of patients with more than three organ failures was small. One possible interpretation of our findings is that the routine use of postoperative PMX for less severely ill patients is not advantageous. Several limitations of this study need to be acknowledged. First, although the study population was large, this study was based on a retrospective observational design without randomization. Although a propensity score method was used to adjust for differences in baseline characteristics and severity of the condition, the results may still have been biased by unmeasured confounders. Examples of such confounders are the cancer stage among patients with malignancies, the degree of fecal dissemination into the abdominal cavity, and changes in the patients’ general condition after the operation; these could have affected the assignment of patients to PMX treatment
The 28-Day Mortality in the Propensity-Matched Groups With Subgroup Analysis by the Use of Noradrenaline and Number of Organ Dysfunctions Table 3.
Polymyxin B Hemoperfusion Variable
No. of Patients
Control
28-Day Mortality (%)
No. of Patients
28-Day Mortality (%)
p
Use of noradrenaline No
293
40 (13.7)
276
38 (13.8)
0.968
Yes
297
61 (20.5)
314
58 (18.5)
0.519
Number of organ dysfunctions 1
78
8 (10.3)
74
8 (10.8)
0.911
2
389
65 (16.7)
404
65 (16.1)
0.814
3
112
23 (20.5)
104
20 (19.2)
0.810
11
5 (45.5)
8
3 (37.5)
1.000
≥4
Critical Care Medicine
www.ccmjournal.org
1191
Iwagami et al
Table 4. Logistic Regression Analysis for 28-Day Mortality in the Propensity-Matched 1,180 Patients Univariate Variable
OR
95% CI
Multivariate p
OR
95% CI
p
Treatment Control
Reference
Polymyxin B hemoperfusion
1.06
Reference 0.78–1.44
0.696
1.10
0.80–1.51
0.569
Age (yr) Reference
≤ 64
Reference
65–74
1.63
0.89–2.97
0.111
1.65
0.89–3.05
0.115
75–84
2.30
1.33–3.99
0.003
2.33
1.31–4.13
0.004
≥ 85
4.19
2.34–7.49
< 0.001
4.35
2.36–8.01
< 0.001
Sex Male
Reference
Reference
Female
1.26
0.92–1.71
0.152
1.12
0.81–1.56
0.496
Maintenance hemodialysis
2.90
1.86–4.53
< 0.001
3.43
2.14–5.51
< 0.001
Use of noradrenaline
1.52
1.12–2.08
0.008
1.42
1.01–1.99
0.043
Use of dobutamine
1.62
1.04–2.52
0.032
1.39
0.87–2.23
0.166
Red cell concentrate transfusion
1.79
1.31–2.45
< 0.001
1.37
0.98–1.93
0.068
Number of organ dysfunctions 1
Reference
Reference
2
1.67
0.96–2.89
0.069
1.26
0.70–2.27
0.434
3
2.11
1.14–3.91
0.017
1.60
0.83–3.10
0.162
≥4
6.18
2.17–17.63
0.001
5.09
1.69–15.34
0.004
OR = odds ratio.
and outcomes. Second, the absence of records on vital signs, blood tests, and blood cultures in the DPC database precluded a more rigorous definition of septic shock (23). Third, successful control of the septic foci by surgery and antibiotics is of paramount importance in patient survival. It is highly unlikely that a surgeon would conclude the surgical intervention without detecting, removing, or repairing the site of perforation or without adequate irrigation of the abdominal cavity. Previous studies have in fact reported that unsuccessful elimination of the septic focus during the initial laparotomy was rare—less than 10% of cases (24, 25). However, our DPC data do not allow us to confirm this assumption. In addition, the number, type, dose, and frequency of antibiotics used varied from patient to patient, and it was extremely difficult to examine the appropriateness of each antibiotic. A lack of more detailed information about surgery and the use of antibiotics is a limitation of this study. Fourth, this study was unable to evaluate changes in physiological variables and dose of vasopressor agents, which have been used as surrogate endpoints in clinical trials. Finally, unlike in randomized trials, the duration and blood flow rate of PMX may vary in actual clinical practice. 1192
www.ccmjournal.org
CONCLUSIONS The present propensity analysis using a large nationwide database demonstrates that postoperative early use of PMX for abdominal septic shock did not improve mortality in the overall study population or any studied subgroup. A large multicentered prospective randomized trial is required to clarify the role of PMX for abdominal and other Gram-negative septic episodes.
REFERENCES
1. Sands KE, Bates DW, Lanken PN, et al; Academic Medical Center Consortium Sepsis Project Working Group: Epidemiology of sepsis syndrome in 8 academic medical centers. JAMA 1997; 278:234–240 2. van Deventer SJ, Büller HR, ten Cate JW, et al: Experimental endotoxemia in humans: Analysis of cytokine release and coagulation, fibrinolytic, and complement pathways. Blood 1990; 76:2520–2526 3. Taveira da Silva AM, Kaulbach HC, Chuidian FS, et al: Brief report: Shock and multiple-organ dysfunction after self-administration of Salmonella endotoxin. N Engl J Med 1993; 328:1457–1460 4. Opal SM, Scannon PJ, Vincent JL, et al: Relationship between plasma levels of lipopolysaccharide (LPS) and LPS-binding protein in patients with severe sepsis and septic shock. J Infect Dis 1999; 180:1584–1589 May 2014 • Volume 42 • Number 5
Clinical Investigations 5. Marshall JC, Foster D, Vincent JL, et al; MEDIC study: Diagnostic and prognostic implications of endotoxemia in critical illness: Results of the MEDIC study. J Infect Dis 2004; 190:527–534 6. Opal SM, Glück T: Endotoxin as a drug target. Crit Care Med 2003; 31:S57–S64 7. Shoji H: Extracorporeal endotoxin removal for the treatment of sepsis: Endotoxin adsorption cartridge (Toraymyxin). Ther Apher Dial 2003; 7:108–114 8. Cruz DN, Perazella MA, Bellomo R, et al: Effectiveness of polymyxin B-immobilized fiber column in sepsis: A systematic review. Crit Care 2007; 11:R47 9. Vincent JL, Laterre PF, Cohen J, et al: A pilot-controlled study of a polymyxin B-immobilized hemoperfusion cartridge in patients with severe sepsis secondary to intra-abdominal infection. Shock 2005; 23:400–405 10. Komatsu S, Shimomatsuya T, Nakajima M, et al: Severity scoring systems for prognosis and efficacy of polymyxin B-immobilized fiber treatment for colonic perforation. Surg Today 2006; 36:807–810 11. Cruz DN, Antonelli M, Fumagalli R, et al: Early use of polymyxin B hemoperfusion in abdominal septic shock: The EUPHAS randomized controlled trial. JAMA 2009; 301:2445–2452 12. Shimizu T, Hanasawa K, Sato K, et al; PMX Treatment Study Group: Direct hemoperfusion with polymyxin-B-immobilized fiber columns improves septic hypotension and reduces inflammatory mediators in septic patients with colorectal perforation. Langenbecks Arch Surg 2009; 394:303–311 13. Shimizu T, Hanasawa K, Sato K, et al; PMX Treatment Study Group in Septic Patients with Colorectal Perforation: The clinical significance of serum procalcitonin levels following direct hemoperfusion with polymyxin B-immobilized fiber column in septic patients with colorectal perforation. Eur Surg Res 2009; 42:109–117 14. Sumitani M, Uchida K, Yasunaga H, et al: Prevalence of malignant hyperthermia and relationship with anesthetics in Japan: Data from the diagnosis procedure combination database. Anesthesiology 2011; 114:84–90
Critical Care Medicine
15. Ono K, Wada K, Takahara T, et al: Indications for computed tomography in patients with mild head injury. Neurol Med Chir (Tokyo) 2007; 47:291–297 16. Griswold ME, Localio AR, Mulrow C: Propensity score adjustment with multilevel data: Setting your sites on decreasing selection bias. Ann Intern Med 2010; 152:393–395 17. Ayanian JZ, Landrum MB, Guadagnoli E, et al: Specialty of ambulatory care physicians and mortality among elderly patients after myocardial infarction. N Engl J Med 2002; 347:1678–1686 18. Yasunaga H, Horiguchi H, Kuwabara K, et al: Outcomes after laparoscopic or open distal gastrectomy for early-stage gastric cancer: A propensity-matched analysis. Ann Surg 2013; 257:640–646 19. Rosenbaum P, Rubin D: Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985; 39:33–38 20. Vincent JL, de Mendonça A, Cantraine F, et al: Use of the SOFA score to assess the incidence of organ dysfunction/failure in intensive care units: Results of a multicenter, prospective study. Working group on “sepsis-related problems” of the European Society of Intensive Care Medicine. Crit Care Med 1998; 26:1793–1800 21. 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 22. Vincent JL: Polymyxin B hemoperfusion and mortality in abdominal septic shock. JAMA 2009; 302:1968; author reply 1969–1970 23. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992; 20:864–874 24. Wacha H, Hau T, Dittmer R, et al: Risk factors associated with intraabdominal infections: A prospective multicenter study. Peritonitis Study Group. Langenbecks Arch Surg 1999; 384:24–32 25. van Ruler O, Mahler CW, Boer KR, et al; Dutch Peritonitis Study Group: Comparison of on-demand vs planned relaparotomy strategy in patients with severe peritonitis: A randomized trial. JAMA 2007; 298:865–872
www.ccmjournal.org
1193
Objectives: To examine the effect of postoperative polymyxin B hemoperfusion on mortality in patients with abdominal septic shock triggered by lower gastrointestinal tract perforation, identifying subpopulations of patients who may benefit from this treatment. Design: Propensity-matched analysis. Setting: We used a nationwide inpatient database in Japan. Patients: We included patients who are 18 years old or older hospitalized during a period of 34 months between July 2007 and October 2011, who had open abdominal surgery on the day of admission (day 0) for perforation of lower gastrointestinal tract, *See also p. 1309. 1 Department of Hemodialysis and Apheresis, The University of Tokyo Hospital, Tokyo, Japan. 2 Department of Health Economics and Epidemiology Research, School of Public Health, The University of Tokyo, Tokyo, Japan. 3 Department of Emergency and Critical Care Medicine, The University of Tokyo Hospital, Tokyo, Japan. 4 Department of Clinical Data Management and Research, Clinical Research Center, National Hospital Organization Headquarters, Tokyo, Japan. 5 Department of Health Informatics and Policy, Tokyo Medical and Dental University Graduate School of Medicine, Tokyo, Japan. Drs. Iwagami and Yasunaga had full access to all the study data and take responsibility for the integrity of the data and accuracy of the data analysis. All the authors directly participated in the planning, execution, or analysis of the study and read and approved the final submitted version of this article. Supported, in part, by grants H22-Policy-031 (Research on Policy Planning and Evaluation) from the Ministry of Health, Labour and Welfare, Japan; 22390131 (Scientific Research B) from Ministry of Education, Culture, Sports, Science and Technology, Japan; and 0301002001001 (Funding Program for World-Leading Innovative R&D on Science and Technology, FIRST program) from the Council for Science and Technology Policy, Japan. The funders had no role in the execution of this study or interpretation of results. Drs. Yasunaga and Fushimi’s institutions received grant support from the Japanese government. The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2013 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0000000000000150
Critical Care Medicine
and who required noradrenaline and/or dopamine. We excluded patients who died on day 0 or 1 and patients starting polymyxin B hemoperfusion on day 2 or later. Measurements and Main Results: The main outcome was 28-day mortality. Of 2,925 eligible patients, 642 received one or two polymyxin B hemoperfusion sessions, starting the first one on day 0 or 1. Propensity score matching created a matched cohort of 1,180 patients (590 pairs with and without polymyxin B hemoperfusion). The 28-day mortality was 17.1% (101 of 590) in the polymyxin B hemoperfusion group and 16.3% (96 of 590) in the control group (p = 0.696). Subgroup analyses by number of polymyxin B hemoperfusion sessions (one or two), timing of polymyxin B hemoperfusion initiation (day 0 or 1), the use of noradrenaline, and number of dysfunctional organs (one to six) did not show any significant difference in 28-day mortality between the groups. Multiple logistic did not show a significant association between the use of polymyxin B hemoperfusion and 28-day mortality (adjusted odds ratio, 1.10; 95% CI, 0.80–1.51; p = 0.569). Age, end-stage renal disease requiring maintenance hemodialysis, the use of noradrenaline, and number of dysfunctional organs were positively associated with 28-day mortality. Conclusions: In this retrospective study, postoperative polymyxin B hemoperfusion did not show any survival benefit for the overall study population or any of the studied subgroups of patients with abdominal septic shock. A large multicentered prospective randomized trial is warranted to identify the true role of polymyxin B hemoperfusion in sepsis caused by Gram-negative bacteria. (Crit Care Med 2014; 42:1187–1193) Key Words: abdominal septic shock; endotoxin; polymyxin b hemoperfusion; propensity-matched analysis
S
epsis, also known as “systemic inflammatory response syndrome” associated with infection, is a serious medical condition with high mortality (1). Endotoxin, a lipopolysaccharide derived from the outer membrane of Gram-negative bacteria, is recognized as a key factor in the sepsis cascade www.ccmjournal.org
1187
Iwagami et al
because it triggers a series of inflammatory reactions that lead to organ dysfunction (2, 3); its high serum level is linked to increased risk of multiple organ failure and death (4, 5). Thus, endotoxin has been regarded as a therapeutic target in treating sepsis (6). Endotoxin adsorption by direct hemoperfusion with polymyxin B-immobilized fiber column has been developed as an endotoxin-targeting therapy because polymyxin B has a high affinity for endotoxin through hydrophobic and ionic forces (7). A meta-analysis suggested that it can improve mortality in addition to mean arterial pressure, dopamine use, and mean Pao2/Fio2 ratio; however, the studies included in that analysis were of suboptimal quality (8). Abdominal infection is a major cause of sepsis (1), and it can be severe, particularly when caused by lower gastrointestinal tract perforation; this is because dissemination of feces containing intestinal Gram-negative bacteria into the abdominal cavity easily leads to bacteremia, with high levels of endotoxin. Many studies have investigated the efficacy of postoperative polymyxin B hemoperfusion (PMX) in patients with abdominal sepsis to identify improvements, mainly in cardiovascular function (9–13). However, controversy remains as to whether postoperative PMX can reduce mortality in such patients, partly because of the paucity of studies with adequate sample size. In Japan, PMX for treating septic shock has been covered by the Japanese health insurance system since 1994 and is therefore widely used in clinical practice. A postmarketing evaluation with a large sample size may help clarify the effectiveness of PMX, especially in reducing mortality. The purpose of this study was to investigate the survival benefit of postsurgical PMX in a homogenous target population of patients with lower gastrointestinal tract perforation requiring open abdominal surgery by means of the Japanese Diagnosis Procedure Combination (DPC) inpatient database.
MATERIALS AND METHODS Data Source The details of the DPC database are described elsewhere (14). Briefly, the DPC is a case-mix patient classification system, whereby administrative claims data and detailed medical data are collected for all inpatients discharged from participating hospitals. Until 2010, data were collected over a 6-month period every year between July 1 and December 31. Since 2011, this data collection has been conducted all year round. The DPC database includes the following information: patients’ age and sex; main diagnosis, comorbidities at admission and postadmission complications classified according to the International Classification of Diseases, 10th Revision (ICD-10) and text data in Japanese; medical procedures, including types of surgery, coded with original Japanese codes; daily records of drug administration and devices used; length of stay; and discharge status. All clinical data for each patient are recorded at discharge by attending physicians. To optimize the accuracy of recorded diagnoses, responsible physicians are obliged to record diagnoses with reference to medical charts. The dates 1188
www.ccmjournal.org
of hospital admission, surgery, discharge, bedside procedures performed, and drugs administered are recorded using a uniform data-submission format. In this study, hospital volume was defined as the average annual number of surgical procedures performed on eligible patients; a unique identifier was used for each hospital. All patient identifiers were removed from this database. Because of the anonymous nature of the data, the need for informed consent was waived. Study approval was obtained from the Institutional Review Board of the University of Tokyo. PMX Since 1994, PMX has been covered by the Japanese health insurance system for patients with sepsis who require vasoactive support with endotoxemia or in whom Gram-negative infection is suspected. A maximum of two hemoperfusion sessions are covered, provided they are not performed on the same day. The only commercially available polymyxinB-immobilized column for adults is Toraymyxin-20R (Toray Industries, Tokyo, Japan). Patient Selection Among 17 million inpatients in the DPC database over a total of 34 months between July 1, 2007, and October 31, 2011 (for 2007–2010, 6 mo annually from July 1 to December 31; for 2011, 10 mo from January 1 to October 31), patients who are 18 years old or older satisfying the following inclusion criteria were selected: 1) diagnosed with nontraumatic perforation of the lower gastrointestinal tract (including the small bowel, colon, and rectum, but not the appendix); 2) underwent open abdominal surgery except exploratory laparotomy on the day of admission (day 0); and 3) started noradrenaline and/or dopamine on day 0. The following patients were excluded: 1) those who died on day 0 or 1 and 2) those who underwent PMX starting on day 2 or later. Among the eligible patients, those whose first PMX session was on day 0 or 1 were identified. For the PMX group, the number of hemoperfusion sessions performed (one or two on consecutive days) was calculated. Variables and Endpoint Regarding underlining conditions, end-stage renal disease that required maintenance hemodialysis was assessed based on the records of the devices used. Liver cirrhosis and malignancy were determined according to ICD-10. The findings associated with surgery were assessed as follows. Perforation sites were categorized into small bowel, colon, rectum, or unknown when described only as “lower gastrointestinal tract perforation,” using ICD-10 in conjunction with the Japanese text data. Red cell concentrate transfusion and fresh-frozen plasma transfusion on day 0 were also taken into account. The presence of organ dysfunction for six organs at admission was defined as follows: 1) cardiovascular dysfunction that required initiation of noradrenaline and/or dopamine on day 0; 2) respiratory dysfunction that required postsurgical continuous mechanical ventilation; 3) renal dysfunction May 2014 • Volume 42 • Number 5
Clinical Investigations
that necessitated initiation of intermittent acute hemodialysis or continuous renal replacement therapy on day 0—except for maintenance hemodialysis patients; 4) hepatic dysfunction as a comorbidity at admission for recorded “liver failure”—except for liver cirrhosis patients; 5) hematologic dysfunction that required platelet concentrate transfusion on day 0; and 6) neurological dysfunction of 100 on the Japan Coma Scale score, which is equivalent to scores of 6–9 on the Glasgow Coma Scale, or more severe. The Japan Coma Scale was recorded in all patients at admission to assess the consciousness level, and it correlated well with the Glasgow Coma Scale (15). The only endpoint used in this study was 28-day mortality. Statistical Analysis Propensity score matching was used to adjust for differences in baseline characteristics and severity of condition at admission between patients with and without PMX (16). This was because previous retrospective observational studies have demonstrated that propensity score methods are a powerful tool for comparing groups with similar observed characteristics without specifying the relationship between confounders and outcomes (17, 18). First, the propensity score was estimated. The log odds of the probability that a patient received PMX treatment was modeled for potential confounders: age; sex; underlining conditions; findings related to surgery; use of noradrenaline, dopamine, and dobutamine; the six organ dysfunctions defined above; hospital volume; and type of hospital (university or nonuniversity hospital). C-statistic was calculated to evaluate the goodness of fit. A one-to-one matched analysis using nearest-neighbor matching was performed based on the estimated propensity scores of each patient. A match occurred when one patient in the PMX group had an estimated score within 0.25 sds of another in the control group (19). Descriptive statistics were presented for all patients and propensity-matched patients. Continuous variables were compared using t test, and categorical variables were compared using chi-square test. For a comparison of 28-day mortality, Fisher exact test or chi-square test was used among the propensity-score-matched groups. Subgroup analyses were also performed among the propensity-score-matched groups according to the number of sessions, day of the first hemoperfusion session, the use of noradrenaline, and number of organ dysfunctions. Patients discharged alive before day 28 were regarded as survivors: lacking hospice facilities or longterm acute-care facilities, the Japanese healthcare system allows patients to stay in the same hospital until recovery is complete. Univariate and multivariate logistic analyses were also performed for the propensity-matched patients to identify independent predictors of 28-day mortality. In a nonstepwise manner, we selected variables showing statistically significant differences between survivors and nonsurvivors with chisquare test. A p value of less than 0.05 was considered statistically significant. All statistical analyses were conducted using IBM SPSS version 20 (IBM SPSS, Armonk, NY). Critical Care Medicine
RESULTS Among 3,139 patients who satisfied the inclusion criteria, we excluded the following: 1) 193 patients who died on day 0 or 1 and 2) 21 patients who underwent the first PMX session on day 2 or later. The remaining 2,925 eligible patients were categorized into the PMX group (n = 642) or control group (n = 2,283), from which 590 propensity score-matched pairs were generated (Fig. 1). C-statistic indicated that the goodness of fit was 0.735 in the propensity score model. Table 1 shows the baseline characteristics of the unmatched PMX and control groups (n = 2,925) and of the propensity score-matched groups (n = 1,180). When the unmatched groups were compared, patients on maintenance hemodialysis, those who underwent colorectal perforation, those who underwent red cell concentrate and fresh-frozen plasma transfusion, those receiving noradrenaline and dobutamine, and those with respiratory, renal, and/or hematologic dysfunction were more likely to receive PMX treatment. After propensity score matching, the baseline patient characteristics were well balanced between the groups. Overall 28-day mortality was 13.4% (393 of 2,925). Table 2 shows the 28-day mortality for the propensity-matched groups: it was 17.1% (101 of 590) for the PMX group and 16.3% (96 of 590) for the control group (p = 0.696). Subgroup analyses according to the number of PMX sessions and day of the first hemoperfusion session revealed no statistically significant difference. Table 3 presents a comparison of 28-day mortality among the propensity-matched groups according to the use of noradrenaline and number of organ dysfunctions; there was no significant difference. Table 4 shows the results of the univariate and multivariate logistic regression analyses for 28-day mortality. No significant association was found between PMX treatment and 28-day mortality (adjusted odds ratio, 1.10; 95% CI, 0.80–1.51; p = 0.569), whereas age, end-stage renal disease requiring maintenance hemodialysis, the use of noradrenaline, and number of organ dysfunctions were positively associated with 28-day mortality.
Figure 1. Outline of patient selection. PMX = polymyxin B hemoperfusion. www.ccmjournal.org
1189
Iwagami et al
Table 1.
Baseline Patient Characteristics in Unmatched and Propensity-Matched Groups Unmatched Groups
Variable
Age (yr, mean ± sd) Sex (male), %
Polymyxin B Hemoperfusion (n = 642)
Control (n = 2,283)
74.5 ± 10.9
74.7 ± 11.9
Matched Groups Polymyxin B Hemoperfusion (n = 590)
Control (n = 590)
p
0.719
74.5 ± 10.8
74.9 ± 11.2
0.565
p
45.6
47.4
0.431
45.8
44.7
0.726
Maintenance hemodialysis
8.1
5.5
0.016
8.5
8.5
1.000
Liver cirrhosis
0.8
0.5
0.367
0.8
0.5
0.726
24.8
26.5
0.389
24.7
25.3
0.840
6.9
17.6
< 0.001
7.5
7.5
0.999
73.2
66.8
72.7
72.4
18.2
14.1
18.0
18.3
1.7
1.5
1.9
1.9
Red cell concentrate transfusion, %
33.5
28.7
0.019
32.5
33.2
0.804
Fresh-frozen plasma transfusion, %
34.6
23.6
< 0.001
32.5
31.9
0.803
Noradrenaline
53.1
36.6
< 0.001
50.3
53.2
0.322
Dopamine
82.4
85.9
0.030
83.2
83.9
0.753
Dobutamine
10.1
7.1
0.010
10.3
11.4
0.574
Underlining conditions, %
Malignancy Findings related to surgery Perforation site, % Small bowel Colon Rectum Unknown
a
Type of vasoactive drug, %
Organ dysfunctions at admission, %b Cardiovascular
100
100
1.000
100
100
1.000
Respiratory
86.1
57.9
< 0.001
84.9
86.6
0.405
Renal
20.4
5.1
< 0.001
13.7
11.9
0.338
Hematologic
4.2
2.5
0.022
4.1
3.9
0.882
Hepatic
0.8
0.4
0.175
0.7
0.5
1.000
Neurological
6.5
4.7
0.066
6.3
4.9
0.311
Described only as “lower gastrointestinal tract perforation,” using the International Classification of Diseases, 10th Revision, in conjunction with the Japanese text data. b Organ dysfunctions at admission were defined as follows: 1) cardiovascular dysfunction that required the use of at least one vasopressor on day 0; 2) respiratory dysfunction that necessitated postsurgical continuous mechanical ventilation; 3) renal dysfunction that required the initiation of intermittent acute hemodialysis or continuous renal replacement therapy on day 0—except for maintenance hemodialysis patients; 4) hepatic dysfunction recorded as “liver failure” at admission—except for liver cirrhosis patients; 5) hematologic dysfunction that necessitated platelet concentrate transfusion on day 0; and 6) neurological dysfunction at admission of 100 on the Japan Coma Scale (scores of 6–9 on the Glasgow Coma Scale) or more severe. a
DISCUSSION In this study, a propensity score analysis was performed using a nationwide database to examine the effect of PMX on mortality in patients with abdominal septic shock. Although the analysis of baseline patient characteristics in the unmatched groups indicated more active utilization of PMX among patients with more severe conditions, the one-to-one propensity score matching successfully balanced patient characteristics between the treated and untreated groups. In particular, adjusting for the proportion of organ dysfunctions was essential because 1190
www.ccmjournal.org
the number of dysfunctional organs and mortality have been shown to be strongly correlated (20, 21). The results showed that PMX treatment did not present any survival benefit for the overall study population—regardless of the number of sessions and the day of initiation. Likewise, none of the subgroup analyses performed in this study indicated a significant association of mortality with PMX treatment. Many studies have examined the effect of postoperative PMX on abdominal sepsis (9–13). A pilot randomized study in 36 postsurgical patients with abdominal sepsis indicated that May 2014 • Volume 42 • Number 5
Clinical Investigations
Table 2.
The 28-Day Mortality in the Propensity-Matched Groups With Subgroup Analyses No. of Patients
28-Day Mortality (%)
p
Polymyxin B hemoperfusion
590
101 (17.1)
0.696
Control
590
96 (16.3)
2 sessions
277
48 (17.3)
1 session
313
53 (16.9)
Control
590
96 (16.3)
Variables
Number of polymyxin B hemoperfusion sessions 0.919
The day of the first session of polymyxin B hemoperfusion Day 0
428
67 (15.7)
Day 1
162
34 (21.0)
Control
590
96 (16.3)
one session of PMX within 24 hours after emergency surgery (within 48 hr after elective surgery) improved cardiac function and reduced the need for continuous renal replacement therapy—but not 28-day mortality (9). A subsequent randomized controlled trial in 64 postsurgical patients with abdominal septic shock demonstrated that two sessions of PMX with a 24-hour interval (the first starting within 24 hr after emergency surgery) improved hemodynamics and Sequential Organ Failure Assessment score (11). However, there was a lack of statistical significance in the odds ratio of the crude 28-day mortality (22). This study had similar findings to the above studies in that the total number of PMX sessions was one or two, and the first session was performed on postoperative day 0 (up to 24 hr after surgery) or day 1 (up to 48 hr after surgery). However, this study included a more homogenous patient population, consisting only of patients with lower gastrointestinal tract perforation and excluding other abdominal conditions, such as cholecystitis, intestinal obstruction, and gastroduodenal perforation. Another difference is that the overall 28-day mortality in this study was lower than that in the studies cited above,
0.279
suggesting that less severe patients may have been included in this study. In this study, patients were stratified into subgroups according to the use of noradrenaline and number of organ dysfunctions as indicators of severity. As a result, none of the studied subgroups benefited from PMX treatment. However, the number of patients with more than three organ failures was small. One possible interpretation of our findings is that the routine use of postoperative PMX for less severely ill patients is not advantageous. Several limitations of this study need to be acknowledged. First, although the study population was large, this study was based on a retrospective observational design without randomization. Although a propensity score method was used to adjust for differences in baseline characteristics and severity of the condition, the results may still have been biased by unmeasured confounders. Examples of such confounders are the cancer stage among patients with malignancies, the degree of fecal dissemination into the abdominal cavity, and changes in the patients’ general condition after the operation; these could have affected the assignment of patients to PMX treatment
The 28-Day Mortality in the Propensity-Matched Groups With Subgroup Analysis by the Use of Noradrenaline and Number of Organ Dysfunctions Table 3.
Polymyxin B Hemoperfusion Variable
No. of Patients
Control
28-Day Mortality (%)
No. of Patients
28-Day Mortality (%)
p
Use of noradrenaline No
293
40 (13.7)
276
38 (13.8)
0.968
Yes
297
61 (20.5)
314
58 (18.5)
0.519
Number of organ dysfunctions 1
78
8 (10.3)
74
8 (10.8)
0.911
2
389
65 (16.7)
404
65 (16.1)
0.814
3
112
23 (20.5)
104
20 (19.2)
0.810
11
5 (45.5)
8
3 (37.5)
1.000
≥4
Critical Care Medicine
www.ccmjournal.org
1191
Iwagami et al
Table 4. Logistic Regression Analysis for 28-Day Mortality in the Propensity-Matched 1,180 Patients Univariate Variable
OR
95% CI
Multivariate p
OR
95% CI
p
Treatment Control
Reference
Polymyxin B hemoperfusion
1.06
Reference 0.78–1.44
0.696
1.10
0.80–1.51
0.569
Age (yr) Reference
≤ 64
Reference
65–74
1.63
0.89–2.97
0.111
1.65
0.89–3.05
0.115
75–84
2.30
1.33–3.99
0.003
2.33
1.31–4.13
0.004
≥ 85
4.19
2.34–7.49
< 0.001
4.35
2.36–8.01
< 0.001
Sex Male
Reference
Reference
Female
1.26
0.92–1.71
0.152
1.12
0.81–1.56
0.496
Maintenance hemodialysis
2.90
1.86–4.53
< 0.001
3.43
2.14–5.51
< 0.001
Use of noradrenaline
1.52
1.12–2.08
0.008
1.42
1.01–1.99
0.043
Use of dobutamine
1.62
1.04–2.52
0.032
1.39
0.87–2.23
0.166
Red cell concentrate transfusion
1.79
1.31–2.45
< 0.001
1.37
0.98–1.93
0.068
Number of organ dysfunctions 1
Reference
Reference
2
1.67
0.96–2.89
0.069
1.26
0.70–2.27
0.434
3
2.11
1.14–3.91
0.017
1.60
0.83–3.10
0.162
≥4
6.18
2.17–17.63
0.001
5.09
1.69–15.34
0.004
OR = odds ratio.
and outcomes. Second, the absence of records on vital signs, blood tests, and blood cultures in the DPC database precluded a more rigorous definition of septic shock (23). Third, successful control of the septic foci by surgery and antibiotics is of paramount importance in patient survival. It is highly unlikely that a surgeon would conclude the surgical intervention without detecting, removing, or repairing the site of perforation or without adequate irrigation of the abdominal cavity. Previous studies have in fact reported that unsuccessful elimination of the septic focus during the initial laparotomy was rare—less than 10% of cases (24, 25). However, our DPC data do not allow us to confirm this assumption. In addition, the number, type, dose, and frequency of antibiotics used varied from patient to patient, and it was extremely difficult to examine the appropriateness of each antibiotic. A lack of more detailed information about surgery and the use of antibiotics is a limitation of this study. Fourth, this study was unable to evaluate changes in physiological variables and dose of vasopressor agents, which have been used as surrogate endpoints in clinical trials. Finally, unlike in randomized trials, the duration and blood flow rate of PMX may vary in actual clinical practice. 1192
www.ccmjournal.org
CONCLUSIONS The present propensity analysis using a large nationwide database demonstrates that postoperative early use of PMX for abdominal septic shock did not improve mortality in the overall study population or any studied subgroup. A large multicentered prospective randomized trial is required to clarify the role of PMX for abdominal and other Gram-negative septic episodes.
REFERENCES
1. Sands KE, Bates DW, Lanken PN, et al; Academic Medical Center Consortium Sepsis Project Working Group: Epidemiology of sepsis syndrome in 8 academic medical centers. JAMA 1997; 278:234–240 2. van Deventer SJ, Büller HR, ten Cate JW, et al: Experimental endotoxemia in humans: Analysis of cytokine release and coagulation, fibrinolytic, and complement pathways. Blood 1990; 76:2520–2526 3. Taveira da Silva AM, Kaulbach HC, Chuidian FS, et al: Brief report: Shock and multiple-organ dysfunction after self-administration of Salmonella endotoxin. N Engl J Med 1993; 328:1457–1460 4. Opal SM, Scannon PJ, Vincent JL, et al: Relationship between plasma levels of lipopolysaccharide (LPS) and LPS-binding protein in patients with severe sepsis and septic shock. J Infect Dis 1999; 180:1584–1589 May 2014 • Volume 42 • Number 5
Clinical Investigations 5. Marshall JC, Foster D, Vincent JL, et al; MEDIC study: Diagnostic and prognostic implications of endotoxemia in critical illness: Results of the MEDIC study. J Infect Dis 2004; 190:527–534 6. Opal SM, Glück T: Endotoxin as a drug target. Crit Care Med 2003; 31:S57–S64 7. Shoji H: Extracorporeal endotoxin removal for the treatment of sepsis: Endotoxin adsorption cartridge (Toraymyxin). Ther Apher Dial 2003; 7:108–114 8. Cruz DN, Perazella MA, Bellomo R, et al: Effectiveness of polymyxin B-immobilized fiber column in sepsis: A systematic review. Crit Care 2007; 11:R47 9. Vincent JL, Laterre PF, Cohen J, et al: A pilot-controlled study of a polymyxin B-immobilized hemoperfusion cartridge in patients with severe sepsis secondary to intra-abdominal infection. Shock 2005; 23:400–405 10. Komatsu S, Shimomatsuya T, Nakajima M, et al: Severity scoring systems for prognosis and efficacy of polymyxin B-immobilized fiber treatment for colonic perforation. Surg Today 2006; 36:807–810 11. Cruz DN, Antonelli M, Fumagalli R, et al: Early use of polymyxin B hemoperfusion in abdominal septic shock: The EUPHAS randomized controlled trial. JAMA 2009; 301:2445–2452 12. Shimizu T, Hanasawa K, Sato K, et al; PMX Treatment Study Group: Direct hemoperfusion with polymyxin-B-immobilized fiber columns improves septic hypotension and reduces inflammatory mediators in septic patients with colorectal perforation. Langenbecks Arch Surg 2009; 394:303–311 13. Shimizu T, Hanasawa K, Sato K, et al; PMX Treatment Study Group in Septic Patients with Colorectal Perforation: The clinical significance of serum procalcitonin levels following direct hemoperfusion with polymyxin B-immobilized fiber column in septic patients with colorectal perforation. Eur Surg Res 2009; 42:109–117 14. Sumitani M, Uchida K, Yasunaga H, et al: Prevalence of malignant hyperthermia and relationship with anesthetics in Japan: Data from the diagnosis procedure combination database. Anesthesiology 2011; 114:84–90
Critical Care Medicine
15. Ono K, Wada K, Takahara T, et al: Indications for computed tomography in patients with mild head injury. Neurol Med Chir (Tokyo) 2007; 47:291–297 16. Griswold ME, Localio AR, Mulrow C: Propensity score adjustment with multilevel data: Setting your sites on decreasing selection bias. Ann Intern Med 2010; 152:393–395 17. Ayanian JZ, Landrum MB, Guadagnoli E, et al: Specialty of ambulatory care physicians and mortality among elderly patients after myocardial infarction. N Engl J Med 2002; 347:1678–1686 18. Yasunaga H, Horiguchi H, Kuwabara K, et al: Outcomes after laparoscopic or open distal gastrectomy for early-stage gastric cancer: A propensity-matched analysis. Ann Surg 2013; 257:640–646 19. Rosenbaum P, Rubin D: Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985; 39:33–38 20. Vincent JL, de Mendonça A, Cantraine F, et al: Use of the SOFA score to assess the incidence of organ dysfunction/failure in intensive care units: Results of a multicenter, prospective study. Working group on “sepsis-related problems” of the European Society of Intensive Care Medicine. Crit Care Med 1998; 26:1793–1800 21. 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 22. Vincent JL: Polymyxin B hemoperfusion and mortality in abdominal septic shock. JAMA 2009; 302:1968; author reply 1969–1970 23. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992; 20:864–874 24. Wacha H, Hau T, Dittmer R, et al: Risk factors associated with intraabdominal infections: A prospective multicenter study. Peritonitis Study Group. Langenbecks Arch Surg 1999; 384:24–32 25. van Ruler O, Mahler CW, Boer KR, et al; Dutch Peritonitis Study Group: Comparison of on-demand vs planned relaparotomy strategy in patients with severe peritonitis: A randomized trial. JAMA 2007; 298:865–872
www.ccmjournal.org
1193
