Intensive Care Med (2000) 26: 1786±1793 DOI 10.1007/s001340000710
Shigeto Oda Hiroyuki Hirasawa Takao Sugai Hidetoshi Shiga Kazuya Nakanishi Nobuya Kitamura Tomohito Sadahiro Takeshi Hirano
Received: 2 April 2000 Final revision received: 26 July 2000 Accepted: 12 September 2000 Published online: 31 October 2000 Springer-Verlag 2000 This work was supported by Grantin-Aid for Scientific Research (C) from The Ministry of Education, Science, Sports and Culture.
)
S. Oda ( ) ´ H. Hirasawa ´ T. Sugai ´ H. Shiga ´ K. Nakanishi ´ N. Kitamura ´ T. Sadahiro ´ T. Hirano Department of Emergency and Critical Care Medicine, Chiba University School of Medicine, 1-8-1 Inohana, Chuou, Chiba 260±8677, Japan E-mail:
[email protected] Phone: +81-43-2 22 71 71 Fax: +81-43-2 26 23 71
O R I GI N A L
Comparison of Sepsis-related Organ Failure Assessment (SOFA) score and CIS (cellular injury score) for scoring of severity for patients with multiple organ dysfunction syndrome (MODS)
Abstract Objective: To evaluate the usefulness of cellular injury score (CIS) and Sepsis-related Organ Failure Assessment (SOFA) score for determination of the severity of multiple organ dysfunction syndrome (MODS). Design: A prospective observational study. Setting: A medical and surgical intensive care unit (ICU) of a teaching hospital. Patients: Forty-seven consecutive MODS patients. Measurements and results: SOFA score and CIS were measured every day for 12 months for 47 MODS patients. Comparison was made of the SOFA score and CIS for usefulness in the scoring of severity of MODS in 26 survivors and 21 non-survivors. In addition, receiver operating characteristics (ROC) analysis was used to determine the usefulness of these two indexes as predictors of prognosis. No significant differences were found on admission between the survivors and non-survivors, but significant differences between the two subgroups (p < 0.001) were found in maximum value within 1 week after admission and maximum value during the course of
Introduction Multiple organ dysfunction syndrome (MODS) still has a poor prognosis and ranks at the top of the list of causes of death in the ICU [1, 2, 3]. In the treatment of MODS, accurate evaluation of severity is important in determining
treatment for both indexes. Analysis of changes after admission indicated that significant differences between survivors and non-survivors began to appear on day 3 of admission for both indexes; at that time SOFA score began to deteriorate in the non-survivors while CIS began to improve in the survivors. ROC analysis demonstrated that the area under the ROC curve was 0.769 for SOFA scores and 0.760 for CIS. Conclusions: Both SOFA score and CIS sequentially reflected the severity of MODS. Furthermore, they were comparable in diagnostic value as predictors of prognosis. These findings may indicate the possibility that MODS is a summation of effects of cellular injury. In addition, sequential evaluation of both SOFA score and CIS would provide a more accurate prediction of prognosis than conventional methods. Key words Index severity illness ´ Multiple organ dysfunction syndrome ´ Intensive care units ´ Sepsis-related Organ Failure Assessment (SOFA) ´ Cellular injury score (CIS)
therapeutic approach, assessing benefits of various treatments, evaluating new therapeutic procedures, comparing lifesaving rates among medical centers and explaining patients' conditions to family members [4, 5, 6]. Various scoring systems have been developed and used clinically to determine the severity of MODS [7,
1787
8, 9, 10]. However, these scoring systems have been designed primarily to compare therapeutic results among multiple centers or to estimate prognosis; they have not been intended to determine the severity of MODS in individual patients as a key factor in making decisions concerning what type of treatment should be provided on a daily basis [11]. Furthermore, many physiological parameters used in such scoring systems are readily affected by treatment and, although such systems are useful for evaluation of severity at admission to the ICU, they cannot be used for serial evaluation during treatment or they are not suited for real-time evaluation of therapy-related improvement of the pathological condition or its progression. Thus, they are not appropriate for determination of therapeutic effects or medical decision-making for individual patients. Particularly with regard to an important issue in current lifesaving medicine, proper utilization of medical resources, they cannot be employed to decide whether treatment should be withheld or withdrawn from patients who are unlikely to survive. Such defects are compensated for by Sepsis-related (or Sequential) Organ Failure Assessment (SOFA) score, which has recently been developed by the working group on sepsis-related problems of the European Society of Intensive Care Medicine [11, 12]. The SOFA score was developed to describe the severity of MODS in individual patients through determination of the severity of organ dysfunction, with focus on morbidity rather than mortality. The SOFA score is also intended for sequential assessment, since it takes temporal factors into account [11]. It has been reported that SOFA score enables individual determination of severity and repeated measurements of multiple organ dysfunction/ failure in multicenter prospective studies, and thereby functions as an index for determining either sequential deterioration or improvement of the pathological condition during treatment [12]. Based on our view of many years that organ dysfunction is the summation of effects of cellular dysfunction in vital organs, we continue to believe that cellular injury should be evaluated if the severity of patients with organ dysfunction is to be correctly assessed [13, 14]. We have developed and clinically used the cellular injury score (CIS) to evaluate the severity of cellular injury in patients with MODS. CIS is calculated from three parameters of intracellular metabolism (arterial ketone body ratio (AKBR) [15, 16], osmolality gap (OG) [17, 18], and blood lactate level). We have already reported that CIS may well reflect the outcome of multiple organ failure (MOF) patients and that, since it is less vulnerable to effects of treatment than other scoring systems, it may be possible to determine changes in the severity of MODS in individual patients on a real-time basis through sequential evaluation of CIS [14].
In this study, we compared the usefulness of the SOFA score and CIS for evaluation of the severity of MODS by measuring them prospectively and sequentially.
Materials and methods Of patients admitted to our general ICU between January and December 1998, 47 were enrolled in the study who had organ dysfunction with a SOFA score [11] of 2 or higher for at least two organs and who remained in the ICU for more than 48 h. The underlying diseases in those patients included severe sepsis/septic shock (13 patients), hematological disease (6), acute hepatic failure (5), circulatory disease (4), respiratory disease, severe acute pancreatitis and autoimmune disease (3 patients each), hemorrhagic shock (2), trauma, burns and poisoning (1 patient each) and 5 others. We continued to measure SOFA score and CIS prospectively every day during the ICU stay. The worst value obtained within 24 h after admission was employed for calculation of a SOFA score. Blood samples for CIS measurement were collected from the arterial line on admission and every morning thereafter. Measurements of AKBR, OG and lactate level and calculation of CIS were performed as already described [14]. Briefly, levels of ketone bodies, acetoacetate and b-hydroxybutyrate, on the basis of which AKBR was calculated, were determined using KETOLEX Sanwa (Sanwa Chemical, Nagoya, Japan) as a reagent by the enzymatic method with a KETO 340 spectrophotometer (Sanwa Chemical, Nagoya, Japan). The serum osmolality, from which OG was calculated, was measured with a Fiske OS osmometer (Fiske Associate, Burlington, Mass.). The predicted osmolality was calculated on the basis of biochemical values obtained every morning and OG was calculated from the difference between the actually determined osmolality and the predicted osmolality [17]. The lactate level in arterial blood was determined by the enzymatic method with a Determiner LA (Kyowa Medex, Tokyo, Japan). All blood drawing was performed in routine clinical fashion. This study was approved by the institutional ethics committee. On the basis of the daily SOFA score and CIS, we compared the scores on admission, the maximum values within 1 week and the maximum values over the course of treatment together with the patterns of changes in values between the subgroups of survivors and non-survivors, in order to determine the usefulness of SOFA score and CIS as severity scoring systems. In addition, we determined the correlation between SOFA score and CIS. We also performed receiver operating characteristics (ROC) analysis [19] in order to compare the usefulness of SOFA score and CIS for prediction of prognosis. The mean values for the subgroups of survivors and non-survivors were compared by two-tailed Student's t-test and values are presented as means standard deviation. Repeated measure ANOVA was used to test the patterns of change in SOFA score and CIS in the subgroups of survivors and non-survivors. The correlation between SOFA score and CIS was determined by simple regression analysis. For ROC analysis, the cut-off points were determined separately for SOFA score and CIS to calculate sensitivity and specificity, on the basis of which ROC curves were prepared. Furthermore, areas under the ROC curves [20] were calculated to compare the usefulness of SOFA score and CIS as predictors of prognosis. Findings of p less than 0.05 were considered significant.
1788
Fig. 1 Comparison of SOFA scores between survivors and non-survivors at ICU admission, maximum value within 1 week and maximum value during ICU stay. Values are means SD; p values determined by unpaired Student's t-test
Results Of the 47 patients enrolled in the study, 26 survived and 21 others died. Table 1 shows the demographic data for the subgroups of survivors and non-survivors. There were no differences between the two subgroups in distribution of age and sex; the mean Acute Physiology And Chronic Health Evaluation II (APACHE II) score on admission was 20.7 10.0 for the survivors versus 22.7 5.6 for the non-survivors, and this difference was not significant. Figure 1 compares SOFA score on admission, maximum value within 1 week and maximum value over the course of treatment for the subgroups of survivors and non-survivors. The mean SOFA score on admission was 9.9 3.2 for survivors and 10.0 3.3 for non-survivors; this difference was not significant. However, the maximum value within 1 week was 11.8 2.9 for survivors and 15.3 3.3 for non-survivors, with the latter significantly higher than the former (p < 0.001). The maximum value over the course of treatment was also significantly higher for non-survivors than for survivors (16.7 3.9 versus 12.3 3.1, respectively; p < 0.001).
Table 1 Demographic data for the study population. Values are given as the means SD Number of patients Male/Female Age (years) APACHE II ICU stay (days)
Survivors
Non-survivors
26 17/9 51 19.1 20.7 10.0 17.5 12.8
21 12/9 60 15.3 22.7 5.6 16.9 16.0
Figure 2 compares CIS data. As for SOFA, the mean value on admission was 2.3 2.2 for survivors and 3.4 2.5 for non-survivors, without a significant difference between the two subgroups. However, the maximum value within 1 week and the maximum value over the course of treatment were 2.8 2.1 and 2.8 2.0, respectively, for the survivors, and 5.1 2.4 and 5.7 1.8, respectively, for the non-survivors, with both of the latter values significantly higher than the former (p < 0.001, respectively). Figure 3 shows how the mean SOFA scores for survivors and non-survivors changed over the 1st week after admission. The survivors continued to have a nearly constant score during that period, while the non-survivors began to have significantly higher scores after admission (p = 0.0013), with the mean for non-survivors significantly higher than that of the survivors on day 3 of ICU admission and thereafter. Figure 4 shows how the mean CIS changed over time. As for SOFA score, significant differences were noted in fluctuations in score between the two subgroups (p = 0.0009), with the difference between subgroups being significant from day 3 of ICU admission onward. When we compared the pattern of change in SOFA score with that of CIS, we found that SOFA score of non-survivors began to rise significantly on day 3 of a ICU stay, whereas the CIS of survivors significantly decreased on that day. We then examined the correlation between SOFA score and CIS. Figure 5 shows the correlation of the maximum value over the course of treatment between SOFA score and CIS for individual patients. There was a significantly positive correlation, though not a strong one, between the two; no patients survived when the
1789
Fig. 2 Comparison of CIS between survivors and non-survivors at ICU admission, maximum value within 1 week and maximum value during ICU stay. Values are means SD; p values determined by unpaired Student's t-test
Fig. 3 Serial changes in SOFA score after ICU admission. Open circles show the means of survivors and closed circles those of non-survivors. Error bars show standard deviations; p values determined by repeated-measures ANOVA. Comparison between survivors and non-survivors was performed using unpaired Student's t-test
maximum value over the course of treatment was 20 or higher for SOFA and 8 or higher for CIS. We then performed ROC analysis in order to determine the usefulness of SOFA score and CIS for prediction of prognosis. For SOFA score, mortality was 76.9 % with a score of 14 or higher and 23.1 % with a score of 13 or lower, with a cut-off of 14 yielding the best results, with sensitivity of 71.3 % and specificity of 76.9 %. For CIS, on the other hand, mortality was 73.7 % with a score of 5 or higher
and 23.1 % with a score of 4 or lower, with a cut-off of 5 yielding a sensitivity of 73.7 % and specificity of 76.9 %. Calculation of areas under the ROC curve for the two scoring systems yielded comparable values (0.769 for SOFA score and 0.760 for CIS).
1790
Fig. 4 Serial changes in CIS after ICU admission. Open circles show the means of survivors and closed circles those of nonsurvivors. Error bars show standard deviations; p values determined by repeated-measures ANOVA. Comparison between survivors and non-survivors was performed using unpaired Student's t-test
Fig. 5 Correlation between maximum CIS and SOFA scores within 1 week after ICU admission
Discussion Since pathological conditions of MODS are complicated, it is difficult to make a correct evaluation of the severity. Thus, various scoring systems have been proposed to facilitate evaluation of the severity of MODS
[7, 8, 9, 10]. These scoring systems may be classified into two major types, one in which dysfunction of individual organs is assessed using organ-specific parameters [7, 8] and the other in which systemic severity is assessed through evaluation of abnormalities in various physiological parameters [9, 10]. The former is repre-
1791
sented by MOF score, proposed by Goris [7], and MOD score, by Marshall [8], while the latter is represented by APACHE II [9] score and APACHE III [10] score. The APACHE II score has been widely used for the evaluation of severity in critically ill patients. Many of the conventional scoring systems for organ dysfunction primarily assess the severity of organ dysfunction by using an organ-specific simple parameter and assessing the extent of its abnormality. However, with recent advances in supportive care, the parameters of organ dysfunction employed in the conventional scoring systems might not always correctly reflect the extent of organ dysfunction at the time of evaluation. For instance, serum creatinine level, which is used as a parameter of renal dysfunction in many scoring systems, is now maintained at a lower level through the use of continuous renal replacement therapy (CRRT) and hence does not any longer reflect the extent of renal dysfunction. Similarly, it is clear that many physiological parameters employed in APACHE II and APACHE III are readily affected by various treatments including administration of vasoactive drugs, mechanical ventilation and renal supportive therapy. These severity scoring systems are therefore suitable for assessing the severity of MODS at the time of admission to the ICU before the patient has been treated, but are problematic if used for sequential evaluation thereafter. Furthermore, they are difficult to use on a daily basis. In addition, it has been reported that APACHE II does not always reflect the severity of MODS in patients belonging to certain groups [21, 22, 23]. In fact, Civetta et al. [21] found that neither the length of stay in the ICU nor mortality correlated with APACHE II for surgical ICU patients. Similarly, Cerra et al. [23] reported that APACHE II did not always predict manifestations of MOF and mortality; they noted, instead, that changes over time in four parameters, PaO2/FIO2 ratio, creatinine, bilirubin and lactate (the latter two not included in APACHE II) were more useful in predicting death. They suggested that these parameters, indicating the severity of cell injury, might better predict prognosis. We continue to believe that organ dysfunction is essentially the summation of injuries to cells forming vital organs, and that the severity of cellular injury should therefore be assessed if organ dysfunction/failure is to be correctly evaluated. To facilitate evaluation of cellular injury, we have developed CIS and reported its usefulness as a scoring system for the severity of MODS/ MOF [13, 14]. CIS is calculated from three parameters of intracellular metabolism (arterial ketone body ratio (AKBR), osmolality gap (OG) [17, 18] and blood lactate). AKBR is the ratio of acetoacetate to b-hydroxybutyrate levels in arterial blood, and reflects the redox state of mitochondria within hepatocytes [15]. AKBR is known to reflect accurately the energy charge within
hepatocytes [16], and it has been reported that hepatocyte injury or reduction in hepatic blood flow can decrease AKBR below the normal level of 0.7 [15]. OG is the difference between the actually determined osmolality and the predicted osmolality calculated from various biochemical values, and is presumed to represent the amount of immeasurable solutes present in the blood, which is spilled over from injured cells [17]. We have already reported that the OG increases in patients with organ dysfunction and correlates well with the severity of MOF [17, 18]. Lactate is a parameter of anaerobic metabolism and is commonly used as a parameter of organ ischemia. The CIS includes these three parameters of intracellular metabolism, scores each from 0 to 3 points and calculates the total. Each of these three parameters can be conveniently measured in the ICU and assessed at the bedside on a real-time basis. We have demonstrated in previous studies that CIS correlates well with the number of failing organs and the mortality of MOF patients, that it is a useful severity scoring system for individual patients since it may be assessed sequentially, and that it is also useful as a predictor of prognosis [13, 14]. Sepsis-related Organ Failure Assessment score was developed as an index to compensate for the defects of the above-mentioned conventional scoring systems for organ dysfunction and to assess the extent of dysfunction of individual organs conveniently and sequentially [11, 12]. The conventional scoring systems for organ dysfunction have employed one parameter for each organ, whereas SOFA score is designed to describe the extent of organ dysfunction in each patient at the time of evaluation without being affected by the treatment, using not only physiological parameters but also the presence or absence of mechanical ventilation, the amount of catecholamines and the amount of urine volume as parameters [11]. It has been demonstrated that SOFA score can describe the severity of dysfunction of each organ in patients with MODS/MOF and that, in addition, it may be useful for prediction of prognosis [12]. In the present study, we measured SOFA score daily, which permitted sequential evaluation together with CIS, a method we employ to determine the severity of MOF in patients with MODS. The purpose of this study was neither to compare the predictability of prognosis for specific patient groups nor to compare results between centers, but to determine whether SOFA score and CIS were useful for making therapeutic decisions for individual patients with MODS and for comparative assessment of therapeutic benefits. The underlying diseases of the patients involved in this study varied, ranging from sepsis to trauma. This patient selection was based on our belief that a scoring system for severity of MODS must reflect pathological conditions of MODS irrespective of the underlying disease [14]. For those groups of patients SOFA score and CIS,
1792
like APACHE II, exhibited no differences at the time of admission to the ICU between the subgroups of survivors and non-survivors, but both yielded significantly higher scores for the subgroup of non-survivors in terms of maximum value within 1 week and maximum value over the course of treatment. Furthermore, monitoring of sequential changes in scores demonstrated that the nonsurvivors began to have significantly higher scores than survivors from the 3rd day of ICU stay. These results suggested that SOFA score and CIS might accurately reflect the severity of MODS and enable real-time evaluation of its severity. Vincent et al. found that total SOFA score could clearly distinguish whether a patient could survive and accurately determine outcome, yet they suggested that total SOFA score should not be used carelessly as a predictor of prognosis [12], since SOFA was originally developed to determine the degree of dysfunction of each organ, and each organ has a variable influence on the prognosis in different patients. However, the present study demonstrated that total SOFA score accurately reflects the severity of MODS in individual patients at the time of evaluation, as does CIS, and that SOFA is comparable to conventional MOD scores [8] in usefulness for determination of severity. Notably, there were differences in pattern of change between SOFA score and CIS. With SOFA score, the survivors continued to have almost the same score as that observed on admission, whereas the non-survivors tended to exhibit increases in score over time. With CIS, in contrast, the non-survivors exhibited consistently higher scores, while the survivors tended to show decreases over time. These findings indicate that CIS reflects the improvement of pathological conditions produced by treatment more accurately than SOFA score does [14], suggesting that monitoring of changes in CIS might be advantageous for evaluation of the therapeutic effect on MODS. Examination of the correlation between maximum SOFA score and CIS over the course of treatment revealed a significantly positive correlation between the two, although the correlation was not very strong. This may be explained as follows : SOFA score, an index reflecting the severity of dysfunction in each organ, may naturally be expected to reach maximum value immediately before death, while CIS, an index of the extent of cellular injury, is likely to reach maximum value when cellular injury in each organ is most deteriorated; it is not always consistent with the time immediately before death [14]. It may thus be presumed that a time lag exists between the degree of organ dysfunction indicated by SOFA score and that of cellular injury indicated by CIS. We have already reported that CIS does not always reach maximum value immediately before death in patients receiving prolonged treatment, although CIS may reflect real-time changes in the severity of individual patients with acute progression of disease [14]. This
suggests that SOFA score might be more useful than CIS for evaluation of the severity of MODS in patients undergoing prolonged treatment. We next compared SOFA score and CIS as predictors of prognosis by performing ROC analysis of the maximum value obtained within 1 week of treatment. We chose the maximum value obtained within 1 week of treatment because we supposed that this would be a sufficient predictor of prognosis (including the progress of treatment) for MODS patients. We found that the best prediction of mortality in those subgroups of patients was obtained with a cut-off of 14 for SOFA score and 5 for CIS. Furthermore, calculation of the area under the ROC curves for the two systems yielded comparable results, indicating that they have a comparable value in terms of predicting prognosis. In their multicenter joint study, Vincent et al. found that, with total SOFA score of 15 or higher, mortality reached 90 % with a sensitivity of 31 % and specificity of 99 % at that time [12]. In the present study, mortality was 76.9 % with a sensitivity of 71.3 % and a specificity of 76.9 % when SOFA reached a maximum value of 14 or higher in the course of treatment. These differences in values might have resulted from differences in therapeutic approaches to MODS as well as from study design (multicenter joint study versus single-center study). In the present study, on the other hand, death could not be avoided for any of the patients with SOFA scores of 20 or higher or CIS of 8 or higher. The results of Vincent et al., as well as our own, suggest that SOFA score and CIS could provide an index for the decision of whether treatment should be restricted or withdrawn for patients for whom preservation of life is considered unlikely. However, when using these scores for decision-making, we have to be very cautious because no severity scoring system, including SOFA and CIS, is sufficiently reliable in making such a critical decision. The two systems were found to be comparable in predicting prognosis, but some differences were noted between them in patterns of fluctuation; in addition, their scores were not always completely correlated with each other. These findings suggest that a more accurate prediction of prognosis can be obtained if changes in values of these two systems are followed simultaneously. Further studies of these two systems appear to be needed, since the present study was a single-center study involving a limited number of patients. Nevertheless, our study demonstrated that both SOFA score and CIS are useful as scoring systems for every patient with MODS. An especially valuable suggestion of this study is that sequential evaluation by combined use of both scoring systems is useful for evaluation of therapeutic outcome and therapeutic policy. In addition, the demonstration of a certain degree of correlation between CIS and SOFA scores may indicate the possibility that organ failure is a summation of the effects of cellular injury [13].
1793
References 1. Members of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference Committee (1992) 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 20: 864±874 2. Tran DD, Groeneveld ABJ, Van der Meulen J, Nauta JJP, Strack van Schijndel RJM, Thijs LG (1990) Age, chronic disease, sepsis, organ system failure and mortality in a medical intensive care unit. Crit Care Med 18: 474±479 3. Deitch EA (1992) Multiple organ failure: pathophysiology and potential future therapy. Ann Surg 216: 117±134 4. Wisner DH (1992) History and current status of scoring systems for critical care. Arch Surg 127: 352±356 5. Arregui LM, Moyes DG, Lipman J, Fatti LP (1991) Comparison of disease severity scoring systems in septic shock. Crit Care Med 19: 1165±1171 6. Barriere SL, Lowry SF (1995) An overview of mortality risk prediction in sepsis. Crit Care Med 23: 376±393 7. Goris RJA, Te Boekhorst TPA, Nuytinck JKS, Gimbrere JSF (1985) Multiple-organ failure: generalized autodestructive inflammation? Arch Surg 120: 1109±1115 8. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL, Sibbald WJ (1995) Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med 23: 1638±1652
9. Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985) APACHE II: a severity of disease classification system. Crit Care Med 13: 818±829 10. Knaus WA, Wagner DP, Draper EA, Zimmerman JE, Bergner M, Bastos PG, Sirio CA, Murphy DJ, Lotring T, Damiano A, Harrell FE Jr (1991) The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized adults. Chest 100: 1619±1638 11. Vincent JL, Moreno R, Takala J, Willats S, De Mendoca A, Bruining H, Reinhart CK, Suter PM, Thijs LG (1996) The SOFA (sepsis-related organ failure assessment) score to describe organ dysfunction/failure. Intensive Care Med 22: 707±710 12. Vincent JL, De Mendonca A, Cantraine F, Moreno R, Takala J, Suter PM, Sprung CL, Colardyn F, Blecher S (1998) Use the SOFA score to assess the incidence of organ dysfunction/failure in intensive care units; results of a multicenter, prospective study. Crit Care Med 26: 1793±1800 13. Hirasawa H, Sugai T, Ohtake Y, Oda S, Shiga H, Matsuda K, Kitamura N (1991) Cellular injury score as an index of the severity of patients with multiple organ failure (abstract). Circ Shock 34: 75 14. Oda S, Hirasawa H, Sugai T, Shiga H, Matsuda K, Ueno H (1998) Cellular injury score for multiple organ failure severity scoring system. J Trauma 45: 304±311 15. Ozawa K (1983) Biological significance of mitochondrial redox potential in shock and multiple organ failure: redox theory. In: Lefer AM, Schumer W, Alan R (eds) Molecular and Cellular Aspects of Shock and Trauma. Liss, New York, pp 39±66
16. Ozawa K, Chance B, Tanaka A, Iwata S, Kitai T, Ikai I (1992) Linear correlation between acetoacetate/beta-hydroxybutyrate in arterial blood and oxidized flavoprotein/reduced pyridine nucleotide in freeze-trapped human liver tissue. Biochim Biophys Acta 1138: 350±352 17. Inaba H, Hirasawa H, Mizuguchi T (1987) Serum osmolality gap in postoperative patients in intensive care. Lancet 2: 1331±1335 18. Hirasawa H, Odaka M, Sugai T, Ohtake Y, Inaba H, Tabata Y, Kobayashi H, Isono K (1988) Prognostic value of serum osmolality gap in patients with multiple organ failure treated with hemopurification. Artif Organs 12: 382±387 19. McNeil BJ, Keeler E, Adelstein SJ (1975) Primer on certain elements of medical decision making. N Engl J Med 293: 211±215 20. Hanley JA, McNeil BJ (1982) The meaning and use of the area under a receiver operating characteristic(ROC) curve. Radiology 143: 29±36 21. Civetta JM, Hudson-Civetta JA, Nelson LD (1990) Evaluation of APACHE II for cost containment and quality assurance. Ann Surg 212: 266±276 22. McAnena OJ, Moore FA, Moore EE, Mattox KL, Marx JA, Pepe P (1992) Invalidation of the APACHE II scoring system for patients with acute trauma. J Trauma 33: 504±507 23. Cerra FB, Negro F, Abrams J (1990) APACHE II score does not predict multiple organ failure or mortality in postoperative surgical patients. Arch Surg 125: 519±522
