This investigation was performed using three different datasets extracted from a multicenter nationwide cohort study conducted in Japan. The first dataset, the J-septic DIC dataset, was compiled in 42 intensive care units (ICUs) between January 2011 and December 2013. 2 The second dataset, the FORECAST dataset, was compiled in 59 ICUs between January 2016 and March 2017, 14 and the third dataset, the SPICE dataset, was compiled in 22 ICUs between December 2017 and May 2018. 15 In the first two datasets, patients were eligible for the registry if they were diagnosed as having severe sepsis or septic shock according to the conventional criteria proposed by the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) consensus conference in 1991 16 and were 18 years of age or older. In the present analysis, we included as the underlying diseases targeted by the JAAM-2 DIC criteria only those of sepsis patients diagnosed using the Sepsis-3 criteria (i.e., SOFA score of 2 or more points). 13

The exclusion criteria included the use of warfarin/acetylsalicylic acid/thrombolytic therapy before study entry; a history of fulminant hepatitis, decompensated liver cirrhosis, or other serious liver disorder; a history of hematologic malignant disease; other conditions increasing the risk of bleeding; treatment with any chemotherapy at study entry; treatment with warfarin before or after study entry; and patients with missing data for any hemostatic markers used for calculating JAAM DIC criteria.

This study followed the principles of the Declaration of Helsinki. The fundamental study protocol was approved by the Institutional Review Board of Osaka General Medical Center (approval numbers: #25–2050, #30-S11–004, and #S201901009). Due to the anonymous and retrospective nature of this study, the board of each hospital waived the need for informed consent.

A case report form was developed for the three datasets used in this study on which the following information was recorded: age, sex, disease severity scores on the day of ICU admission, the source of ICU admission, pre-existing conditions, new organ dysfunction, primary source of infection, and concomitant therapies against sepsis. The severity of illness was evaluated at study entry according to the Acute Physiology and Chronic Health Evaluation (APACHE) II score and SIRS score. The Sequential Organ Failure Assessment (SOFA) score was used to assess organ dysfunction, which was defined as a SOFA subscore ≥2 for each organ. 17 The primary outcome measure was all-cause in-hospital mortality.

We proposed novel DIC criteria named the JAAM-2 DIC criteria that were modified from the original JAAM DIC criteria. The underlying diseases targeted by the JAAM-2 DIC criteria, which comply with those of the original JAAM DIC criteria, are shown in Table 1 . 9 The SIRS score component from the JAAM DIC criteria was omitted, and the cutoff value for diagnosing DIC was set at 3 points or more ( Table 2 ).

To identify the differences between the original JAAM and modified JAAM-2 DIC criteria that distinguished nonsurviving from surviving patients with sepsis, receiver operating characteristic (ROC) analyses were performed. The target condition was set as in-hospital mortality.

We evaluated associations between in-hospital mortality and anticoagulant therapy according to the status of DIC or not to clarify the significance of the two DIC criteria in determining when to introduce anticoagulant therapy against sepsis. Participants were categorized into two groups: the anticoagulant group, comprising patients who received any anticoagulant therapy such as antithrombin, recombinant human thrombomodulin, heparin/heparinoid, and serine protease inhibitors, and the control group, comprising patients who received no anticoagulant therapy. Due to the retrospective nature of this analysis, there were baseline imbalances between the two groups; therefore, an adjusted mortality analysis was performed using propensity scoring as described below. The SPICE dataset was not used for this analysis due to the lack of data on anticoagulant therapy.

Descriptive statistics were calculated as medians (interquartile range) or proportions, as appropriate. Univariate differences between the groups were assessed using the Mann–Whitney U test, Kruskal–Wallis test, chi-squared test, or Fisher's exact test. A p -value of <0.05 indicated statistical significance. All statistical analyses were performed using STATA software version 15.0 (Stata Corp, College Station, Texas, United States).

The overall effectiveness of anticoagulant therapy on mortality was assessed using a Cox regression model with inverse probability-of-treatment weighting using the propensity scores. The propensity score for receiving anticoagulant therapy was calculated using multivariate logistic regression and included 25 independent variables for the J-septic DIC cohort and 30 variables for the FORECAST cohort, including age, sex, disease severity, source of ICU admission, past medical history of severe conditions, new organ dysfunctions, ICU characteristics, primary source of infection, causal microorganisms, anticoagulant therapy not for DIC, and other therapeutic interventions ( Supplementary Table S1 [available in the online version]). Hazard ratio and estimated 95% confidence interval were calculated along with estimated survival curves.

The patient flow diagram is shown in Fig. 1 . During the study period, 3,195 consecutive patients fulfilling the inclusion criteria were registered in the J-Septic DIC registry database. After excluding 1,040 patients who met at least one exclusion criterion, we analyzed 2,154 patients in the final study cohort. The anticoagulant group comprised 1,089 patients, and the control group comprised 1,065 patients. Similarly, we enrolled 817 patients from the FORECAST registry and 608 patients from the SPICE-ICU registry in the final study cohort.

Baseline characteristics of the study population are shown in Table 3 , Supplementary Table S2 and S3 (available in the online version). Patient characteristics such as age and sex were similar between the three datasets. After applying an inverse probability of treatment weighting with propensity score, patient characteristics, such as illness severity, as indicated by SOFA, APACHE II, and DIC scores and the rate of new organ dysfunction, were well matched between the anticoagulant and control groups.

ROC curves for the original JAAM and modified JAAM-2 DIC criteria as predictors of in-hospital mortality are shown in Fig. 2 . Consistent with the three different datasets, the curves for both the JAAM and JAAM-2 DIC criteria were almost entirely consistent with each other. These data suggested that in predicting short-term mortality, use of the JAAM DIC and JAAM-2 DIC criteria was considered to be equivalent.

Survival curves for the anticoagulant and control groups in the propensity score-weighted prediction model according to DIC status diagnosed using the two criteria are shown in Fig. 3 (J-septic DIC dataset) and Fig. 4 (FORECAST dataset). Consistent with both criteria and both datasets, favorable effects of anticoagulant therapy were observed only in the patient subsets with DIC, whereas differences in mortality between the anticoagulant and control groups in the subsets without DIC were not significant. These findings were consistent between the two datasets and suggested that to determine the optimal target of anticoagulant therapy for sepsis, use of the JAAM DIC and JAAM-2 DIC criteria for diagnosing DIC was considered to be equivalent.

On the basis of the study results, we proposed modified JAAM DIC criteria that omitted the SIRS criteria but included platelet count, fibrin degradation products (or D-dimer), and prothrombin time, and the cutoff value for diagnosing DIC was set at 3 points or more. We named these new DIC criteria the JAAM-2 DIC criteria.

Using three different datasets constructed in Japan, we verified the prognostic value and diagnostic value of the JAAM-2 compared with the original JAAM DIC criteria for initiating anticoagulation. Consequently, ROC analysis revealed that curves for both the original JAAM and JAAM-2 DIC criteria were almost consistent with each other. Survival analysis revealed that the curves of the anticoagulant and control groups for both the original JAAM and JAAM-2 DIC criteria were also almost entirely consistent with each other. Thus, the newly proposed JAAM-2 DIC criteria could potentially be used as an alternative to the original JAAM DIC criteria in clinical practice.

Several different clinical practice guidelines for DIC have been developed by societies in Britain, 18 Japan, 19 and Italy, 20 along with the harmonized guidance by the ISTH. 21 Some distinct discrepancies in the appraisal of diagnostic criteria for DIC exist between these guidelines. The Japanese and Italian clinical practice guidelines recommend the use of either the JMHW, ISTH, or the JAAM criteria, whereas the British guideline recommends the use of the ISTH criteria. The guidelines do not offer consistent recommendations on diagnosing DIC, and thus, there is currently no definitive agreement as to which of these criteria is superior to the other. The present study does not aim to discuss the diagnostic value of the several DIC criteria because we have no gold standard for DIC diagnosis. No meta-analysis has been conducted so far to compare the prognostic performance among the several available DIC criteria. Nonetheless, we showed that the clinical usefulness of the proposed JAAM-2 DIC criteria was nearly equivalent to that of the traditional JAAM DIC criteria. While a discussion on superiority would be worthless, we showed that the performance of the JAAM-2 scoring system is almost identical to that of the JAAM criteria in terms of mortality prediction and determining treatment timing.

Modification of the JAAM DIC criteria has been discussed in several studies so far. Umemura et al 22 proposed unified DIC criteria involving several hemostatic endothelial molecular markers based on the JAAM DIC criteria and showed that the addition of protein C activity and plasminogen activator inhibitor 1 to the original JAAM DIC criteria resulted in greater prognostic value than the original criteria. Iba et al 23 proposed replacing the SIRS score with antithrombin activity in the JAAM DIC criteria. They validated the proposed criteria using a dataset of 819 sepsis patients and found that using AT-based DIC criteria makes it possible to discriminate a more coagulation disorder-specific population. All of these previous attempts were in addition to or replacements of the other variables instead of the SIRS score, and thus, the burden on clinicians still remained. In the present study, we simply omitted the SIRS score, so this modification of the JAAM DIC criteria should allow a totally clinical-friendly approach.

We intended to evaluate the severity of sepsis by adding “SIRS score ≥ 3”; however, the present study showed the prognosis to be not different without this item included. Therefore, we think it is reasonable to omit the SIRS item from the JAAM criteria. In the present analysis, as the JAAM-2 DIC criteria have been shown to increase clinical simplicity without diminishing any diagnostic performance, the educational activities led by our academic society will aid in the replacement of original JAAM with JAAM-2 DIC criteria in Japan. Furthermore, it will be necessary to verify coherence with the sepsis-induced coagulopathy criteria 24 proposed by the ISTH in the future.