This is a prospective, single-center study investigating the diagnostic performance of plasma adrenomedullin (total and mature AM) for sepsis and septic shock defined by Sepsis-3, and the prognostic performance of AM for 28-day mortality of patients with sepsis. In this study, we demonstrated that (1) AM is a competent biomarker with the greatest diagnostic accuracy for sepsis and septic shock among the biomarkers we evaluated, (2) AM can accurately diagnose sepsis and septic shock at ICU admission, and (3) AM has the potential to prognose 28-day mortality for patients with sepsis, especially from Day 2 to Day 5 post-ICU admission, as do SOFA score and lactate levels.
Sepsis-3 definitions and sepsis biomarkers
Sepsis is a life-threatening event caused by microbial infection-induced multiple organ failure. In 1992, the American College of Chest Physicians and the Society of Critical Care Medicine (SCCM) first proposed the definition and diagnosis of sepsis based on the SIRS criteria, referred to as Sepsis-1 [35]. Subsequently, in 2001, the SCCM, the American College of Chest Physicians, the European Society of Intensive Care Medicine, the American Thoracic Society, and the Surgical Infection Society refined the diagnostic criteria of sepsis, called Sepsis-2, and defined “severe sepsis” [36]. The SCCM and the European Society of Intensive Care Medicine published the Sepsis-3 definition and diagnostic criteria in 2016, which removed SIRS and severe sepsis from the original Sepsis-1 concept, and redefined sepsis using the SOFA score [29]. As described by Fang and colleagues, when two or more SIRS criteria (Sepsis-1 and − 2) or the increment in SOFA score ≥ 2 (Sepsis-3) were applied to predict 21-day all-cause mortality in infected patients without prior chronic organ dysfunction, the sensitivity was 96.0% or 91.0%, and the specificity was 8.3% or 21.9%, respectively [37]. Sepsis-3 diagnostic criteria may narrow the sepsis population, because the SOFA score represents the severity of the condition, whereas the SIRS score represents a clinically evident host response to infection.
Diagnosis by Sepsis-3 definitions is based on identifying the infection and SOFA scoring. Although hematological, biochemical, and microbiological laboratory tests are essential to diagnose sepsis, culture-based infection diagnosis takes time [8], and failure to identify sepsis in the early stages delays effective treatment, resulting in high mortality. Thus, major efforts have been made to find biomarkers that allow early diagnosis of sepsis [7, 8]. Recently, TREM-1, HMGB1, PCT, and PSEP were identified and validated as sepsis biomarkers. PCT, widely regard as one of the most useful sepsis biomarkers, is recommended by the Surviving Sepsis Campaign Guideline (SSCG) [29]; however, PCT is not an ideal diagnostic biomarker. For patients on steroids, or patients with local infection alone or infection with atypical bacteria, PCT results can be falsely negative. Moreover, PCT levels may increase after cardiac arrest or surgery, and in patients with severe trauma [8]. Thus, in this study, we evaluated the blood concentration of AM, which is elevated in sepsis [19–28], as an additional potential biomarker to diagnose sepsis and septic shock defined by Sepsis-3.
Two molecular forms of AM in circulation
Kitamura et al. reported that two major molecular forms of AM circulate in the blood of humans, biologically active mature AM (mAM) with an amidated C-terminus, and inactive intermediate AM (iAM) with a non-amidated C-terminal glycine [30, 31]. Numerous prior studies assessed AM levels in plasma and tissue, but the immunoreactive AM detected in these studies was the tAM level, which includes both mAM and iAM. The clinical significance of iAM is still unclear. Some reports showed similar changes in mAM and iAM [33, 38, 39]. Although iAM is biologically inactive, one study demonstrated that following conversion to mAM, it exerted vasodilatory action in isolated rat aortas ex vivo and took much longer to achieve maximum relaxation than mAM, suggesting that iAM may act as a hormone reservoir [40]. Bioactive AM would be an informative biomarker; however, its short half-life (approximately 22 minutes) and the AM-binding protein, complement factor H, hinder the reliable measurement of bioactive AM [27, 28, 41]. To address these problems, mid-regional pro-adrenomedullin (MR-proADM or proADM), another fragment of the AM precursor peptide, has been introduced as an alternative biomarker [27, 28]. Although it had been presumed that MR-proADM represented the sum of synthesized iAM and mAM, the ratio between MR-proADM and bioactive AM can vary, thus MR-proADM does not always reflect the actual amount of bioactive AM [40]. Recently, Weber et al. reported a new assay to reliably measure the bioactive mature form of AM (which they referred to as bio-ADM) [42], and bio-ADM in patients with sepsis is more strongly associated with clinical outcome relative to MR-proADM concentrations [24]. In our mAM assay, the limits of detection and quantitation were 0.133 and 0.085 pM, respectively, according to Clinical and Laboratory Standards Institute (CLSI) protocols. Intra- and inter-assay coefficients of variation were 1.8% and 5.1%, respectively [33]. However, we did not assess the impact of complement factor H on our mAM assay. Thus, we believe our mAM assay is as reliable as the bio-ADM assay reported by Weber et al., unless complement factor H concentration is especially high.
AM and other clinical study parameters
Consistent with previous studies, we observed that at ICU admission, tAM and mAM were moderately elevated in patients with sepsis (without shock), whereas, they were highly elevated in patients with septic shock [19–28]. In contrast, SOFA score, PSEP, and lactate levels in patients with sepsis (without shock) were not significantly different from those in the non-sepsis group. The SOFA score and lactate levels of non-sepsis patients were elevated because the non-sepsis group contained patients who had severe diseases with organ failure or organ dysfunction such as acute heart failure, acute pancreatitis, acute hepatic failure, and acute interstitial pneumonitis. When non-sepsis patients were admitted to the ICU, the average PSEP level was 663 ng/L, which was as same as cutoff value (670 ng/L) for sepsis diagnosis reported by Nakamura et al [43]. As described in previous reports [43, 44], this high PSEP level may be associated with renal dysfunction, because the creatinine level of non-sepsis patients was also elevated.
Diagnostic performance of AM at ICU admission for patients with sepsis and septic shock, as defined by Sepsis-3
Although multiple reports have claimed that bioactive AM and MR-proADM could prognose the severity of sepsis, the progression to septic shock, or sepsis-related mortality [19–28], there has been little mention of the performance of AM for diagnosing sepsis using the latest Sepsis-3 definitions. In this study, we found that among all clinical parameters examined, tAM had the highest diagnostic accuracy for sepsis at ICU admission, and the diagnostic accuracy of mAM for sepsis was also high (Table 3).
Bernal-Morell et al. reported that to diagnose sepsis, the AUCs of MR-proADM, CRP, PCT, and lactate were 0.771 (95% CI: 0.692–0.850), 0.643 (95% CI: 0.547–0.739), 0.695 (95% CI: 0.604–0.786), and 0.483 (95% CI: 0.383–0.583), respectively [45]. Wacker et al. reported that the AUC of PCT for sepsis diagnosis was 0.85 (95% CI: 0.81–0.88, sensitivity: 77%, specificity: 79%) [46]. Kondo et al. reported that the AUC of PCT and PSEP were 0.84 (95% CI: 0.81–0.87, sensitivity: 80%, specificity: 75%) and 0.87 (95% CI: 0.84–0.90, sensitivity: 84%, specificity: 73%), respectively [47]. These reports suggest that PCT and PSEP are high-quality, diagnostic biomarkers of sepsis. The diagnostic performance of PCT in our study was consistent with these previous reports. Additionally, we found that the ability of tAM and mAM to diagnose sepsis was equivalent to that of PCT; however, the diagnostic performance of PSEP was lower than that reported by previous studies. As described previously, the PSEP levels in the non-sepsis group in our study may have been elevated due to renal dysfunction [44].
The diagnostic performances of tAM and mAM for septic shock were highest among the clinical parameters evaluated at ICU admission, and were equivalent to the diagnostic performance of lactate and PCT for septic shock. Chen et al. reported that the AUC for the diagnostic performance of AM for severe sepsis and septic shock defined by Sepsis-2 was 0.847 (95% CI: 0.797–0.898, cutoff value: 41.24 ng/L, sensitivity: 67.6%, and specificity: 90.0%) [25]. The AUC for diagnostic performance of AM for sepsis and septic shock in our study is consistent with previous reports, although we used Sepsis-3 diagnostic criteria and prior studies used Sepsis-1 or Sepsis-2 criteria. These findings suggest that AM levels increase as the severity of sepsis increases, thus, AM may be a suitable diagnostic biomarker for point-of-care testing.
Predictive performance of AM for the prognosis of patients with sepsis
Prognosing the clinical outcome of sepsis patients with MR-proADM and bioactive AM levels has also been described [19–21, 23, 24, 26, 27]. Chen et al. reported that the AUC of AM to predict the risk of in-hospital mortality was 0.773 (95% CI: 0.738–0.808, cutoff value: 34.49 ng/L, sensitivity: 81.6%, specificity: 60.8%) [26]. Marino et al. demonstrated that bio-ADM levels at ICU admission strongly correlated with 28-day mortality. Patients with ADM levels > 70 ng/L had a 28-day survival rate of 55%. For patients in this group whose ADM levels remained above 70 ng/L through Day 4 post-admission, the survival rate was 36%. In contrast, when ADM levels were below 70 pg/mL the survival rate was 100% [24]. Furthermore, Mebazaa et al. reported that in patients with bio-ADM > 70 ng/L on admission, a decrease in bio-ADM levels to below 70 ng/L on Day 2 was associated with recovery of organ function on Day 7 and better 28-day outcomes (9.5% mortality), whereas persistently elevated bio-ADM levels on Day 2 were associated with prolonged organ dysfunction and high 28-day mortality (38.1% mortality) [20]. In our study, tAM and mAM levels on Day 1 could not predict 28-day sepsis-related mortality; however, their levels on Day 3 showed a high predictive value for sepsis-related mortality by ROC analysis. In fact, the AUCs of tAM and mAM were the highest among the biomarkers evaluated. These results suggest that both tAM and mAM levels reflect the severity of organ damage and mortality will increase if organ damage continues.
Previous reports suggested that increasing AM levels correlated worsening prognoses for patients with sepsis [48]. In contrast, AM can have a tissue protective effect in vitro and in vivo. Moreover, AM is highly anticipated as a new therapeutic agent for inflammatory bowel disease [10]. Although some preclinical animal studies have evaluated the effects of AM or AM antibody therapy for sepsis [49], it is still unclear if AM will have beneficial therapeutic effects for patients with sepsis. There is an ongoing, multi-center study evaluating AM-binding antibody therapy for sepsis (AdrenOSS-2) [50]. The results of this study may finally characterize the efficacy of AM therapy for patients with sepsis.
Limitations
Our study has several limitations. The effect of complement factor H on the mAM assay used in our study is not yet known. The sample size was small. The laboratory data, which are obtained in daily routine and used by the practicing clinicians (except for PSEP) were not measured using the same blood sample used for AM measurement. Furthermore, this study was performed as a small, single-center study focusing on the diagnostic performance of AM for sepsis and septic shock based on Sepsis-3 definitions. Thus, future randomized, controlled, multi-center studies should seek to verify our study results.