Background

Vasodilatory shock, a form of life-threatening generalized acute circulatory failure [1, 2], affects many patients in intensive care [3] and is associated with high mortality [4]. Vasodilatory shock has many etiologies, including but not limited to sepsis (the most common cause), inflammatory shock without infection (e.g., pancreatitis), postsurgical vasoplegia, endocrine shock, and spinal shock [5]. The primary goal of the hemodynamic treatment of such patients is to restore adequate mean arterial pressure (MAP) [6] with fluid resuscitation and/or vasopressors [7,8,9]. However, some patients are resistant to vasopressor therapy and require high doses to reach target MAP. This catecholamine-resistant vasodilatory shock (CRVS) is associated with adverse events [10, 11] and high mortality rates [12,13,14], but its pathophysiology is not well understood.

The peptide angiotensin I (ANG I) is an integral part of the renin-angiotensin-aldosterone system, which regulates blood pressure and is converted by the angiotensin-converting enzyme (ACE) to ANG II, making the ANG I/II ratio a marker of ACE function [15, 16]. Low levels of ANG II, a potent vasoconstrictor, are associated with increased mortality in severe sepsis [17], vasodilatory shock [18], and acute respiratory distress syndrome [19], all of which are conditions with endothelial injury, decreased endothelium-bound ACE activity, and decreased capacity to convert ANG I to ANG II [18,19,20]. Thus, the ANG I/II ratio may be elevated in CRVS and predict worse clinical outcomes. These considerations have become increasingly relevant since synthetic human ANG II was approved in the USA and Europe to increase MAP in patients with vasodilatory shock [21].

Accordingly, as part of the randomized, double-blind, phase 3 ATHOS-3 (Angiotensin II for the Treatment of High-Output Shock) trial (ClinicalTrials.gov, NCT02338843), we measured ANG I and II levels of patients with CRVS before initiation of synthetic human ANG II infusion and calculated their ratio. We hypothesized that such patients would have elevated ANG I levels and an increased ANG I/II ratio compared with healthy controls and that a higher ANG I/II ratio would be associated with increased norepinephrine requirements at baseline and with increased mortality.

Methods

Patients

Patients with vasodilatory shock

The ATHOS-3 study protocol, including patient characteristics, has been previously published [22, 23]. In brief, patients with catecholamine-resistant hypotension (defined as those with a total vasopressor dose >0.2 mcg/kg/min for ≥6 h) and high-output shock (defined as central venous oxygen saturation >70% with central venous pressure >8 mmHg or cardiac index >2.3 L/min/m2) were randomized and treated with either ANG II or placebo, plus standard of care. Blood samples were drawn and stored after randomization and prior to administration of study drug. Collected blood was centrifuged (2000 g for 10 min) and stored at –80 °C until shipped for analysis.

Healthy control sera

As part of the ANG I and ANG II assay validation, ANG I and ANG II levels were measured in banked sera donated by healthy volunteers.

ANG I and ANG II assessments

Endogenous serum concentrations of ANG I and ANG II were measured by ultra-performance liquid chromatography with tandem mass spectrometry detection, capable of measuring angiotensin peptide levels as low as 10 pg/mL (inVentiv Health Clinique, Quebec City, Quebec, Canada). Following rapid thawing of the serum, samples were stabilized with a combination of aliskiren, pepstatin A, and o-phenanthroline in acidified dimethyl sulfoxide combined with a mixture of EDTA and 4-(hydroxymercury) benzoic acid in phosphate-buffered saline. All samples were spiked with stable-isotope-labeled internal standards for ANG I and ANG II at a concentration of 50 pg/mL. Following protein precipitation using acetonitrile with 1% formic acid and solid-phase extraction (Oasis MCX; Waters Corporation, Milford, MA, USA) of the supernatant, samples underwent liquid chromatography-tandem mass spectrometry analysis using a reverse-phase analytical column (Acquity CSH C18; Waters Corporation) operating in line with an XEVO TQ-S triple quadrupole mass spectrometer (Waters Corporation) in multiple reaction monitoring. The sum of the signal from three different mass transitions per peptide was measured, and angiotensin concentrations were calculated by relating the ratio of peptide signal to internal standard signal.

Statistical analyses

Analyses of baseline ANG I, ANG II, and ANG I/II ratio and association with survival were prespecified. All other analyses, including comparison to healthy controls, were post hoc. Wilcoxon rank-sum test, Fisher’s exact test for binary outcomes, and chi-square statistic for other categorical outcomes were used for comparisons. Survival from the time of randomization to time of death from any cause was analyzed by the Kaplan–Meier formula. Estimates and confidence intervals were calculated by the product limit method and Greenwood’s formula for the variance and included the difference between treatment arms. For missing data in time-to-event analyses, including mortality at day 28, censored data techniques were utilized. Patients with missing data were censored on the last known survival date up to the specified endpoint (i.e., day 28).

Differences in survival between ANG I/II ratios above and below the median were analyzed by a two-sided log-rank test for mortality to day 28. Multivariate analyses were conducted for mortality to day 28, which included a stratified log-rank test using baseline strata and covariates that were not balanced. To adjust for the impact of multiple comparisons, a P <  0.01 was used to infer statistical significance.

Results

We studied 321 patients with vasodilatory shock. Sera from 24 healthy subjects formed the control group. Baseline ANG I and II levels are summarized in Table 1. In comparison to healthy controls, vasodilatory shock patients had substantially (roughly 6-fold) higher ANG I levels (253 pg/mL [interquartile range (IQR) 72.30–676.00 pg/mL] vs 42 pg/mL [IQR 30.46–87.34 pg/mL]; difference P <  0.0001) and higher ANG I/II ratios (1.63 [IQR 0.98–5.25] vs 0.4 [IQR 0.28–0.64]; difference P <  0.0001). In contrast, ANG II levels were not different between groups (84 pg/mL [IQR 23.85–299.50 pg/mL] vs 97 pg/mL [IQR 35.27–181.01 pg/mL]; difference P = 0.9895). Distribution of baseline ANG I and II levels and ANG I/II ratio for vasodilatory shock patients can be found in Additional file 1: Figures S1–S3 (Table 1).

Table 1 Baseline angiotensin I, angiotensin II, and angiotensin I/II ratio in ATHOS-3 patients and healthy controls
Full size table

Angiotensin I/II ratio

The median ANG I/II ratio across treatment arms at baseline was 1.63 (IQR 0.98–5.25). Patient demographics and disease characteristics by baseline median ANG I/II ratio were largely similar between groups (Table 2). However, recent exposure to ACE inhibitors was significantly more common in patients with a ratio above the median. Moreover, patients with a higher ANG I/II ratio had higher serum albumin concentrations and were receiving a higher dose of vasopressor support (norepinephrine-equivalent dose) at baseline. Baseline ANG I/II ratios were similar between the placebo (n = 139) and ANG II treatment arms (n = 142) (Table 2).

Table 2 Baseline demographics and disease characteristics for patients with vasodilatory shock
Full size table

Survival by baseline ANG I/II ratio

Mortality in the trial’s placebo treatment arm was 64.7% in those with baseline ANG I/II ratio above the median and 45.2% in those with a ratio below the median (Fig. 1). In a multivariate analysis of mortality in the placebo arm, the baseline ANG I/II ratio was a significant predictor of overall mortality (hazard ratio 0.54; P = 0.0111) (Table 3, Fig. 1)

Fig. 1
figure 1

Survival to day 28 by baseline ratio of angiotensin I/II (<1.63 or ≥1.63, the population median). ANG angiotensin, CI confidence interval, Est estimate

Full size image
Table 3 Multivariate analyses of survival in placebo treatment arm
Full size table

Discussion

We measured the plasma concentrations of ANG I and ANG II and calculated their ratio at baseline in patients enrolled in the ATHOS-3 study. We found that, in patients with CRVS, ANG I levels were higher than in healthy controls. We also found that despite much higher ANG I concentrations in the ATHOS-3 patients, ANG II levels were similar to those in healthy controls; this led to increased ANG I/II ratios. These observations suggest that ACE function and the conversion of ANG I to ANG II may be disordered in vasodilatory shock. Moreover, we found that ANG I/II ratios above the median were associated with specific baseline features (i.e., recent use of ACE inhibitor, greater dose of norepinephrine-equivalent administration, and greater severity of illness). Finally, we found that a high ANG I/II ratio predicted increased mortality.

Relationship to previous studies

Previous studies have reported that the baseline ANG I/II ratio averaged 0.38 in otherwise healthy patients with hypertension [15]; this is consistent with the ratio of 0.4 in healthy sera measured. The median ratio value of 1.63 for patients in the present study suggests a possible pathological decrease in conversion of ANG I to ANG II in patients with CRVS. Endothelial injury is common during septic shock. Thus, endothelial membrane–bound ACE activity may be reduced during shock. Logically, reduced ACE activity should lead to decreased ANG I to ANG II conversion and an increased ratio. A significant proportion of ATHOS-3 patients had high ANG I/II ratios, suggesting decreased ACE activity. Low levels of ANG II and ACE activity on day 1 have been previously reported in patients with sepsis and appear associated with a poor prognosis [17]. Decreased ACE activity could be due to an intrinsic defect in ACE function [20] or to small peptides with ACE inhibitory properties [24]. In addition, at least two pro-inflammatory cytokines (tumor necrosis factor-α [TNF-α] and interleukin-1β) downregulate ACE in cultured human endothelial cells [25]. Finally, while not examined in this study, different single-nucleotide polymorphisms of ACE can affect ACE activity and are associated with mortality rates in septic shock [26], possibly through interactions between TNF-α and such polymorphisms [27]. It appears biologically plausible that a high ANG I/II ratio may reflect decreased ACE activity. In keeping with this notion, the recent use of ACE inhibitors was markedly more common in patients with a high ANG I/II ratio in our study.

Another key enzyme, ACE2, can also affect the ANG I/II ratio. ACE2 catalyzes the conversion of ANG II to ANG (1–7) [28], and increased ACE2 activity may also decrease ANG II levels and increase ANG I/II ratios. Therefore, high ACE2 activity may contribute to a high ANG I/II ratio in vasodilatory shock.

Study implications

Our findings suggest that in many patients with CRVS, there is an imbalance between ANG I and ANG II levels. This imbalance may be related to changes in ACE1 and/or ACE2 activity, which may relatively diminish ANG II generation and can be exacerbated by recent ACE inhibitor administration. Moreover, the findings imply that diminished ability to convert ANG I to ANG II may contribute to a catecholamine-resistant vasodilatory state and increase the risk of death. In their aggregate, these findings suggest that there is a biological rationale for the exogenous administration of ANG II in CRVS.

Strengths and limitations

To our knowledge, this is one of the first studies to evaluate serum ANG I and ANG II levels and the ANG I/II ratio in patients with CRVS. Only a single recent pilot study found that increased ANG I levels were correlated with mortality [29]. In comparison, our study was much larger and involved several hundred patients in multiple countries and continents, thus providing a high level of external validity. In addition, this study utilized a double-blind, placebo-controlled, phase 3 registration trial design, assuring that characteristics and outcomes were collected prospectively and were independently monitored; this minimized selection and ascertainment bias. The measurements of ANG I and ANG II were performed by an independent laboratory blinded to clinical characteristics, thus further minimizing bias. Moreover, the analysis of such data followed a prespecified protocol. Finally, the associations observed appear logical and consistent with current knowledge of the physiology and pathophysiology of ANG I, ANG II, and ACE1 and ACE2 activity in inflammatory states.

Our study had limitations. We dichotomized ANG I/II ratios as part of our assessment. Such an approach simplifies comparisons but is insensitive to the continuous nature of biological variables. Thus, the correct specific cutoff point to inform clinical decisions remains unknown. Follow-up was to 28 days only, so implications for longer survival windows could not be made. In addition, ACE activity was not measured directly; rather, ACE activity was inferred from the ratio of ANG I/II in this study. However, patients with prior exposure to ACE inhibitors appeared to be particularly prone to a high baseline ANG I/II ratio, indicating that, in at least some patients, a high baseline ratio very likely resulted from decreased ACE activity. We did not measure the ANG I/II ratio in real time. However, ANG I and II concentrations were collected prospectively as part of a prespecified analysis. We did not measure ACE2 activity as part of the ATHOS-3 study. Thus, our suggestion that increased ACE2 activity may affect the ANG I/II ratio remains speculative. Further studies will require a more detailed assessment of the increasingly complex angiotensin family of molecules and their interaction with ACE1 and ACE2 activity.

Conclusions

In CRVS, both ANG I and the ANG I/II ratio are elevated. High ANG I/II ratios are associated with specific baseline clinical features and predict increased mortality. These observations provide a biological rationale for interventions aimed at correcting such imbalance.