Methylene blue versus hydroxocobalamin for septic shock: a real-world comparative effectiveness study using TriNetX

Abstract

Background

Adjunctive therapies targeting pathologic vasodilation in septic shock are increasingly used when standard vasoactive strategies fail. Although methylene blue (MB) and high-dose hydroxocobalamin (HC) have shown potential hemodynamic benefits, comparative data in sepsis are lacking. We compared outcomes among septic patients treated with MB or HC using a large multicenter electronic health record network.

Methods

We conducted a retrospective cohort study using the TriNetX US network (65 health systems). Adults hospitalized with sepsis, severe sepsis, or septic shock from 2016 to 2025 who received MB or HC were included. Patients with nitroprusside exposure, methemoglobinemia, cardiogenic shock, or neither therapy were excluded. Propensity score matching (1:1) balanced 91 covariates.

Results

A total of 732 matched patients were analyzed (366 MB, 366 HC). Mortality and acute kidney injury rates were similar between groups. HC was associated with higher continuous renal replacement therapy initiation at 3–7 days and 7 days after treatment (both P < 0.001). Stroke and prolonged vasopressor use were more frequent with HC during days 3–7 but not by day 7. Other secondary outcomes were comparable.

Conclusions

MB and HC demonstrated similar early mortality, but MB was associated with lower continuous renal replacement therapy use and fewer early strokes and prolonged vasopressor requirements. These findings suggest MB may be the more favorable adjunctive therapy, although prospective comparative studies are needed to confirm these observational results.

Septic shock is characterized by widespread vasodilation, consequential hypotension, and multiorgan dysfunction. Despite advances in supportive care, it remains a leading cause of mortality worldwide.^1–4 The pathophysiology of septic shock is partially driven by a dysregulated production of vasodilatory gasotransmitters, such as nitric oxide (NO) and hydrogen sulfide (H₂S).^5,6 The standardized Sepsis-6 or Surviving Sepsis Campaign bundles do not directly target these mediators,^4,7–10 which has prompted investigation into adjuvant therapies that may modulate this underlying vasodilatory mechanism.

Methylene blue (MB), which has been well validated for methemoglobinemia and vasoplegic shock, is one such potential adjuvant due to its ability to attenuate NO-mediated vasodilation through guanylate cyclase inhibition.^11–13 Current randomized controlled trials and meta-analyses suggest that it may reduce vasopressor requirements, duration of vasopressor requirements, length of stay in the intensive care unit (ICU), and mortality.^14–19 Yet this evidence is limited by small sample sizes, heterogeneity in study design, and variable endpoints. As a result, the optimal timing, dosing, and patient selection for MB adjuvant therapy remain largely unclear.

Hydroxocobalamin (HC), which has traditionally been used to treat cyanide toxicity, is another such potential adjuvant due to its ability to scavenge both NO and H₂S, two potent vasodilatory gasotransmitters.^6,20–24 However, the literature supporting its use is also limited. Observational studies, case series, and a randomized trial in septic shock have demonstrated that high-dose HC may reduce vasopressor requirements and circulating H₂S levels.^6,20,25 Systematic reviews and meta-analyses in vasoplegic shock, including those with cardiac surgery populations, suggest that while an improvement in hemodynamics exists, the mortality benefit is unclear.^20,22,26 Furthermore, direct comparisons to MB are limited. It is crucial to note that many of these data do not specifically look at septic shock.

Given the mechanistic rationale, preliminary clinical evidence for each agent, and lack of head-to-head comparison studies, there is a vital need to further delineate the real-life clinical effect of MB and HC in patients with septic shock. This study therefore aimed to compare them in adults with septic shock across a range of clinically important outcomes. Specifically, this study represents a comparative effectiveness analysis within a population that clinicians have already determined requires escalation to rescue therapy, rather than a comparison against patients who did not receive salvage therapy.

METHODS

We conducted a retrospective cohort study using the US Network of the TriNetX Analytics Platform, an electronic health record and claims database containing deidentified inpatient and outpatient data from over 117 million patients across 76 American healthcare organizations.²⁷

Study population

Adult patients (age $\ge$18 years) who were hospitalized with septic shock and were administered MB or HC within 1 week of diagnosis between January 1, 2016, and January 1, 2025, were included. These patients were identified using codes from the International Classification of Diseases, 10th Revision (ICD-10). All patients with R65.2 (severe sepsis), A41.9 (sepsis, unspecified organism), R65.21 (severe sepsis with septic shock), or T81.12 (postprocedural septic shock) and required vasopressors within 7 days on or after any of these instances of sepsis were considered to have septic shock. Vasopressors included epinephrine, norepinephrine, phenylephrine, dobutamine, dopamine, and vasopressin. We used RxNorm codes to identify these medications. Patients with R57.0 (cardiogenic shock) were excluded, as initiation of vasopressors could be attributed to an alternative etiology. Similarly, patients with a history of a cardiac procedure were excluded, as (1) they maintain a higher risk of cardiac-related vasoplegia rather than a pure septic etiology^28–32 and (2) the goal was to capture septic shock only and not postprocedure complications. Patients were also excluded if they had another indication for MB or HC therapy, such as hydrogen cyanide toxicity (T53.7), methemoglobinemia, nitroprusside exposure, or B group vitamin deficiency (E53.8 and D51). Patients who did not receive either MB or HC were excluded to avoid treatment-severity bias. Additionally, patients who received both MB and HC were excluded from the primary comparative analysis; this subgroup represents individuals with a greater degree of refractory shock in whom the initial salvage therapy was insufficient, and their inclusion would compromise attribution of outcomes to either individual agent. Supplemental Material, Appendix A, provides additional information on full inclusion criteria, exclusion criteria, and ICD and RxNorm codes used.

Outcomes

Primary outcomes of interest included all-cause mortality at two time points: between 3 and 7 days posttherapy and at 7 days posttherapy. Secondary outcomes included acute kidney injury (AKI) (N17.9, N17), continuous renal replacement therapy (CRRT) initiation (90945, 90947), atrial fibrillation, ventricular tachycardia, ventilator/hospital-acquired pneumonia, stroke, vasoplegia requiring midodrine, and vasopressor use after 3 days. We selected the 3- to 7-day window to monitor for secondary outcomes (end-organ dysfunction) since MB and HC are known free radical scavengers but require time to take effect. We selected 7 days posttherapy to assess for all-cause mortality since they are salvage therapies. A 0- to 72-hour analysis was also performed to provide a baseline reference on patient status prior to the salvage effects of MB and HC. Patients were excluded from the analysis if they experienced an outcome prior to the specified time window since we were unable to verify the exact timing of the last outcome occurrence. Arrhythmias were included as an outcome of interest since they are known complications of sepsis, secondary to physiological stress, electrolyte abnormalities, and the like. Supplemental Material, Appendix B, provides the full list of ICD-10 and Current Procedural Terminology (CPT) codes used.

Statistical analysis

Direct propensity score matching using 91 covariates across demographics, comorbidities, procedures, medications, and laboratory values was applied. Supplemental Material, Appendix C, provides more information on all covariates. TriNetX was used for analysis. Hypothesis testing was two-sided, and alpha was set at 0.05.

RESULTS

After matching, 732 patients across 65 US healthcare organizations were included. The MB and HC groups each had 366 patients. Postmatch patient characteristics are summarized in Table 1. Appendix C lists describes prematch patient characteristics.

Table 1.

Basic demographic characteristics of included patients after propensity score matching

Variable	Methylene blue (MB)	Hydroxocobalamin (HC)	P value
Current age (years)	58.8 ± 17.4	60.0 ± 16.2	0.31
Age at index (years)	55.1 ± 17.5	56.3 ± 16.5	0.33
White	68.3%	67.2%	0.75
Black	16.1%	17.5%	0.62
Male	62.0%	63.1%	0.76

It is important to note that for several outcomes, <366 patients were included since the outcome was recorded prior to the time window of interest. This included CRRT initiation, atrial fibrillation, ventricular tachycardia, ventilator or hospital-acquired pneumonia, and stroke at both time points. Table 2 notes the number of excluded patients per outcome at each time point.

Table 2.

Excluded patients per outcome of interest^*

Outcome	3 to 7 days posttherapy		At 7 days posttherapy
	MB	HC	MB	HC
Mortality	0	0	0	0
AKI	0	0	0	0
CRRT	62	81	11	10
AFib	132	123	80	79
VTach	54	49	39	36
VAP/HAP	136	146	91	96
Stroke	39	30	29	19
Midodrine	0	0	0	0
3-day pres	0	0	0	0

^*Patients were excluded if they experienced the outcome of interest prior to the time window of interest. 3-day pres indicates vasopressor requirement at 3 days posttherapy; AFib, atrial fibrillation; AKI, acute kidney injury; CRRT, continuous renal replacement therapy initiation; HC, hydroxocobalamin; MB, methylene blue; midodrine, vasoplegia requiring midodrine; VAP/HAP, ventilator-acquired pneumonia/hospital-acquired pneumonia; VTach, ventricular tachycardia.

Once the patient exclusions as described in Table 2 were applied, primary and secondary outcomes were analyzed. Mortality did not differ between MB and HC at 3 to 7 days (9.8% vs 12.6%, P = 0.24) or at 7 days (29.5% vs 35.0%, P = 0.11). Of note, CRRT initiation was higher after HC at both time points (3–7 days: 10.5% vs 3.3%, P < 0.001; 7 days: 29.0% vs 16.9%, P < 0.001). Stroke was more frequent with HC in the 3- to-7-day window (3.0% vs 0.0%, P = 0.002) but not at 7 days. Ongoing vasopressor use beyond 72 hours was higher with HC in the 3- to 7-day window (51.1% vs 39.1%, P = 0.001) but not at 7 days. Other secondary outcomes were similar. All proportions and hypothesis tests are summarized in Table 3. Importantly, our 0- to 72-hour analysis was statistically insignificant.

Table 3.

Proportion of patients with primary and secondary outcomes of interest^*

	3 to 7 days posttherapy					At 7 days posttherapy
	MB	HC	P value	HR	P value	MB	HC	P value	HR	P value
Primary outcome
Mortality	9.8%	12.6%	0.24	1.367	0.57	29.5%	35.0%	0.11	1.222	0.58
Secondary outcomes
AKI	20.5%	19.9%	0.85	1.008	0.34	66.1%	61.2%	0.17	0.912	0.46
CRRTa	3.3%	10.5%	<0.01b	3.756	0.30	16.9%	29.0%	<0.01b	1.804	0.01b
AFiba	4.3%	4.1%	0.93	N/A	N/A	18.9%	15.7%	0.31	N/A	N/A
VTacha	3.2%	3.2%	0.97	N/A	N/A	4.9%	4.2%	0.69	0.870	0.77
VAP/HAPa	4.3%	4.5%	0.92	N/A	N/A	17.8%	19.6%	0.59	N/A	N/A
Strokea	0.0%	3.0%	0.002b	N/A	N/A	3.0%	3.7%	0.57	N/A	N/A
Midodrine	4.6%	6.3%	0.33	1.393	0.32	9.3%	11.2%	0.39	1.236	0.84
3-day pres	39.1%	51.1%	0.001b	1.523	0.59	93.7%	96.4%	0.09	1.032	0.04b

^*The proportions were calculated after applying the patient exclusions described in Table 2. Some hazard ratios could not be calculated due to small sample sizes after patient exclusions, denoted by “N/A.” 3-day pres indicates vasopressor requirement at 3 days posttherapy; AFib, atrial fibrillation; AKI, acute kidney injury; CRRT, continuous renal replacement therapy initiation; HC, hydroxocobalamin; HR, hazard ratio; MB, methylene blue; midodrine, vasoplegia requiring midodrine; VAP/HAP, ventilator-acquired pneumonia/hospital-acquired pneumonia; VTach, ventricular tachycardia.

^a An outcome with patients excluded from either one or both time cohorts since they experienced the outcome prior to the time window.

^b An outcome with a statistically significant difference between exposure groups under a specified time window.

Kaplan-Meier analyses of each outcome were respectively concordant. Supplemental Material, Appendix D, provides additional information.

DISCUSSION

In this study, we examined retrospective patient data across 65 US healthcare organizations using the TriNetX platform to determine real-world clinical effects of MB and HC when used as salvage therapy in septic shock. There was no statistically significant difference in all-cause mortality between MB and HC at either 3 to 7 days or at 7 days posttherapy; this finding reflects the absence of a between-group difference and does not imply lack of benefit compared with untreated patients. Similarly, there was no statistically significant difference in the rates of vasoplegia requiring midodrine at either time point. A proportional difference in vasopressor requirements was noted at the 3- to 7-day posttherapy window that disappeared at 7 days, but the hazard ratio difference became statistically significant. This may be explained by differences in timing effects that do not change the overall 7-day risk and may suggest that MB has a statistically significant but delayed effect. All in all, these findings corroborate the existing literature, which suggests that despite these agents’ abilities to acutely reduce vasopressor needs, sustained vasoactive effects and definitive survival benefits are not well established.^{14–16,22,26}

There were no notable differences in the rates of atrial fibrillation or ventricular tachycardia within the specified posttherapeutic time windows. Furthermore, a large proportion of patients with either arrhythmia were excluded since they were documented with the outcome prior to the posttherapeutic windows. The findings for ventilator-acquired pneumonia and hospital-acquired pneumonia were similar. These data suggest that neither agent increased arrhythmogenic risk nor impacted nosocomial infection rates, which further aligns with the current literature.^14–16,22

With regards to renal effects, an interesting picture emerges. There was no statistically significant difference in AKI rates at either time window, which agrees with the currently available and limited renal safety data.²² However, MB was associated with a statistically significantly lower rate of CRRT initiation when compared to HC at both time points, a novel finding that has not yet been clearly established.^22,26 It is important to note that the hazard ratio became statistically significant only at 7 days. The mechanisms behind this result are likely multifactorial. MB and HC improve vasoplegia through different pharmacodynamic mechanisms and remain statistically similar in the long term. Our results showed that MB was associated with a decreased vasopressor requirement at 3 days posttherapy, but the hazard ratio remained statistically insignificant, suggesting that group differences appear to reflect cumulative rather than sustained effects. In addition to vasoactive effects, MB contributes well-established antiinflammatory effects by modulating proinflammatory cytokines and upregulating antiinflammatory cytokines.^19,33 The CRRT hazard ratio becoming statistically significant at 7 days may reflect this, as biological differences in modulating inflammation are time dependent. With its vasoactive and antiinflammatory effects combined, MB may exert renoprotective effects. Indeed, it has been associated with a decreased rate of renal failure and improved markers of renal function versus standard care,^17,34 but our study did not explore renal outcomes beyond 7 days to monitor for this effect’s sustenance.

Our findings indicate that stroke risk was decreased with MB in the short term, but this became statistically insignificant after 7 days. This is not consistent with the existing literature, which has not demonstrated such an effect^22,26 and may represent a spurious association due to our small sample size. However, this effect may also be partially due to the fact that the 7-day window included strokes from days 0 to 7 vs just 3 to 7 days. A decrease in stroke rate from 3 to 7 days is biologically plausible given that MB and HC exhibit protective effects against free radicals and therefore protection from stroke onset by hypoperfusion exacerbating existing plaques. Nevertheless, this warrants further investigation.

This study has limitations. First, we utilized retrospective data that rely on ICD-10 coding. Consequently there is an inherent risk of misclassification and subsequent biases of exposure misclassification and outcome ascertainment. Causal inference is also limited by observational design. Second, we had a small sample size. This was especially the case once patients who experienced the outcomes of interest outside of the specified time window were excluded. Nevertheless, our primary and secondary outcomes largely align with the existing literature. Third, we were unable to verify the dose or timing of vasopressors, as this information was not included in the database. Fourth, we analyzed outcomes such as CRRT and stroke, which may reflect a sicker patient population. However, our propensity score match included past medical conditions, laboratory values, and existing medications prior to analysis to minimize their confounding effects. Fifth, there was no comparison between these therapies against a placebo. This was not possible as a database study since both MB and HC are typically deployed as salvage therapies. Including a patient population that received neither increases the risk of bias through confounding by indication, as these patients likely experienced a decreased severity of shock. Similarly, patients who received both agents were not analyzed, as they likely experienced an increased severity of shock.

MB may offer practical advantages over HC as an adjunctive treatment in septic shock, particularly with renal replacement therapy initiation and stroke. Although the existing literature has more fully established the renoprotective effects of MB, there is currently no known association with improved rates of renal replacement therapy. Additionally, the existing literature has not demonstrated a decreased stroke risk with either agent. Prospective trials are therefore warranted to explore their effects further and fully on renal replacement therapy initiation and stroke risk.

Disclosure statement/Funding

The authors report no funding or conflicts of interest.