Nitrous Oxide for the Treatment of Complex Regional Pain Syndrome: A Randomized Blinded Trial

Abstract

Introduction:

Complex Regional Pain Syndrome (CRPS) is a debilitating neuropathic condition often refractory to conventional treatments. N-methyl-D-aspartate (NMDA) receptor antagonists has a well-established role in the development and modulation of chronic neuropathic pain. Nitrous oxide is widely-used and generally safe anesthetic gas with NMDA receptor antagonist activity. We therefore tested the hypothesis that brief periods of nitrous oxide exposure reduce pain in patients with CRPS.

Methods:

Patients with diagnosis of CRPS were randomized to either 2 hours nitrous oxide exposure on three alternating days (Nitrous Oxide) vs. a placebo air/oxygen mixture (Air-Oxygen). Our primary outcome was patient-reported pain scores at 1-week and 1-month. Secondary and exploratory outcomes were physical and mental health (PRMOIS-29 v2 survey), specific neuropathic pain symptoms (McGill short form questionnaire), and opioid consumption.

Results:

44 patients participated in the study; 20 randomized to Nitrous Oxide and 24 assigned to Air-Oxygen. Pain scores did not differ significantly, with estimated difference in means (Nitrous Oxide – Air-Oxygen) of −0.57 (95% CI: −1.42, 0.28) points, P = 0.19. There were also no differences detected in secondary outcomes, with estimated difference in mean Z-scores for physical health (Nitrous Oxide – Air-Oxygen) of 0.13 (95% CI: −0.16, 0.43), mental health 0.087 (95% CI: −0.31, 0.48), and Patient Global Impression of Change score −0.7 (95% CI: −1.85, 0.46).

Conclusions:

Compared to air/oxygen, 2 hours of nitrous oxide/oxygen exposure for three sessions did not provide meaningful therapeutic potential for patients with chronic CRPS. Our results do not support using nitrous oxide for treatment of CRPS.

Introduction

Complex Regional Pain Syndrome (CRPS) is a debilitating neuropathic pain condition that is characterized by out-of-proportion and long-lasting severe pain and reduced range-of-motion in an affected limb. The condition profoundly impairs life quality.¹ Patients may lose functional use of the affected extremity and may also be unable to work or properly care for themselves which worsens the disease’s already substantial socioeconomic consequences.²

Treatment of CRPS is multimodal and includes various classes of analgesics, physical therapy, nerve blocks, peripheral and central neural stimulation, and intrathecal drugs.³ Over-activation of NMDA receptor signaling is believed to contribute to hypersensitivity within the central nervous system and play a fundamental role in the pathophysiology of CRPS.^{4; 5} Consequently, patients who are refractory to many treatments are sometimes given subanesthetic ketamine infusions, typically administered in sessions over a period of days.⁶ Ketamine is a well-known N-methyl-D-aspartate (NMDA) receptor antagonist, and this pharmacologic mechanism is believed to underpin its efficacy in CRPS. Although widely used, there are few double-blind placebo-controlled trials of ketamine for CRPS published. Available studies suggest significant pain relief but limited functional improvement. According to a recent meta-analysis, ketamine infusions may provide more than 30% pain reduction in 58% of patients with CRPS during the 1–3 month follow-up period.⁶

Ketamine infusions are costly, require intravenous access, and necessitate continued patient monitoring. Ketamine also may cause various side effects, including hallucinations, feelings of inebriation, agitation, confusion, nausea/vomiting, headaches, hypertension, cardiac arrhythmias, and hepatobiliary dysfunction.⁷ Though potentially beneficial for certain patients, ketamine is not an ideal medication for CRPS.

Aside from ketamine, there are few commonly used NMDA receptor antagonists, one of which is nitrous oxide. Nitrous oxide is a low potency inhaled anesthetic used for both general anesthesia and sedation. Unlike ketamine, nitrous oxide causes sparse side effects. Furthermore, ventilation and oxygenation are unaffected, and the gas is rapidly eliminated via the lungs, making nitrous oxide relatively easy to administer.

Various studies suggest that intraoperative nitrous oxide use may reduce opioid requirements and the development of chronic post-surgical pain.^8–10 Nitrous oxide also inhibits opioid-induced hyperalgesia in humans with acute surgical pain.¹¹ Animal studies report that nitrous oxide alleviates hypersensitivity in rats with neuropathic pain. Furthermore, the reduction in pain hypersensitivity provided by nitrous oxide treatment demonstrates long-lasting relief from neuropathic pain, surpassing the efficacy of other commonly prescribed medications like gabapentinoids or NMDAR antagonists.¹² Another neuropathic pain rat model study showed that a single exposure to nitrous oxide reduced pain hypersensitivity in an injured extremity, abolished pain hypersensitivity in the contralateral extremity, and completely prevented thermal allodynia — effects that persisted for more than a month.¹³ Furthermore, nitrous oxide can also activate TWIK-related potassium channels and blocks calcium channels in the ascending pathway, reducing neuronal excitability. It also attenuates the activation of glutamatergic receptors and may block sodium channels. In the descending pathway, nitrous oxide promotes the release of endogenous opioid ligands and stimulates the release of norepinephrine. Additionally, nitrous oxide may cause epigenetic changes by inhibiting methionine synthase, an important enzyme for DNA and RNA methylation. These effects could explain why N2O, as a short-acting analgesic, exhibits long-lasting anti-pain sensitization effects in animal models of chronic pain. ¹⁴

Given that nitrous oxide is an NMDA antagonist plus effect on other receptors and data from pain studies in animals, our goal was to determine whether nitrous oxide reduces pain and improves CRPS symptoms. Specifically, we tested the primary hypothesis that 2 hours of nitrous oxide exposure for three sessions in CRPS patients decreases pain scores 1 week and 1 month after the last inhalation session. Secondarily, we hypothesized that 2 hours of nitrous oxide exposure over three sessions improves PROMIS-29 v2 and Global Impression of Change scores 1 week and 1 month after the last inhalation session.

Methods

This double-blind randomized trial was conducted at Cleveland Clinic Main Campus. The study was approved by the Cleveland Clinic Institutional Review Board (code number: 19–150, approval date: March 8, 2019), and all patients provided written informed consent. The trial was registered at ClinicalTrials.gov on March 2019, before the first patient was enrolled (NCT03879538). Patients were recruited from July 2019 to July 2022 from the Cleveland Clinic Pain Management Center, including satellite locations.

Subject selection

We include patients aged 18–65 years, previously diagnosed with CRPS (Type I or Type II) based on the revised International Association for the Study of Pain criteria and with disease duration of at least 6 months. Patients were excluded if they had significant cardiopulmonary disease, required high doses of opioids for other chronic pain conditions, or reported no benefit from a previous ketamine infusion. Patients with spinal cord stimulators were allowed to participate. The full list of exclusion criteria is presented in Appendix 1.

Protocol

Participants were randomized 1:1 via a web-based randomization system with stratification by spinal cord stimulator status. The web system was activated after written consent had been obtained and shortly before each procedure, thus concealing allocation as long as practical.

Participants randomized to nitrous oxide study group were given 50% nitrous oxide mixed with 50% oxygen (Nitrous Oxide). Control patients were given 50% air mixed with 50% oxygen mixture (Air-Oxygen). Both gases were given through a mask breathing circuit (MXR Flowmeter, Porter, Hatfield, PA) that was approved by the FDA. Patients in each group inhaled the assigned gas over three 2-hour-long sessions over the course of one week (Monday, Wednesday, and Friday).

Gases were administered by an anesthesiologist who was aware of the assigned gas. However, patients and all other investigators were blinded to treatment. Because nitrous oxide can cause sedation or an intoxicated feeling, participants in each group were given 2 mg of intravenous midazolam to maintain blinding.

At the end of each inhalation session, trial gases were discontinued and patients returned to breathing room air. Patients were monitored for an additional 30 minutes before being discharged to home.

Measurements

Demographic and morphometric characteristics were recorded, as were duration, location, and severity of CRPS.

Before the initial treatment, patients completed PROMIS-29 v2.0, SF-MPQ-2 (6 neuropathic pain items) and previous 7-day opioid usage surveys in person. After participation in three breathing treatments over 1 week, patients were contacted via telephone appointment 1 week and 1 month after the last treatment and asked to again complete the same three surveys, along with a survey on patient’s global impression of change (PGIC).

Primary and Secondary Outcomes

Our primary outcome was pain (scales ranging from 0 – “no pain” to 10 – “worst pain ever”) based on the PROMIS-29 v2.0 questionnaire. Secondary outcomes were 1) physical and mental health Z-scores based on PROMIS-29 v2.0 profile data, and 2) patient disease perception as measured by the PGIC survey (a seven-point scale ranging from 1 – “no change or condition has gotten worse” to 7 – “very much improve”). Exploratory outcomes were: 1) average daily opioid consumption in IV morphine equivalents (mg) based on the survey on previous 7-day opioid use, and 2) the 6 neuropathic symptom items included in the short form McGill pain questionnaire 2 (SF-MPQ-2, with each item ranging from 0 –”no pain” to 10 –”worst pain).

During each treatment session, vital signs were recorded at 5-minute intervals whereas oxygen saturation was monitored continuously.

Statistical methods

We assessed balance of randomized groups on baseline and procedural characteristics using absolute standardized difference (ASD), defined as the absolute difference in means, mean ranks, or proportions divided by the pooled standard deviation. Baseline variables with $\text{ASD}>0.59$ (i.e. $1.96\times \sqrt{\frac{1}{n_1}+\frac{1}{n_2}}$, where $n_1$, $n_2$ are sample size in 2 groups) were considered imbalanced.¹⁵

Analyses were modified intent-to-treat and thus included all randomized patients who received some amount of study intervention. The statistical software SAS 9.4 (SAS Institute, Cary, North Carolina) was used for all analyses.

We assessed the effect of Nitrous Oxide versus Air-Oxygen on pain scores (primary outcome) at 1-week and 1-month follow-up time points using a linear regression model with an unstructured within-subject correlation structure to accounting for repeated measurements. The difference in means of pain score between 2 groups was estimated by the coefficient of treatment group variable and 95% confidence interval, with a significance level of 0.05.

The interaction between treatment group and follow-up time points was tested using a statistical significance criterion of P < 0.15. A statistically significant interaction would suggest that the treatment effect varies over time, in which case we would estimate the treatment effect separately at each follow-up time point. We similarly evaluated treatment-by-baseline pain score interaction using a statistical significance criterion of P < 0.15.

Our secondary outcome #1 was physical and mental health Z-scores from the PROMIS-29 survey. As intermediate scores, T-scores for each of the 7 domains used in the survey need to be estimated from raw data in order to obtain final summary Z-scores measuring physical and mental health. T-scores were obtained from an on-line service provided by the HealthMeasures Scoring Service (https://www.assessmentcenter.net/ac_scoringservice).

We again used a repeated measurement linear regression model with an unstructured within-subject correlation structure to modeling physical health Z-score measured at 1-week and 1-month follow-up time points while adjusting for participants’ baseline physical health scores. Mental health scores were similarly analyzed. A repeated measurement linear regression model was used to model our secondary outcome #2 - PGIC scores. The significance level of 0.05 was used in analyzing each of these secondary outcomes.

Only summary statistics by treatment group for our exploratory outcomes are reported without statistical testing.

Based on the rule of modified intent-to-treat analysis, missing values in pain score at 1-week or 1-month follow-up time point were imputed by the multiple imputation procedure. We generated 10 imputed datasets using the fully conditional specification method and reported the combined analysis results from the 10 models for our primary outcome.

Sample size consideration

Sample size was based on being able to detect a change of 2 units or more in mean pain scores (the primary outcome) using analysis of covariance on the outcomes at the 1-week follow-up time point, as we planned to analyze the primary outcome on 1-week and 1-month follow-up separately if there was a significant treatment-by-follow-up time interaction. Using an estimated standard deviation (SD) of 2 units for pain scores (0–10) with standard type I and type II error rates (alpha = 0.025 due to Bonferroni correction for multiple testing, beta = 0.20), we calculated that 21 patients per group (total sample size of 42 patients) would be needed.

Results

Of the 499 potential participants who were screened for eligibility, 442 were excluded for not meeting screening criteria. From the remaining patients, 13 declined their participation at the screening phone interview. The final sample, consisting of 44 patients, 20 were randomized to Nitrous Oxide and 24 to Air-Oxygen, Figure 1. Potentially confounding baseline variables were presented in Table 1 and were considered balanced as all absolute standardized differences were less than the threshold value of 0.59.

Figure 1.

CONSORT flow diagram describing enrollment and follow-up for the Nitrous Oxide trial.

Table 1.

Patient characteristics by treatment group (N=44)

Factor#	Nitrous Oxide1 (N=20)	Air-Oxygen2 (N=24)
	Statistics^	Statistics^	ASD*
Age (years)	44.4 ± 12.3	46.3 ± 13.4	0.147
Female	16 (80.0)	22 (91.7)	0.339
Spinal cord stimulators used	5 (25.0)	7 (29.2)	0.094
Race			0.535
White/Caucasian	20 (100.0)	21 (87.5)
Black/African American	0 (0.0)	1 (4.2)
Hispanic/Latino	0 (0.0)	1 (4.2)
Multicultural	0 (0.0)	1 (4.2)
Duration of disease (months)	53.5 [36.5, 126.0]	108.0 [28.5, 122.5]	0.099
Extremity affected			0.426
Single location	8 (40.0)	5 (20.8)
Multiple locations	12 (60.0)	19 (79.2)
Pain score at baseline	7.3 ± 1.7	6.8 ± 1.5	0.293

^#No missing value for any factor.

¹50% nitrous oxide mixed with 50% oxygen (Treat).

²100% Oxygen (Control).

^{^}Statistics presented as Mean ± SD, Median [P25, P75], or N (%).

^*Absolute standardized difference (ASD), baseline variable with ASD > 0.59 were considered imbalanced.

Primary analysis

Means ± SDs of pain scores are reported by study group at baseline and 2 follow-up times in Tables 1 and 2, Figure 2. There was no evidence that treatment effect varied over time, with P-values range from 0.40 to 0.94 for the treatment-by-follow-up time interaction based on 10 imputed data sets. When collapsing over time, we did not find a statistically significant difference between the study groups on mean pain score at the 1-week and 1-month follow-ups, with overall difference in means (Nitrous Oxide – Air-Oxygen) of −0.57 (95% CI: −1.42, 0.28; P = 0.19) resulted from the combined analyses results based on multiple imputations, Table 2. The futility boundary was crossed in our final analysis, Figure 3.

Table 2.

Effect of Nitrous Oxide (Treat) versus Air-Oxygen (Control) on the primary outcome

	Nitrous Oxide (N=20)			Air-Oxygen (N=24)			Difference in means* (95% CI)
Primary Outcome	n	n missing	Mean ± SD	n	n missing	Mean ± SD	Nitrous Oxide–Air-Oxygen	P-value§
Pain scores (PS)║							−0.57 (−1.42, 0.28)	0.19
1-week follow-up	19	1	5.9 ± 1.5	23	1	6.0 ± 2.5
1-month follow-up	17	3	6.5 ± 2.0	21	3	6.8 ± 2.2
Patients with low baseline PS (baseline PS < 7.0)							0.48 (−1.09, 2.05)	0.55
1-week follow-up	5	0	5.0 ± 1.0	9	1	3.9 ± 2.0
1-month follow-up	5	0	4.6 ± 1.3	9	1	4.9 ± 2.0
Patients with high baseline PS (baseline PS ≥ 7.0)							−1.00 (−1.91, −0.09)	0.03
1-week follow-up	14	1	6.3 ± 1.6	14	0	7.4 ± 1.7
1-month follow-up	12	3	7.3 ± 1.6	12	2	8.3 ± 0.8

^*Effect of Nitrous Oxide versus Air-Oxygen on pain scores was estimated from a repeated measurement linear regression model with an unstructured correlation structure. This model adjusted for baseline pain score value. CI, confidence interval.

^§P-value was obtained from a 2-sided t-test using a test statistic defined as, $T=\frac{\widehat{{\beta }_1}}{{\text{SE}}_{{\beta }_1}}$, where $\widehat{{\beta }_1}$ is the model-based estimate of treatment effect on the primary outcome, ${\text{SE}}_{{\beta }_1}$ is the standard error of the treatment effect.

^║Based on multiple imputation procedure (10 imputed data sets), P-values for treatment-by-follow-up time interaction ranged from 0.40 to 0.94, suggesting no evidence of treatment effect heterogeneity over time; and P-values for treatment-by-baseline pain score interaction ranged from 0.012 to 0.097, indicating evidence of varying treatment effect among different baseline pain scores. To facilitate assessing this interaction effect, we dichotomized patients’ baseline pain score into “low” and “high” levels using the median value of baseline pain scores (7.0), i.e., baseline pain score was classified as “low” if it was less than 7.0; otherwise, it was deemed as “high”. When collapsing over time, we reported difference between the treatment groups on mean pain scores for “low” and “high” baseline pain score levels, separately.

Figure 2.

Box plots of pain scores in Nitrous Oxide versus Air-Oxygen group at baseline, 1-week and 1-month follow-up time points based on raw data, where box plots displaying the minimum, 25^th percentile, mean (in diamond symbol), median, 75^th percentile, and maximum values of pain score with circles representing possible outliers.

Figure 3.

Treatment effect on pain score (Nitrous Oxide – Air-Oxygen) based on the final analysis and previous interim analyses without treatment-by-baseline pain score interaction considered, plotted with pre-determined efficacy and futility boundaries for pain scores. The horizontal axis denoted study enrollment, while the vertical axis denoted the standardized effect size (Z), defined as the difference in means of pain score collapsed over time divided by the standard error of the difference. Results for the interim analysis (n=28), updated interim analysis (n=36), and final analysis (n=44) are marked by (+), with Z-statistics of −0.24 (n=28), −0.15 (n = 36), and −1.32 (n=44). The futility boundary was crossed at all analyses. Group sequential design was based on 90% power for the study with an alpha of 0.05.

Using our final data without multiple imputations on missing pain score values, we again did not find a statistically significant difference between the study groups on mean pain score at the 1-week and 1-month follow-ups, with overall difference in means (Nitrous Oxide – Air-Oxygen) of −0.52 (95% CI: −1.40, 0.35; P = 0.24) when collapsing over time. It was consistent with our result based on multiple imputation procedure.

A treatment-by-baseline pain score interaction effect remained possible (interaction P-values range from 0.012 to 0.097), indicating that treatment effect might depend on baseline pain. To facilitate assessing the treatment-by-baseline pain score interaction effect, we dichotomized patients’ baseline pain score into “low” and “high” levels using the median value of baseline pain scores (7.0). (There were no missing values for patients’ baseline pain scores.) Among patients with low baseline pain scores, mean ± SD scores were 5.0 ± 0.0 in Nitrous Oxide patients versus 5.3 ± 0.8 in those assigned to Air-Oxygen group; while among patients with high baseline pain scores, the mean ± SD was 8.0 ± 1.1 in Nitrous Oxide patients versus 7.9 ± 0.7 for Air-Oxygen patients.

The interaction between treatment and dichotomous baseline pain scores status (“low” or “high”) was non-significant based on the pre-specified criterion for claiming significant interaction effect (P < 0.15), with median [p25, p75] interaction P-values of 0.22 [0.17, 0.32] across the 10 imputed datasets. After collapsing over time, we did not find a statistically significant difference between the study groups on mean pain scores for patients with low baseline pain scores, but found a statistically significant result for patients with high baseline pain scores. Specifically, the difference in means of pain score (Nitrous Oxide – Air-Oxygen) was 0.48 (95% CI: −1.09, 2.05; P=0.55) for low baseline scores and −1.00 (95% CI: −1.91, −0.09; P=0.03) for high baseline scores, Table 2.

Secondary analyses

Mean ± SD of baseline summary Z-scores (Nitrous Oxide vs. Air-Oxygen) were (1) physical health: −1.7 ± 0.60 vs. −1.6 ± 0.51; and (2) mental health: −1.3 ± 0.72 vs. −1.2 ± 0.67. There was no evidence of treatment effect heterogeneity over time in modeling physical health Z-score, with P = 0.32 for the treatment-by-follow-up time interaction effect; similar result held in modeling mental health Z-score, with treatment-by-follow-up time interaction P = 0.43. There was also no evidence of treatment effect heterogeneity across baseline physical or mental health score levels in modeling corresponding health scores, as the treatment-by-baseline Z-score interaction P-value was 0.45 for physical health and 0.57 for mental health.

When collapsing over time, we did not find a statistically significant difference between the treatment groups on mean physical nor mental health Z-scores, with differences in means (Nitrous Oxide – Air-Oxygen) being 0.13 (95% CI: −0.16, 0.43); P = 0.36 for physical health and 0.087 (95% CI: −0.31, 0.48); P = 0.66 for mental health, Table 3.

Table 3.

Effect of Nitrous Oxide (Treat) versus Air-Oxygen (Control) on the secondary outcomes

	Nitrous Oxide (N=20)			Air-Oxygen (N=24)			Difference in means5 (95% CI)
Secondary outcomes	n1	n2 missing	Mean ± SD	n	n missing	Mean ± SD	(Nitrous Oxide – Air-Oxygen)	P-value6
PROMIS-29 survey
Physical health score3							0.13 (−0.16, 0.43)	0.36
One-week follow-up	19	1	−1.4 ± 0.71	23	1	−1.4 ± 0.78
One-month follow-up	17	3	−1.4 ± 0.58	21	3	−1.5 ± 0.68
Mental health score4							0.087 (−0.31, 0.48)	0.66
One-week follow-up	19	1	−1.01 ± 0.81	23	1	−0.91 ± 1.00
One-month follow-up	17	3	−1.2 ± 0.88	21	3	−1.1 ± 0.91
PGIC scale 7							−0.7 (−1.85, 0.46)	0.23
One-week follow-up	19	1	2 [1, 5]	23	1	4 [1, 5]
One-month follow-up	17	3	2 [1, 4]	21	3	2 [1, 5]

¹No. of available data.

²Missing values were not imputed.

³Treatment-by-follow-up time interaction P = 0.32 and treatment-by-baseline physical health score interaction P = 0.45, suggesting no evidence of treatment effect heterogeneity over time or among different baseline physical health score levels.

⁴Treatment-by-follow-up time interaction P = 0.43 and treatment-by-baseline mental health score interaction P = 0.57, suggesting no evidence of treatment effect heterogeneity over time or among different baseline mental health score levels.

⁵Effects of treatment A versus B on secondary outcomes were estimated from a repeated measurement linear regression model with an unstructured correlation structure. This model adjusted for the corresponding baseline score in modeling physical or mental health score at follow-up times.

⁶P-value was obtained from a 2-sided t-test using a test statistics defined as $T=\frac{\widehat{{\beta }_1}}{{\text{SE}}_{{\beta }_1}}$, where $\widehat{{\beta }_1}$ is the model-based estimate of treatment effect on the primary outcome, ${\text{SE}}_{{\beta }_1}$ is the standard error of the treatment effect.

⁷Median [P25, P75] was reported. Treatment-by-follow-up time interaction P = 0.68, indicating no evidence of treatment effect heterogeneity over time. In addition, P-values based on Kruskal-Wallis test on PGIC scales at one-week follow-up are 0.27 and 0.54 at one-month follow-up.

For the secondary outcome #2, there was no evidence of treatment effect heterogeneity over time with P = 0.68 for the treatment-by-follow-up time interaction in modeling PGIC scales. When collapsing over time, there was no statistically significant difference between the groups on mean PGIC scales, with difference in means (Nitrous Oxide – Air-Oxygen) of −0.7 (95% CI: −1.85, 0.46); P-value = 0.23, Table 3.

Descriptive analyses for exploratory outcomes

At least half of the patients in each study group did not use any opioid during the previous 7-day at baseline, 1-week, and 1-month follow-up time points. The 75^th percentile of average daily opioid consumption (mg IV morphine equivalent) in the Nitrous Oxide and Air-Oxygen group were 10.6 vs. 3.0 at baseline, 11.3 vs. 0 at 1-week post-treatment, and 10.0 vs. 2.5 after 1-month, Table 4.

Table 4.

Summary statistics for daily opioid use and six neuropathic pain items¹

Exploratory outcomes	Median [p25, p75]
	Nitrous Oxide	Air-Oxygen
Daily opioid use 2
Baseline	0 [0, 10.6]	0 [0, 3.0]
One-week follow-up	0 [0, 11.3]	0 [0, 0]
One-month follow-up	0 [0, 10.0]	0 [0, 2.5]
Neuropathic pain 3
Hot-burning pain
Baseline	7.5 [5.5, 10.0]	7.5 [5.0, 8.0]
One-week follow-up	6.0 [4.0, 8.0]	7.0 [4.0, 7.0]
One-month follow-up	6.0 [5.0, 9.0]	7.0 [5.0, 8.0]
Cold-freezing pain
Baseline	6.0 [3.0, 9.0]	5.5 [0.0, 6.0]
One-week follow-up	5.0 [2.0, 7.0]	5.0 [1.0, 8.0]
One-month follow-up	6.0 [3.0, 7.0]	5.0 [0.0, 7.0]
Pain caused by light touch
Baseline	6.0 [4.0, 8.5]	6.0 [3.5, 8.0]
One-week follow-up	4.0 [3.0, 6.0]	6.0 [2.0, 9.0]
One-month follow-up	5.0 [4.0, 8.0]	7.0 [5.0, 7.0]
Itching
Baseline	3.0 [0.5, 5.0]	3.5 [0.0, 6.5]
One-week follow-up	2.0 [1.0, 5.0]	0.0 [0.0, 6.0]
One-month follow-up	4.0 [1.0, 6.0]	4.0 [0.0, 7.0]
Tingling or ‘pins and needles’
Baseline	7.0 [5.0, 9.5]	7.0 [5.0, 8.0]
One-week follow-up	6.0 [4.0, 8.0]	5.0 [2.0, 7.0]
One-month follow-up	6.0 [4.0, 7.0]	5.0 [4.0, 8.0]
Numbness
Baseline	7.0 [6.0, 9.0]	6.0 [4.0, 8.5]
One-week follow-up	7.0 [2.0, 8.0]	5.0 [3.0, 7.0]
One-month follow-up	6.0 [3.0, 8.0]	7.0 [5.0, 8.0]

¹Missing values - (1) baseline: none except 1 value is missing for the ‘Cold-freezing pain’ variable in Nitrous Oxide group;

(2) one-week follow-up: each study group had 1 missing value for all variables; (3) one-month follow-up: each study group had 3 missing values for all variables.

²Average daily opioid consumption during the previous 7-day period at baseline, one-week and one-month follow-up time points, reported in IV morphine equivalents (mg).

³The pain intensity for each of 6 items are measured on 0 (no pain) to 10 (worst pain) scale.

Among six neuropathic pain items surveyed at baseline and both follow-up times, the largest absolute difference in medians of these variables between 2 groups was equal to 2 on an 11-point scale, Table 4.

Missing data

For the primary outcome - pain score, there were 2 missing values at 1-week follow-up survey (1 for each group) and 6 missing ones at 1-month survey (3 for each group), Table 2. Here we assumed the missing data were missing at random, and missing values were imputed based on the multiple imputation procedure in our analysis. For the secondary outcomes, the pattern of missing values were similar with those for pain score. Since PROC MIXED procedure in SAS can use partial repeated outcome measurements, i.e., observations having missing outcome value at only one of 2 follow-ups, not both; and only 1 patient had missing secondary outcomes at 2 follow-ups, we did not impute missing values in our secondary analyses.

Among 44 patients included in our analyses, 2 patients were withdrawn from this study (see Appendix Table 1 for details on withdrawals including reason for withdrawal). The first withdrawn patient only missed pain score reported at 1-month follow-up survey, and the second withdrawn patient missed both pain scores at 2 follow-up time points. 6 protocol deviations occurred, Appendix Table 2. There were no serious adverse events.

Discussion

Treating chronic neuropathic pain with NMDA receptor antagonists has been of considerable interest to the pain management community. Nonetheless, it has proven challenging to translate the theoretical advantages of NMDA antagonists into consistent meaningful clinical benefit. Our results illustrate the challenge: 2 hours of 50% nitrous oxide exposure every other day across three sessions did not improve pain by a statistically significant or clinically meaningful amount. Lack of benefit in patients with CRPS is consistent with our previous trial in which nitrous oxide proved ineffective in patients with chronic low back pain with neuropathic symptoms.¹⁶ Available evidence therefore suggests that nitrous oxide — despite being an NMDA antagonist — does not provide meaningful prolonged analgesia in patients with chronic neuropathic pain. Similarly, nitrous oxide exposure did not impact chronic pain development after surgery. ¹⁷ In contrast, there is no question that inhaled nitrous oxide provides excellent acute analgesia.

Nitrous oxide and ketamine both are NMDA receptor antagonists, but are distinctly different drugs with different properties. Ketamine is a more potent NMDA receptor antagonist than nitrous oxide and also how they block the receptor is different which may be why we observed no benefit.¹² Interestingly ketamine also interacts with various other receptors including alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA), Gamma-aminobutyric acid (GABA), dopamine, serotonin, and mu, delta and kappa opioid receptors.¹⁸ These interactions could contribute to ketamine’s effect on CRPS pain. Furthermore, ketamine has active metabolites which nitrous oxide lacks which further contribute to NMDA antagonism.

PROMIS-29 v2 scores and global impression of change scores were included to formally assess functional status and mental health, neither of which demonstrated benefit from inhalation of nitrous oxide. A previous randomized trial of outpatient ketamine infusions for CRPS also evaluated functional status, using an accelerometer incorporated into a watch. Ketamine had no apparent effect on pre- and post-treatment values. Thus, while different measures were used, the results were similar between nitrous oxide and ketamine.¹⁹

The McGill Short Form questionnaire was included to characterize specific types of neuropathic pain. Statistical analysis on these exploratory data was not performed, but the results do not suggest any clinically meaningful effect of nitrous oxide inhalation. In distinct contrast, a previous trial of ketamine for CRPS reported a 35% decrease in the total McGill score — a benefit that was sustained at the final study evaluation at 9–12 weeks post-treatment.¹⁹ Benefit with ketamine, but not nitrous oxide, may be related to relative impotency of nitrous oxide on NMDA receptors. Fewer than half of our patients took opioid medications during the trial period which reduced our ability to evaluate the effect of nitrous oxide on opioid consumption. Although we avoided statistical testing for exploratory outcomes including opioid use, there was no evidence that nitrous oxide reduced opioid use.

Nitrous oxide dosing was empirical and based on what might be practical in routine use. Nitrous oxide is well tolerated at 50%, but higher concentrations routinely provoke nausea and vomiting. We used a 2-hour exposure because it seemed unlikely that all patients would tolerate longer periods, especially those assigned to Air-Oxygen group. Nitrous oxide could have been given more often or for longer but would become increasingly difficulty for patients to tolerate. Thus, while a more intense or longer exposure might provide benefit, practicality would suffer.

Other limitations include the purely subjective nature of pain which was our primary outcome. Validity of the trial thus depends critically on adequacy of blinding. Because many patients can detect inhaled nitrous oxide, we gave all participants midazolam which has similar effects. But it remains possible that some patients could distinguish nitrous oxide from air, and were thus biased. But we note that bias would presumably have augmented apparent effects of nitrous oxide, whereas there was actually no apparent benefit. We restricted enrollment to patients with CRPS and have previously reported no benefit in patients with chronic back pain. It nonetheless remains possible that nitrous oxide may prove useful for other pain syndromes.

In summary, there were no statistically significant or clinically meaningful improvements in pain scores, functional and mental health evaluated by PROMIS-29 v2 scores, or global impression of change scores in patients given 2-hours of 50% nitrous oxide every other day for three sessions. Our results do not support using nitrous oxide in current dosing and period for treatment of CRPS.

Key Points.

What is Already Known on this Topic?

Treatment of CRPS is multimodal and includes NMDA receptor antagonists such as ketamine infusions. Nitrous oxide is a low potency inhaled anesthetic with NMDA receptor antagonist activity. Unlike ketamine, nitrous oxide may cause sparse side effects.

What This Study Adds

Inhaled nitrous oxide did not provide meaningful therapeutic potential for patients with CRPS.

How This Study Might Affect Research, Practice or Policy

Our results do not support using nitrous oxide for treatment of CRPS.

Glossary of Terms

CRPS

Complex Regional Pain Syndrome

NMDA

N-methyl-D-aspartate

FDA

Food and Drug Administration

AMPA

alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid

GABA

Gamma-aminobutyric acid