Randomized Clinical Trial of Efficacy and Safety of a Single 2-mg Intravenous Dose of Hydromorphone Versus Usual Care in the Management of Acute Pain

Authors


  • Presented at the Society for Academic Emergency Medicine annual meeting, Chicago, IL, May 2012.
  • Clinical Trial Number: NCT01429298.
  • The authors have no relevant financial information or potential conflicts of interest to disclose.

Address for correspondence and reprints: Andrew K. Chang, MD, MS; e-mail: ACHANG@montefiore.org.

Abstract

Objectives

The objective was to test the efficacy and safety of 2 mg of intravenous (IV) hydromorphone (Dilaudid) against “usual care” in emergency department (ED) patients with acute severe pain.

Methods

This was a randomized clinical trial. Patients allocated to 2 mg of IV hydromorphone received their medication in a single dose. Those randomized to usual care received any IV opioid, with type, dose, and frequency chosen by the ED attending. All patients received 2 L/min. nasal cannula oxygen. The primary outcome was the difference in the proportion of patients who achieved clinically satisfactory analgesia by 30 minutes. This was defined as the patient declining additional analgesia when asked the question, “Do you want more pain medicine?” A 10% absolute difference was chosen a priori as the minimum difference considered clinically significant.

Results

Of 175 subjects randomized to each group, 164 in the 2 mg hydromorphone group and 161 in the usual care group had sufficient data for analysis. Additional pain medication was declined by 77.4% of patients in the 2 mg hydromorphone group at 30 minutes, compared to 65.8% in the usual care group. This difference of 11.6% was statistically and clinically significant (95% confidence interval [CI] = 1.8% to 21.1%). Safety profiles were similar and no patient required naloxone. There was more pruritus in the hydromorphone group (18.3% vs. 8.7%; difference = 9.6%, 95% CI = 2.6% to 16.6%).

Conclusions

Using a simple dichotomous patient-centered endpoint in which a difference of 10% in proportion obtaining adequate analgesia was considered clinically significant, 2 mg of hydromorphone in a single IV dose is clinically and statistically more efficacious when compared to usual care for acute pain management in the ED.

Resumen

Ensayo Clínico Aleatorizado de la Eficacia y la Seguridad de una Dosis Única de 2 mg de Hidromorfona Frente a la Atención Estándar en el Manejo del Dolor Agudo

Objetivos

Determinar la eficacia y la seguridad de 2 mg de hidromorfona (Dilaudid) intravenosa (IV) frente a la atención estándar en los pacientes con dolor agudo grave en el servicio de urgencias (SU).

Metodología

Se trata de un ensayo clínico aleatorizado. Los pacientes asignados a 2 mg de hidromorfona IV recibieron la medicación en una dosis única. Aquéllos asignados a la atención estándar recibieron cualquier opioide intravenoso, con tipo, dosis y frecuencia elegidos por el médico del SU. Todos los pacientes recibieron oxígeno en gafas nasales a 2 L/min. El resultado principal fue la diferencia en la proporción de pacientes que alcanzaron una analgesia satisfactoria a los 30 minutos. Ésta se definió como el paciente que declinó analgesia adicional cuando se le preguntó la cuestión: ¿quiere más medicación para el dolor? Una diferencia absoluta del 10% se eligió a priori como la diferencia mínima considerada clínicamente significativa.

Resultados

De los 175 sujetos asignados aleatoriamente a cada grupo, se obtuvieron suficientes datos para el análisis en 164 del grupo de hidromorfona 2 mg y en 161 del grupo de atención estándar. El 77,4% de los pacientes del grupo de hidromorfona 2 mg iv declinó medicación adicional para el dolor a los 30 minutos, en comparación con el 65,8% del grupo de atención estándar. Esta diferencia del 11,6% fue estadística y clínicamente significativa (IC 95% = 1,8% a 21,1%). Los perfiles de seguridad fueron similares y ningún paciente requirió naloxona. Hubo más prurito en el grupo hidromorfona (18,3% vs 8,7%; diferencia 9,6%, IC 95% = 2,6% a 16,6%).

Conclusiones

Cuando se usa un resultado simple dicotómico basado en el paciente y que considera clínicamente significativa una diferencia del 10% en la proporción de obtener una adecuada analgesia, 2 mg de hidromorfona en una dosis única IV es clínica y estadísticamente más eficaz cuando se compara con la atención estándar para el manejo del dolor agudo en el SU.

Provision of rapid and effective relief of acute, severe pain is a fundamental imperative of good clinical medicine. There is evidence that doses of analgesics given in practice are not large enough to achieve satisfactory analgesia for many patients.[1-5] Titration of analgesia appears to constitute the ideal strategy for management of acute pain.[3, 6, 7] However, two large multicenter studies have showed that pain scores are usually obtained only once in emergency department (ED) patients.[2, 4] This suggests a lack of reassessment after initial treatment, which would presumably preclude implementation of analgesic titration. Such practice seems entirely plausible, given the difficulty of reassessment of patients' pain in an overcrowded ED. If the use of a larger dose of opioid than is usually administered can safely provide pain relief to more patients than usual care, pain control may be able to be achieved with less need for individual titration.

In a prior study,[8] we found that administration of a single dose of 2 mg intravenous (IV) hydromorphone (Dilaudid) led to a large and rapid reduction in pain. However, approximately one-third of the study cohort had mild, transient oxygen desaturation (defined conservatively as < 95%).8 The clinical significance of this remains unclear, particularly when considered in light of the degree of desaturation reported in normal subjects during sleep[9] and air travel.[10-12] Nevertheless, for this study we chose to reexamine use of 2 mg IV hydromorphone as a single dose, with a protocol modification requiring that all enrolled patients receive 2 L of oxygen via nasal cannula. Unlike the original investigation,[8] which was a safety and efficacy study without a comparison group, in this study we directly compared the 2 mg IV hydromorphone protocol against “usual care.” Usual care was defined simply as use of any IV opioid at any dose, interval, or frequency, as determined by the patient's emergency physician (EP). Our objective was to test the efficacy and safety of 2 mg IV hydromorphone against usual care of ED patients with acute severe pain, using a dichotomous and patient-centered endpoint.

We chose to test the utility of a simple, dichotomous (yes/no) endpoint because we believe that it may be more useful at the bedside than the widely employed numerical rating scale (NRS). Some potential clinical advantages of use of such a binary endpoint include the following: 1) A yes/no response does not require the patient to be readily conversant with numbers or even be numerically literate; 2) unlike the NRS, a yes/no response (a nod or shake of the head is as easily understood) potentially leads directly to a therapeutic intervention; 3) this endpoint is squarely patient-centered in the sense that it invites the individual in pain to participate directly in his or her treatment; and 4) a summary yes/no response serves as a kind of “omnibus index” that can be used by individual patients to take into consideration, not just the balance of pain severity and incremental relief, but also to weigh the myriad of other relevant variables, such as common opioid side effects (e.g., nausea, vomiting, or pruritus), all of which seem unlikely to be satisfactorily summarized by a single number.

Methods

Study Design

This was a randomized clinical trial comparing the efficacy and safety of 2 mg IV hydromorphone to that of usual care for the ED treatment of acute severe pain. The study was approved by the Montefiore Medical Center institutional review board (IRB). All patients provided written informed consent in English or Spanish, depending on patient language preference. Both versions were approved by our IRB. The study was registered at clinicaltrails.gov.

Study Setting and Population

The study took place in an academic inner-city ED with an annual census of over 100,000 adults. Patients 21 to 64 years of age presenting to the ED with acute pain were enrolled if their pain was of sufficient severity to warrant use of IV opioids in the judgment of the EP. Exclusion criteria were as follows: patients managed by any of the EPs who were investigators in this study, history of allergy to hydromorphone or morphine, hypotension (systolic blood pressure < 90 mm Hg), room air oxygen saturation < 95%, alcohol or other drug intoxication as judged by the EP, use of other opioids within the past 7 days, use of a monoamine oxidase inhibitor, pregnancy, history of a chronic pain syndrome (such as sickle cell disease or fibromyalgia), and weight less than 150 pounds. We excluded patients weighing less than 150 pounds because we were concerned that 2 mg IV hydromorphone might be a potentially dangerous dose for such patients.

Research associates (RAs) consented and enrolled patients referred to them by EPs. The RAs were fluent in English and Spanish and enrolled patients 24 hours per day, 7 days per week, from March to November, 2010.

Study Protocol

Patients were randomized to the 2 mg IV hydromorphone group or usual care; allocation was generated with www.randomization.com, using sealed opaque envelopes opened in sequential order by the RAs immediately following enrollment. Patients randomly allocated to usual care received an initial dose of IV opioid; the type and dose of which was determined by the treating EP. Patients in the 2 mg hydromorphone group were allocated to receive 2 mg IV hydromorphone, administered slowly over 2 to 3 minutes. All patients were placed on 2 L O2 by nasal cannula. Subjects were blinded to the treatment they were assigned. At 30 minutes, both groups were asked the following scripted question: “Do you want more pain medication?” Patients in either group who answered or otherwise indicated “yes” had their treating attending physician notified, who then decided on further pain management. Those who answered or otherwise indicated “no” did not receive additional analgesia at that time. Attending physicians were thus able to treat patients' pain in any manner they deemed fit once this primary study endpoint was reached.

In addition to acquisition of the primary endpoint, patients were also asked to rate their pain on a previously validated and reproducible[13] standard verbal NRS ranging from 0 (“no pain”) through 10 (“worst pain possible”) at 15, 30, 45, 60, 90, and 120 minutes following administration of the initial opioid dose at time 0.

For safety reasons, patients were monitored for a total of 120 minutes (i.e., 90 minutes past the primary study endpoint) to determine adverse effects. Systolic blood pressure, heart rate, oxygen saturation, nausea, vomiting, and pruritus were assessed at baseline and at 15, 30, 45, 60, 90, and 120 minutes after initial administration of opioid.

Patients who experienced oxygen desaturation (defined as < 95%) were gently aroused if sleeping, asked to take several deep breaths, and repositioned into a sitting position if they had been in a reclined position. Nasal cannula oxygen was also increased to 4 L, and the treating attending physician was notified. Subsequent management, including the use of naloxone, was per the treating attending physician's discretion.

Measures

The primary efficacy outcome was the difference in the proportion of patients in each group who achieved satisfactory analgesia, defined a priori as declining additional pain medication when asked at 30 minutes. We chose to use a dichotomous outcome based on the scripted question, “Do you want more pain medication?” for the several reasons listed previously. Secondary efficacy outcomes, all of which were determined a priori and measured concurrently with the primary outcome at 30 minutes, included change in NRS, proportion of patients achieving ≥ 50% decline in NRS,[14, 15] proportion of patients achieving an absolute NRS score ≤ 3,[16, 17] and proportion of patients' describing their pain as “none” or “mild” (in contrast to descriptions of “moderate” or “severe” on a modified four-category Likert scale).

The primary safety outcome was use of naloxone at any point during the study, which could be given for any reason as determined by ED attending judgment. Based on prior work, our secondary safety outcomes were adverse effects of desaturation < 95% at any time during the study[8]; hypotension (<90 mm Hg), or bradycardia (<50 beats/min); and side effects of nausea, vomiting, or pruritus. Information about all additional medications administered to the patients, including name of medication, dose, route of delivery, and time of administration were abstracted from the medical record in real time.

Data Analysis

Data were recorded on a standardized data collection instrument and entered into SPSS Data Entry (SPSS, Inc., Chicago, IL) by a trained data clerk. Double entry of key variables was performed by a second trained clerk and any discrepancies were reconciled by referral to the original data collection instrument.

All variables are shown as means with standard deviations (SDs), medians with interquartile ranges (IQRs), or proportions with 95% confidence intervals (95% CIs). To estimate precision and statistical significance, we used Wilson's method of calculating a 95% CI around differences between proportions.[18] To calculate CIs around differences between means, standard methods were used.[19] To facilitate meaningful comparison between the two groups, we calculated the amount of opioid analgesic administered at different time points in the study in morphine equivalent units (MEU). Because estimates of the morphine equivalence of hydromorphone varies, we chose to use an intermediate value of 1 mg hydromorphone = 7 mg morphine.[20] Thus, 1 mg morphine = 1 MEU and 1 mg hydromorphone = 7 MEU. No other IV opioid analgesics were prescribed. SPSS version 17 (SPSS Inc., Chicago, IL.) was used to conduct all data analyses.

Sample Size Calculation

We calculated that a sample size of 320 patients (160 per group) would be needed to detect at least a 10% difference in percentage of patients declining additional pain medication when asked at 30 minutes postbaseline, using a two-tailed alpha of 0.05 and power of 0.80. To ensure enrollment of a minimum of 320 patients for analysis, an additional 30 patients (approximately 10%) were randomized to account for potential protocol violations and missing data. We used nQuery Advisor version 6.0 (Statistical Solutions, Los Angeles, CA) to calculate the sample size.

Results

A total of 175 subjects were randomized to each group for a total enrollment of 350 patients. As shown in the CONSORT flow diagram (Figure 1), 10 patients (four in the 2 mg IV hydromorphone group and six in the usual care group) were enrolled but were never given study medications, and hence had no study data available for analysis. Two patients (both in the usual care group) were enrolled twice and their second visits were removed. Four patients (two in each group) were missing the primary endpoint and were removed. Five patients in the hydromorphone group and four in the usual care group received IV ketorolac while in the ED but before the beginning of the study and were excluded (patients in the usual care group who were coadministered IV ketorolac in conjunction with the initial opioid were included). Removal of these 16 patients (Figure 1) left a total of 164 patients in the 2 mg IV hydromorphone group and 161 patients in the usual care group, for a total of 325 patients available for analysis, thus exceeding the target sample size of 320 patients.

Figure 1.

CONSORT diagram.

Table 1 displays baseline characteristics of the sample. The majority of patients had abdominal pain and reported their pain severity as 10 on the NRS. Baseline features were sufficiently balanced that no multivariable adjustment for chance maldistribution was necessary.

Table 1. Baseline Characteristicsa
Characteristic2 mg Hydromorphone Group (n = 164)Usual Care Group (n = 161)
  1. Values reported as mean (±SD) or n (%).

  2. NRS = numeric rating scale.

  3. a

    Some numbers do not add up to total sample size due to missing data.

Sex
Female101 (61.6)91 (56.5)
Male63 (38.4)70 (43.5)
Race/ethnicity
Hispanic113 (68.9)103 (64.0)
African American38 (23.2)41 (25.5)
White10 (6.1)12 (7.5)
Other3 (1.8)5 (3.1)
Age (years) 41.4 (±12.2)41.2 (±12.3)
Weight (lbs) 195.4 (±39.0)201.4 (±42.8)
Location of pain
Abdomen101 (61.6)110 (68.3)
Other63 (38.4)51 (31.7)
Pain intensity (NRS)
3 to 719 (11.6)21 (13.0)
822 (13.4)22 (13.7)
920 (12.2)15 (9.3)
10103 (62.8)103 (64.0)
Nauseated or vomited before receiving opioids in the ED
Yes96 (58.9)99 (61.5)
No67 (41.1)62 (38.5)

Table 2 displays the single primary and four secondary efficacy outcomes, all determined at 30 minutes postbaseline and all chosen a priori. More patients (77.4%) in the 2 mg IV hydromorphone group declined additional pain medication at 30 minutes, compared to 65.8% in the usual care group. This difference of 11.6% was both clinically and statistically significant (95% CI = 1.8% to 21.1%), according to a priori criteria.

Table 2. Primary and Secondary Efficacy Outcomes
Outcomes2 mg Hydromorphone Group (n = 164)Usual Care Group (n = 161)Difference (95% CI)
  1. Values are n (%) unless otherwise noted

    NRS = numeric rating scale.

Primary
Declined additional medication at 30 minutes 127 (77.4)106 (65.8)11.6 (1.8–21.1)
Secondary
Mean (±SD) change in NRS pain score baseline to 30 minutes6.7 (±2.9)5.6 (±3.2)1.2 (0.5–1.8)
Achieved ≥ 50% decline in NRS pain score at 30 minutes136 (82.9)111 (69.0)13.9 (4.3–22.6)
Achieved absolute pain score ≤ 3 at 30 minutes114 (69.5)90 (55.9)13.6 (3.1–23.7)
Reported no pain or mild pain at 30 minutes122 (74.4)95 (59.7)14.6 (4.4–24.5)

The secondary outcomes employed more traditional NRS pain scores and descriptive verbal ratings of pain. Consistent with the primary endpoint, each of the secondary outcomes favored the 2 mg IV hydromorphone protocol over usual care, as shown in Table 2.

After conversion of all opioids administered in both groups into MEU, we determined that patients in the usual care arm received an initial median opioid dose of 6 MEU (IQR = 4 to 7 MEU) that was slightly less than half that of the 2 mg hydromorphone arm, who all received (by protocol) 14 MEU. One-third of the usual care subjects received IV hydromorphone as their initial analgesic; the remaining patients allocated to usual care received IV morphine. In the usual care arm, in which additional analgesia could be given at any time and in any amount, five patients received additional opioid between baseline and 30 minutes: four received an additional 1 mg IV hydromorphone (7 MEU), and a single patient received an additional 6 mg IV morphine (6 MEU) during this time. Nine patients in the usual care group were coadministered IV ketorolac in addition to their initial dose of opioid.

By 120 minutes postbaseline (90 minutes after the primary endpoint had occurred), the total median dose of opioid in the usual care group was 7 MEU (IQR = 4 to 14 MEU) versus 14 MEU (IQR 14 to 14 MEU) in the 2 mg hydromorphone group. Similar to the findings at 30 minutes, by 120 minutes, the usual care group received about half that of those in the 2 mg hydromorphone arm. Significantly more patients in the usual care arm received additional opioids between 30 and 120 minutes: 32% (n = 52) of patients in the usual care group required additional IV opioids in contrast to 13% (n = 22) in the IV hydromorphone group, for a difference of 18.5% (95% CI = 9.6% to 27.1%).

Table 3 summarizes the incidence of adverse effects and side effects by group. No patient required use of IV naloxone. Two patients, both in the usual care group, experienced transient desaturation (defined a priori as oxygen saturation < 95%) at one or more time points after administration of opioid. Bradycardia and hypotension were uncommon and similar in both groups. No adverse effect required intervention, other than gentle arousal and, if appropriate, repositioning from the supine to sitting position. With respect to opioid side effects, both pruritus (18.3% vs. 8.7%, for a difference of 9.6%, 95% CI = 2.6% to 16.6%) and nausea (26.9% vs. 14.5%, for a difference of 12.4%, 95% CI = −1.8% to 25.8%) were both substantially more common in the hydromorphone group, although only the difference in pruritus achieved statistical significance.

Table 3. Adverse Effects and Side Events by Group
Effects and Events2 mg Hydromorphone Group (n = 164)Usual Care Group (n = 161)Difference in percentage (95% CI)
  1. Values are reported as n (%).

  2. a

    Refers only to those patients who developed nausea or vomiting after receiving opioids, but had neither symptom prior to receiving opioids.

Administration of naloxone0 (0)0 (0)0 (0)
Oxygen saturation < 95%0 (0)2 (1.2)–1.2 (–4.4 to 1.2)
Heart rate < 50 beats/min5 (3.0)4 (2.5)0.6 (–3.5 to 4.7)
Systolic blood pressure < 90 mm Hg3 (1.8)1 (0.6)1.2 (–1.8 to 4.7)
Pruritus30 (18.3)14 (8.7)9.6 (2.6 to 16.6)
Nauseaa18/67 (26.9)9/62 (14.5)12.4 (–1.8 to 25.8)
Vomitinga5/67 (7.5)3/62 (4.8)2.6 (–6.8 to 12.0)

Discussion

In a randomized clinical trial in which a difference of 10% in proportion obtaining adequate analgesia was considered clinically significant, we found that 2 mg IV hydromorphone was both clinically and statistically more efficacious than usual care for treatment of acute pain in the ED. About three-quarters of the patients randomly allocated to the 2 mg IV hydromorphone arm achieved the primary endpoint of declining further analgesia at 30 minutes postadministration. This is in contrast to two-thirds of the patients receiving usual care, for an absolute between-group difference of 11.6%. We specified a priori that a difference of 10% would constitute the minimum threshold for declaration of a clinically meaningful effect.

We also examined several secondary measures of efficacy based on the quantitative NRS pain scores and verbal descriptors of pain (Table 2). A frequently used measure in pain research is the minimum clinically significant difference in pain defined as the arithmetic difference in mean change in pain scores of approximately 1.3 NRS units or more.[13, 21-24] The mean difference in change in pain in this study based on an NRS scale ranging from 0 (no pain) to 10 (worst possible pain) scale was 1.2 NRS units (95% CI = 0.5 to 1.8). While the difference between groups in this study did not formally meet this standard criterion of 1.3 NRS units, the CI is well within that of other studies.[13, 21-24] Two other measures based on the NRS that have been used in other studies,[14-17] plus categorical descriptors of pain at 30 minutes after initiation of treatment, also indicated better pain control in the patients who received 2 mg IV hydromorphone. The proportion of patients who achieved a 50% or greater proportionate decrease in NRS pain score, an absolute NRS score of ≤3, or described their pain as “none” or “mild” at 30 minutes, was about 15% higher in the 2 mg hydromorphone group compared to the usual care group.

Two mg IV hydromorphone, which is equivalent to approximately 14 mg IV morphine,[20] is substantially higher than the amount of opioid typically given in emergency practice and therefore carries with it the possibility of increased adverse events and side effects. The design of the protocolized intervention arm of this study was drawn from an earlier safety and efficacy study of 2 mg IV hydromorphone administered as a single bolus.[8] In that study we found that this dose resulted in rapid and marked pain relief. However, approximately one-third of patients experienced mild, transient oxygen desaturation (defined arbitrarily as <95%). Although the clinical importance of this is unknown, we modified the protocol in the current study to place all patients on oxygen (2 L via nasal cannula) prior to the initial bolus of opioid. It is possible that this was overly conservative, since greater levels of desaturation are uniformly seen in normal subjects on commercial airline flights[10-12] and during sleep,[9] without any known short- or long-term adverse effects. With this modification, the oxygen saturation of two patients, both randomized to usual care, transiently fell below 95%, returning immediately to above 95% by combining gentle arousal with repositioning of the patient from the supine to sitting position. No patient randomized to 2 mg IV hydromorphone desaturated, and no patients in either group required naloxone.

Pruritus occurred in about 10% more patients randomized to 2 mg IV hydromorphone than in those receiving usual care. The incidence of pruritus was also approximately two to three times higher than that found in earlier work among patients who received 1 mg of hydromorphone,[25-27] generating the hypothesis that this may represent a dose effect. There was also some suggestion of an increased incidence of nausea in the 2 mg IV hydromorphone group, although the difference did not reach statistical significance.

We anticipated that a 2 mg dose of hydromorphone would meet the initial opioid requirements of most ED patients in acute pain. However, one-quarter of these patients wanted more analgesia when asked at 30 minutes. While patients in the 2 mg hydromorphone group received approximately twice the dose of opioid when compared to the usual care group in morphine equivalent units, only about 12% more experienced adequate analgesia than those treated with standard care. We speculate that the lack of a clear, linear inversely proportionate relationship between amount of opioid administered and need for additional analgesia when asked reflects the tremendous amount of interpatient variability in opioid requirement.[16, 28, 29] This is presumably driven, at least in part, by substantial differences in the “objective” magnitude of the painful stimulus; its subjective perception; and the poorly understood physiologic variation in opioid receptor sites, their type, regulation, saturation kinetics, and relationship to analgesia, in addition to broad variation in other aspects of opioid pharmacokinetics, including rate of drug metabolism.[30-33]

The marked natural variation in opioid requirement is well illustrated by a study of morphine titration in 621 ED patients by Lvovschi et al.3 Patients received 2- to 3-mg boluses of morphine every 5 minutes administered by nurses until patients reported a visual analog scale (VAS) score of 30 or less, the study's prespecified definition of adequate analgesia. The mean (±SD) dose needed to achieve this level of analgesia was 11 (±6) mg morphine. Assuming that the sampling distribution of the dose of morphine administered is normal (not an unreasonable assumption with a sample size of over 600 patients), the dose required to achieve adequate analgesia can be estimated to be between 5 and 17 mg for the 64% of patients whose dose was within 1 SD of the mean dose.3 We draw attention to this previously reported threefold variation in opioid dose needed to achieve a threshold endpoint, believed by these investigators to represent adequate analgesia, primarily to place in context our findings of a twofold variation in opioid dose required to reach a different but analogous endpoint. It is noteworthy that the threefold variation is seen among the two-thirds of the study population closest to the mean. The remaining third, by definition, can only demonstrate greater variance in opioid requirement. This indicates that our data appear to be well within a conservative assessment of the marked interpatient variation in quantity of opioid needed, as determined by the patient presenting in acute, severe pain.

Our finding that one-quarter of patients did not achieve adequate analgesia, despite receiving the equivalent of 14 mg of morphine, simply represents the “other side” of the broadly splayed distribution that appears to represent the natural variation inherent in opioid dose required for achievement of satisfactory analgesia in the patient population studied. Generalizability of these and similar findings, which is at least biologically plausible, requires confirmation by other investigators in populations with different demographic characteristics.

Limitations

This study's limitations center around the a priori decision to use a primary categorical endpoint, expressed by the patient as a simple yes/no to the question, “Do you want more pain medication?” rather than the more traditional ordinal/continuous endpoint measured as a number on an NRS or VAS.

The first problem with such a choice is that the majority of previous pain studies,[34-38] including our own,[8, 14, 26, 39-43] have used the NRS or VAS scores, thus making our results difficult to evaluate by using the traditional scientific approach of examining new findings within the context of comparable prior work.[8, 14, 26, 34-43] In spite of this important limitation, we chose to depart from the traditional NRS primary endpoint and use a categorical one because it seemed to us to be more clinically relevant to aim at adequate individual analgesia as a target endpoint, rather than a delta NRS, which represents the smallest improvement in pain score worth detecting. We reasoned further that such a tradeoff of endpoints would be clinically worthwhile because the standardized question and its simple dichotomous response struck us as more likely to serve the patient as a true “omnibus index.” By the term “omnibus index,” we mean that a simple yes/no response allows the individual to take into consideration not just the pain severity and extent of relief measured by the NRS, but also to place one's current degree of analgesia in the context of unwanted concomitants of opioid use, such as nausea, vomiting, or pruritus, as well as other clinically important side effects that are difficult to articulate, such as the sensation of being “high” or out of control, which many individuals find unsettling.

A second limitation of our choice of a binary endpoint is inherent in the mathematics of traditional frequentist (non-Bayesian) sample size calculation, i.e., any given delta expressed as a proportion (such as the endpoint used in this study) will require a larger sample size, and therefore a greater expenditure of resources, than the “analogous” delta expressed as a mean (such as the NRS). In addition, our sample size (though relatively large) was not large enough to detect small differences in the incidence in adverse effects.

A third limitation is that the choice of a novel categorical endpoint requires selecting a clinically meaningful difference (the delta, which drives the sample size calculation, referred to above) in the proportion of patients who choose to decline additional analgesia. The delta NRS, which constitutes the minimum clinical difference worth detecting, has already been identified and shown to be valid and reproducible.[13, 23, 44] To our knowledge, there is no evidence supporting (or literature recommending) how large a difference in the absolute proportion of patients attaining satisfactory analgesia constitutes a clinically meaningful delta. We chose a 10% absolute difference in successful treatment between groups to define a minimum threshold for declaring clinical significance.

It could be argued that a smaller between-group difference in the proportion of patients achieving self-defined adequate analgesia, such as 5%, is also clinically meaningful because it represents the real goal of acute pain management, rather than the minimum clinically important difference in pain severity. However, as mentioned above in the second limitation of a necessarily increased sample size due to the choice of a binary rather than continuous endpoint, the need to consider carefully the cost of conducting clinical studies, and to remain within the slim budget allocations of current grant awards, requires all investigators to compromise between the scientific need of not missing the smallest treatment difference judged to be worth detecting, and real-world considerations, which require that one remain within the range of an obtainable sample size, taking into account time, cost, and other resources that might be deployed more usefully elsewhere.

Thus, our choice of a 10% delta represents the same kind of limitation inherent in all threshold criteria newly chosen by investigators to define clinical significance: it is necessarily an arbitrary, though not capricious, compromise and cannot be otherwise. We invite other investigators to consider whether such a difference (and choice of a dichotomous, rather than a dimensional/ordinal NRS delta as an endpoint) seems clinically sensible in measuring outcomes that are most important to patients.

A fourth limitation is that this is a single-blind randomized clinical trial. By design, the medical staff were not (and could not be) blinded to study group allocation, because the attending physicians were responsible for choosing the analgesic treatment for patients randomly allocated to usual care. Knowledge of the relatively large dose of opioid given to the protocolized group may have influenced the physicians' use of the significantly more potent opioid hydromorphone as well as driven administration of higher doses of morphine in the usual care group. The effect of this potential performance bias would be to underestimate the difference between the 2 mg hydromorphone group and usual care by driving the target outcome toward the null.

A fifth limitation is that both groups were placed on 2 L of supplemental nasal cannula oxygen, and this is not likely standard practice in most EDs. We built this into our protocol, however, based on a prior study in which approximately one-third of patients experienced oxygen desaturation when given 2 mg IV hydromorphone.[8]

The population served is largely Latino, African American, urban, and poor and weigh at least 150 pounds. It is possible that requesting additional pain medication may be influenced by many factors including culture and expectation of pain relief.[45, 46] Thus the results may not generalize to other patient populations with different sociodemographic characteristics and patients who weigh less than 150 pounds. Given that the racial and weight distribution of the two arms of the study were similar, this is not a problem for comparison of the efficacy and safety of 2 mg hydromorphone versus usual care and should not compromise internal validity.

Conclusions

In this randomized clinical trial, in which a difference of 10% in proportion obtaining adequate analgesia was considered clinically significant, the administration of 2 mg of hydromorphone in a single IV dose is both clinically and statistically more efficacious when compared to usual care for acute, severe pain management in the ED. The incidence of pruritus was higher among patients randomly allocated to 2 mg IV hydromorphone, which may be a consequence of higher opioid dosing. Despite the size of this dose, equivalent to about 14 mg morphine, one-quarter of the patients in this arm wanted additional analgesia when asked at 30 minutes. This, combined with other findings, suggests that the opioid requirement of a substantial number of ED patients commonly exceeds the traditional range of doses administered. We hypothesize that this is a consequence of marked interindividual variation in amount of opioid required to achieve satisfactory analgesia, further highlighting the need for individualized titration as the most rational approach to acute pain management in the ED.

Ancillary