Meta-analyses report similar numbers needed to treat for nonsteroidal antiinflammatory drugs (NSAIDs) and opioids. Differences in baseline pain intensity among the studies from which these numbers needed to treat were derived may have confounded the results. NSAIDs have an opioid-sparing effect, but the importance of this effect is unclear. Therefore, the authors sought to compare the proportions of subjects who obtain pain relief with ketorolac versus morphine after surgery and to determine whether the opioid-sparing effect of an NSAID reduces the magnitude of opioid side effects.
The study was a double-blind, randomized controlled trial. The authors randomly assigned 1,003 adult patients to receive 30 mg ketorolac or 0.1 mg/kg morphine intravenously. They calculated the proportion of subjects who achieved at least 50% reduction in pain intensity 30 min after analgesic administration. Further, so long as pain intensity 30 min after analgesic administration was 5 or more out of 10, patients received 2.5 mg morphine every 10 min until pain intensity was 4 or less out of 10. The authors assessed the presence of opioid-related side effects.
Five hundred patients received morphine and 503 received ketorolac. Fifty percent of patients in the morphine group achieved pain relief, compared with 31% in the ketorolac group (difference, 19%; 95% confidence interval, 13-25%). The ketorolac-morphine group required less morphine (difference, 6.5 mg; 95% confidence interval, -5.8 to -7.2) and had a lower incidence of side effects (difference, 11%; 95% confidence interval, 5-16%) than the morphine group.
Opioids are more efficacious analgesics than NSAIDs, although historic data for these two drugs yield similar numbers needed to treat. Adding NSAIDs to the opioid treatment reduces morphine requirements and opioid-related side effects in the early postoperative period.
META-ANALYSES of randomized, double-blind, single-dose studies that evaluate opioids and nonsteroidal antiinflammatory drugs (NSAIDs) for acute postoperative pain have yielded overlapping numbers needed to treat (NNTs) for NSAIDs and opioids.1,,2,,3 The NNT is a measure of the effect size of a treatment that can be used to rank treatments on the basis of effectiveness. The NNTs to achieve at least 50% pain relief over 4–6 h with a single dose of the following drugs are 2.7 for 400 mg oral ibuprofen (95% confidence interval [CI], 2.5–3.0), 2.6 for 30 mg intramuscular ketorolac (95% CI, 2.3–3.1), and 2.9 for 10 mg intramuscular morphine (95% CI, 2.6–3.6). These NNTs mean that of three patients exposed to any of these medications, one will achieve at least 50% pain relief over 4–6 h. These results contrast with clinical experience and consensus recommendations that opioids are more effective than NSAIDs for moderate-to-severe acute pain.5,,6
The discrepancy between clinical experience and the results of meta-analyses could be due to the fact that the studies that served as a basis to estimate NNTs for opioids and NSAIDs did not directly compare agents in these two classes. Instead, in such studies, NSAIDs and opioids have been compared with placebo. The result of this indirect comparison may overestimate the efficacy of NSAIDs. The validity of the adjusted indirect comparisons depends on the internal validity and similarity of the included trials. Therefore, differences in baseline pain intensity of subjects enrolled in the studies could have confounded the results because the higher the baseline pain intensity is, the greater the decrease in pain intensity must be to obtain a similar degree of pain relief.
The use of both NSAIDs and opioids concurrently for the treatment of acute pain is common practice. Adding an NSAID to an opioid regimen reduces opioid requirements, which could lead to a decrease in opioid-related side effects. A key weakness of the evidence for such practical benefit, based on studies that have evaluated the opioid-sparing effect of NSAIDs, is that they were not designed to detect a difference in the incidence and severity of side effects. To detect a difference in side effects between treatment groups, in general, larger numbers of patients are needed than simply to evaluate analgesia or opioid requirements.
Therefore, we designed a large study, first to compare the analgesic efficacy of an NSAID and an opioid in a head-to-head trial by determining the proportion of subjects who obtained adequate postoperative pain relief 30 min after analgesic administration, and second to determine whether the opioid-sparing effect of NSAIDs decreases the risk of opioid side effects, by comparing opioid requirements and side effects in patients who received ketorolac plus morphine or morphine alone.
Materials and Methods
This study was approved by the Institutional Review Board of San Ignacio Hospital (Bogota, Colombia), a tertiary care university teaching hospital. Individuals who agreed to participate provided written informed consent before surgery. We recruited patients from May 2003 until November 2003. Patients aged between 18 and 60 yr who underwent surgical procedures were potentially eligible for the study. We excluded subjects older than 60 yr because in this population, it is necessary to decrease the dose of opioids to reduce the likelihood of side effects. Patients were ineligible if they were pregnant, had asthma, were allergic to NSAIDs, had a history of gastric ulcer or abnormal renal function, had received NSAIDs 6 h before surgery or during surgery, had a history of long-term opioid use, had received an opioid within 5 days before surgery, or had postoperative hemodynamic instability, e.g. , due to intraoperative blood loss.
We placed no restrictions on the type of anesthetic or other intraoperative medication except that the use of NSAIDs was not allowed and that fentanyl was the only opioid permitted. The intraoperative dose of fentanyl used, if any, was recorded.
We conducted a randomized, double-blind, controlled trial. In the postanesthesia care unit, we randomly assigned patients who reported pain of at least moderate intensity to receive ketorolac or morphine. To ensure that the number of subjects with moderate to severe pain intensity was similar in each treatment arm, we performed a stratified randomization, using a computer-generated random number program.
The randomization was concealed; the operating room pharmacy was in charge of treatment assignment, preparation of the solutions, and handing the corresponding solution to the evaluator.
When patients arrived in the postanesthesia care unit, we asked them to describe their pain as none, mild, moderate, or severe. Patients were randomized if their pain intensity was at least “moderate.” The analgesic infusion of 30 mg intravenous ketorolac or 0.1 mg/kg morphine over 3–5 min then began, from the start of which we asked patients to rate their pain intensity at rest on a 0–10 numerical rating scale, where 0 represents no pain and 10 represents the worst pain imaginable, every 10 min.
If the pain intensity 30 min after initiation of the analgesic infusion was 5 or higher, patients in both groups received a bolus of 2.5 mg intravenous morphine every 10 min until their pain intensity was 4 or less. Therefore, one group received only morphine for postoperative analgesia (morphine group) and the other group received ketorolac followed by morphine (ketorolac–morphine group).
Simultaneously with the pain intensity evaluation, patients rated their pain relief using a five-point Likert scale that ranged from no relief to complete pain relief. We documented the number of rescue doses required in each group.
The dose of ketorolac is that recommended by the manufacturer, and the dose of morphine is that commonly used for pain relief and in pharmacokinetic studies.10,,11,,12 In addition, this morphine dose is within the dose range (6–10 mg) used in studies that evaluate the equipotency of ketorolac.13,,14,,15,,16,,17,,18
Clinical evaluations were done by observers who were blinded to patient assignment. The evaluators received the solutions from the pharmacy personnel in 100-ml normal saline bags that were identical in appearance. Each solution was infused over 3–5 min. The randomization code was kept concealed in the operating room pharmacy.
The study was powered (500 patients/group) to detect a 10% intergroup difference in the proportion of patients who achieved at least 50% decrease in initial pain intensity, with an α level of 0.05 (two tailed) and a β level of 0.9. We assumed, based on a published meta-analysis, that 46% of the subjects in the morphine group would achieve at least a 50% decrease in pain intensity, and we hypothesized that 36% of subjects in the NSAID group would achieve this degree of relief.
The enrollment of 500 patients/group also permitted us to detect, with 80% power, a 5% difference in the incidence of a side effect with prevalence as low as 10%, such as would be the case with pruritus. For more common side effects such as sedation, nausea, vomiting, and dizziness, the study was overpowered.
Assessment of Outcomes and Study Procedure
The primary outcome was the proportion of subjects who reported a 50% decrease in pain intensity 30 min after the initiation of the intravenous infusion.
We chose a threshold pain intensity decrement of 50% because this value has been extensively applied to evaluate pain treatments and to estimate NNTs and represents a clinically meaningful decrease in pain intensity from the patients' point of view.
We chose to evaluate the analgesic response at 30 min because ketorolac and morphine not only have a similar onset of analgesic action—time to peak plasma concentration for 30 mg intravenous ketorolac is around 5 min and that for morphine is 10 min—but also a similar first-order rate constant Keo (15–20 min for both ketorolac and morphine). Keo determines the equilibration of drug between plasma and effect site; the effect site concentrations determine the pharmacodynamic effects. This pharmacokinetic similarity is confirmed by the fact that the analgesic effects of both ketorolac and morphine are observed at this time period.11,,22 In addition, the evaluation at 30 min is clinically relevant and reflects clinical judgment—that it would be unethical and impractical to delay supplemental analgesic therapy by more than 30 min after patients have received an initial therapeutic dose of an analgesic.
Evaluation of Side Effects
Simultaneously with pain intensity, patients evaluated the presence and severity of sedation, nausea, vomiting, pruritus, and dizziness as absent, mild, moderate, or severe. We recorded the number of episodes of vomiting and instances of antiemetic therapy.
We conducted an intention-to-treat analysis. For analgesic effectiveness, we calculated the proportion of subjects who reported a 50% or greater decrease in pain intensity 30 min after the initiation of the intravenous analgesic infusion. The percentage of pain reduction was calculated as 100 times the difference between pretreatment and posttreatment pain intensity, divided by the pretreatment pain intensity. In addition, we calculated risk differences and the NNT for analgesia, along with 95% CIs.
To evaluate the effects of sex and baseline pain intensity on the odds of having 50% or greater decrease in the numerical rating scale, we applied a logistic regression model. This model included these variables as well as the interaction between treatment and baseline pain intensity to determine which of the two analgesics was more efficacious in relation to baseline pain intensity.
In addition to pain relief, we assessed whether pain intensity differed between the morphine and ketorolac–morphine groups, using a linear regression. Because each patient had multiple evaluations (until pain intensity was ≤ 4 out of 10) and these measures were not independent, we used an analysis of repeated measures using generalized estimating equations. That method takes this lack of independence into consideration by adjusting the standard errors.
To determine whether there was a difference in opioid requirements between the groups, we used a robust linear regression, because the total dose of morphine had outliers (patients with unusually high requirements). Robust regression does not assume normality and minimizes the impact of influential observations; it produces valid and more efficient estimates than the traditional linear regression. It assigns less weight to observations that have larger residuals or observations that are very influential. This technique has been previously used in the analysis of analgesic studies.28,,29
To determine whether opioid requirements were different in men and women, we estimated the dose of morphine in milligrams per kilogram according to sex and performed a robust linear regression.
To evaluate side effects, we created a variable that summarized the presence of any side effect during the duration of the study and estimated the 95% CI for the difference in proportions. Because the overall incidence of any adverse effect differed significantly between the groups, we compared each side effect. We also estimated the number needed to harm as an index of the safety and tolerability of a medication and the 95% CI.
The sample size calculations, randomization, and analyses were performed with STATA® software version 7SE (College Station, TX). P values less than 0.05 were considered statistically significant.
We followed the CONSORT guidance for reporting results. We randomized 1,003 patients. Five hundred subjects received morphine, and 503 received ketorolac. There were no protocol violations, and all patients were included in the analyses ().
Demographic characteristics, duration and type of surgery, type of anesthesia, intraoperative fentanyl, and baseline pain intensity were similar in both groups. Fifty-eight percent of patients had severe pain at the start of the infusion ().
Fifty-nine percent of all subjects did not achieve a 50% or greater decrease in pain intensity. The probability of not achieving 50% or more of pain relief increased when the initial pain intensity was severe ().
Morphine provided a superior analgesic effect to ketorolac for patients with both moderate and severe pain (). Fifty percent of the subjects in the morphine group achieved a 50% or greater decrease in pain intensity versus 31% in the ketorolac group. Therefore, the difference in the proportion of patients who achieved a 50% decrease in pain intensity was 19% (95% CI, 13–25%), which corresponds to an NNT of 5 (95% CI, 4–10).
The degree of pain relief was also higher in the morphine group. Forty-nine percent of the subjects in the morphine group reported at least “much improvement,” in contrast with the ketorolac group, in which 39% of the subjects reported at least “much improvement” (P = 0.005).
Effect of Sex and Baseline Pain Intensity
Patient sex did not affect the odds of having adequate pain relief (odds ratio = 1.0; 95% CI, 0.7–1.3). Baseline pain intensity was a good inverse predictor of having at least a 50% or greater decrease in pain intensity. That is, if the baseline pain intensity was severe, the odds of having a 50% decrease in pain intensity were lower than when the pain was moderate (odds ratio = 0.4; 95% CI, 0.3–0.5). The interaction between baseline pain intensity and analgesic treatment was not significant (P = 0.2).
Pain Intensity throughout the Duration of the Study
Pain intensity was higher in the ketorolac–morphine group during the first 30 min after analgesic administration (when patients in this group had received only ketorolac). The difference was 1.0 unit (95% CI, 0.5–1.6). Afterward, pain intensity was similar in both groups (intergroup difference = 0.1 units; 95% CI, −0.4 to 0.2]) ().
Opioid requirements in each group are displayed in . The morphine group required more morphine to achieve the desired pain intensity levels than did the ketorolac–morphine group (). This estimate includes the 0.1-mg/kg dose given at the outset of the protocol for the morphine group. The intergroup difference in morphine requirements was 6.5 mg (95% CI, 5.8–7.2).
As morphine consumption increased, the risk of development of any side effects also increased (). For each 1-mg increment of morphine consumed, the risk of development of side effects increased 1.3% (95% CI, 1.3–1.4).
Sex Effect on Opioid Requirements and Side Effects
Women required more morphine than men to achieve a similar degree of pain relief. On average, women required 20% more morphine in milligrams per kilogram (95% CI, 8–34%). Women also had a 6% higher risk of development of side effects than men (relative risk = 1.06); however, this difference did not achieve statistical significance (95% CI, 0.97–1.16).
Side Effects at 30 Minutes of Evaluation (before Any Rescue Medication)
The proportion of patients who experienced any side effect by 30 min after the start of the initial analgesic infusion was 54% in the ketorolac–morphine group and 68% in the morphine group. Therefore, the difference in the proportion of subjects who exhibited any side effect was 14% (95% CI, 8–20), which corresponds to a number needed to harm for morphine of 7 (95% CI, 5–12).
Side Effects throughout the Study
The incidence and severity of side effects was higher in the morphine group. The proportions of patients who experienced any side effect were 63% in the ketorolac–morphine group and 74% in the morphine group, so the absolute risk difference for development of any side effect was 11% lower in the ketorolac–morphine group than in the morphine group (95% CI, 5–16) (). This difference corresponds to a number needed to harm for morphine compared with the combination of ketorolac–morphine of 9 (95% CI, 6–20).
Looking at each side effect, the morphine group had a higher incidence of sedation, dizziness, and pruritus. The occurrence of nausea/vomiting was similar in both groups ().
We found that morphine is a more efficacious analgesic than ketorolac. Our results showed that the NNT of ketorolac compared with morphine is 5, meaning that of 5 patients treated with ketorolac, 1 did not obtain adequate pain relief who would have experienced it had he or she received morphine.
How can we explain that morphine provided superior analgesia compared with ketorolac despite the fact that published NNTs for morphine and NSAIDs suggest similar analgesic efficacy? First, ketorolac and morphine were not directly compared in earlier calculations of NNT.3,,31,,32 Arithmetically, the NNT is the inverse of the absolute risk reduction; therefore, changes in the baseline risk yield different NNTs. In analgesic studies, the intensity of baseline pain could be analogous to the baseline risk. We found that if pain is severe, the probability of having pain relief is lower than when pain is moderate. Therefore, systematic differences in the baseline pain intensity among studies that evaluated NSAIDs and opioids could explain the discrepancy.
Second, the sample size of the trials included in the meta-analysis could also be a contributor. The calculation of the NNT for ketorolac was based on a total of only 176 patients. Discrepancies in the results between small and larger trials are not uncommon34,,35 and can be explained by inclusion of patients with diverse baseline risks of experiencing the event of interest34,,36 or by the uneven allocation of these subjects in the small trials.36,,37 In the current study, we directly compared ketorolac and morphine, and included more than 1,000 subjects.
Third, the moment of evaluation also could explain why morphine was more efficacious than ketorolac. We evaluated pain intensity and pain relief 30 min after analgesic administration, not 4–6 h, as the studies that served for the NNT calculations did. Although the onset of action of ketorolac and morphine is similar, ketorolac has a longer half-life, 5–6 h, than morphine, 2–3 h. Last, we administered the medication intravenously, and many studies from which the NNTs were obtained used the intramuscular route. The evaluation at 30 min and the intravenous administration of analgesics are more appropriate for the treatment of acute postoperative pain of moderate or severe magnitude.
Various studies have compared ketorolac with strong opioids with conflicting results. The findings range from ketorolac being less effective than strong opioids5,,38,,39,,40,,41,,42,,43,,44,,45,,46,,47 to being equally effective13,,14,,15,,16,,18,,48,,49,,50,,51,,52,,53,,54 or to being more effective than opioids.55,,56,,57,,58
The studies that have reported no difference in the degree of analgesia have at most one fifth the number of subjects we studied, so differences between ketorolac and morphine could have been missed; in addition, very small doses of opioids (2–4 mg morphine or morphine equivalent) were used.59,,60,,61,,62 Of the studies in which ketorolac was more effective than morphine, two evaluated analgesia for renal colic.55,,56 Opioids produce tonic contraction of smooth muscle and so may have a spasmogenic effect on ureteric activity, possibly by inhibiting the release of nitric oxide. On the contrary, NSAIDs have a spasmolytic effect and therefore may be a superior choice for this kind of pain. The study that found ketorolac to be superior to morphine included a total of 119 patients undergoing major surgery and allocated to three groups (10 and 30 mg ketorolac, and morphine). It is difficult to reconcile these varied findings with those of the current study.
The superior analgesic effect of morphine in our trial was balanced by an increased occurrence of side effects. The number needed to harm of morphine at 30 min of observation indicates that for seven patients treated with morphine, one will experience an adverse effect that would not have occurred if that patient had received ketorolac.
Nonetheless, to limit analgesics in the ambulatory setting only to NSAIDs, as has been suggested, could result in inadequate pain relief, and pain by itself may produce nausea. We found that adding an opioid to an NSAID regimen, if pain is not adequately controlled, led to a much lower opioid consumption. Although some authors have suggested that the opioid-sparing effect of NSAIDs does not have clinical importance,67,,68 we found that the reduction in opioid requirements was marked: 4.5 mg versus 11.1 mg. The confidence interval of this estimate indicates that the smallest decrease in opioid requirements that one could expect would be 48%, which is still a considerable reduction. A very recent systematic review that evaluated the morphine-sparing effect of acetaminophen combined with morphine patient-controlled analgesia found a 20% (9-mg) reduction in morphine requirement in 24 h.
As the dose of morphine increased, the risk of side effects increased, a finding that echoes many (but not all) of the conclusions of a recent meta-analysis by Marret et al. that assessed the morphine-sparing effect of NSAIDs in 22 trials enrolling 2,307 patients. Other research largely confirms these findings and likewise supports the existence of a dose–response relation for postoperative opioid adverse effects. Therefore, it is not surprising that the reduction in opioid requirements was associated with a clinically important reduction in opioid side effects. Such a finding is a principal benefit of multimodal analgesia, which is now advocated in numerous clinical practice guidelines on acute pain management.
We found that the combination of ketorolac and morphine decreased the occurrence of opioid-related side effects, including pruritus. However, we did not find any difference in the incidence of nausea/vomiting between the groups. This last finding contrasts with the meta-analysis by Marret et al. Possible explanations for this difference are the short window of observation and that our patients were not yet mobilized. It is believed that movement increases the likelihood of nausea/vomiting after opioid exposure. In addition, patients received antiemetic therapy if, at arrival in the postanesthesia care unit, they reported moderate or severe nausea or were vomiting. The prescription of antiemetics could have hindered detection of any difference, but we, as others, considered it inappropriate to withhold routine symptomatic treatment because patients were participating in the study.73,,74
We found that women required more morphine to achieve a similar degree of analgesia than did men. This result confirms previous findings. However, women did not exhibit more side effects than men, as has also been reported. Still, the confidence interval of our estimate is wide, and so we cannot rule out the presence of sex differences.
The duration of follow-up in the current study was short, and this limitation deserves discussion. We do not know whether the magnitude of the opioid-sparing effect and the decrease in opioid side effects are persistent. In addition, we did not consider the toxicity of repeated NSAID administration, such as gastrointestinal toxicity, operative site bleeding, or renal dysfunction, which are major clinical concerns during postoperative pain management.
Although the use of broad inclusion criteria and unrestricted use of anesthetic techniques in this trial could be criticized, we see this inclusiveness as a strength. We designed a pragmatic trial that emulated the conditions in which the results would be applied. A trial with these characteristics assures the generalizability and usefulness of the results, and it does so without compromising scientific rigor. The randomization, masked allocation, and large sample size of this study assure group balance so that risk factors for greater or lesser analgesic responses or for development of side effects are similarly distributed and therefore would not bias the results.
In summary, clinicians should be aware of the superior analgesic properties of morphine compared with ketorolac, despite the fact that historic data for these two drugs yield similar NNTs. There is a trade-off, however, in that one of seven patients treated with morphine will experience an adverse effect that would not have occurred if that patient had received ketorolac. The addition of a strong opioid to an NSAID decreases the risk of opioid side effects in the early postoperative period due to a considerable reduction in opioid requirements. Given the safety of short-term (less than 5 days) administration of NSAIDs such as ketorolac after major operations, we encourage the use of this combination when there is no contraindication to NSAID therapy.