Reduction of Postburn Hyperalgesia after Local Injection of Ketorolac in Healthy Volunteers

The experimental protocol was approved by the Human Investigations Committee of Yale University School of Medicine. Healthy volunteer subjects were evaluated for history of hypersensitivity to NSAIDs, bleeding disorders, ulcers, allergies, nasal polyps, asthma, cardiac irregularities, and ingestion of NSAIDs within the previous 7 days. The protocol was explained to subjects both verbally and by written informed consent. Volunteers were shown the contact thermal stimulator and invited to test the various heat stimuli before agreeing to participate.

The experimental protocol involved a single session consisting of three thermal testing phases at each of six sites (Figure 1). With the subject sitting in a room with an ambient temperature of 23–25 degrees Celsius, the volar forearms were exposed and immobilized in a vacuum-molded bean bag. Test sites were chosen approximately 5 cm apart and marked on the skin (three on each arm). They were numbered (distal to proximal) 1, 2, and 3 on the "bradykinin model" forearm, and 4, 5, and 6 on the "burn model" forearm.

In each of the testing phases, a series of nine successive test stimuli (5-s duration, applied at 25-s intervals) was delivered to each test site with a 1-cm x 1-cm contact Peltier thermal stimulator (constructed by the Medical Instrumentation Facility of Yale University School of Medicine). [35]Thus, a total of 54 stimuli were delivered during a given phase. The desired temperature was servo-controlled via electronic circuitry and feedback from a thermocouple located at the interface between the thermal stimulator and the skin. A base temperature of 38 degrees C was maintained for 1 min before the first stimulus and for 23 s between subsequent stimuli. The temperature of each test stimulus was reached at the end of 1 s with a linear ramp increase up to the desired value, maintained for 5 s, and then decreased within 1 s to the base temperature. Before each thermal testing sequence, a 47 degrees Celsius stimulus was delivered to eliminate the impact of surprise on magnitude estimations; the response to this stimulus was not recorded. The remaining nine stimuli consisted of a computer-generated random sequence using each integer temperature between 43 degrees C and 51 degrees Celsius, including a new 47 degrees C stimulus. A given site received the same thermal test sequence in every phase of the study (baseline, after pretreatment, and after injury;Figure 1).

In addition to the pain induced by the thermal testing sequences, the subjects were exposed to the pain of the pretreatment injection (ketorolac or saline) and the pain of the actual injury (bradykinin injury or burn injury). The measurements of pain and the testing procedures are detailed later.

Before initiation of the study phases, subjects were familiarized with the method of magnitude estimation of pain [36,37]during delivery of sample stimuli at a remote site on the dorsal forearm. They were told to assign the number zero to nonpainful sensations and to assign a positive numeric value to each painful stimulus, such that the relative values assigned to different stimuli would be proportional to the subject's estimate of their relative pain intensities.

As originally suggested by Stevens, [37]each subject was instructed to call the first painful stimulus any number that seemed appropriate and to then assign numbers to subsequent painful stimuli in such a way that they reflected the relative degree of perceived pain. Consistent with other applications of magnitude estimation, which normalized values within each subject to facilitate intersubject comparisons of experimental conditions, [36–40]all scores above zero in a given subject were normalized to the subject's maximum estimate (i.e., highest pain score) evoked by the stimuli presented to the untreated, uninjured control site (site 4). The normalized scores were used for all comparisons of pain intensity during the thermal testing.

In addition to recording the degree of pain, we also determined the pain threshold at each site. This was defined as the lowest temperature at which pain was felt at that site, unless no pain was felt at the next higher temperature; in that case, the second lowest temperature at which pain was felt was considered to be the pain threshold.

The sequence of testing phases is illustrated in Figure 1and detailed below.

Phase 1. Baseline Thermal Testing. Before pretreatment or injury, each site was tested with its prescribed thermal sequence; and the pain threshold and normalized pain scores at that site were recorded.

Phase 2: Pretreatment and Repeat Thermal Testing. After baseline responses were determined, sites 2 and 3 were pretreated by intradermal injection of 10 micro liter of either 0.9% preservative-free saline or 0.3 mg of ketorolac tromethamine (Toradol, Syntex Laboratories, Palo Alto, CA). These pretreatment solutions were randomly assigned in a double-blind fashion so that one site received saline and the other received ketorolac. Sites 5 and 6 were likewise randomly assigned to the saline and ketorolac pretreatments. Because there was no obvious precedent for determining the volume of pretreatment injectant in these human volunteer models, we selected volumes such that, when equal volumes of pretreatment and bradykinin were injected, the final dose and volume of bradykinin would be consistent with prior studies in which bradykinin was administered alone. [5–7,11].

Subjects were asked to rate the acute pain induced by each injection at 20-s intervals for 60 s using the method of magnitude estimation described earlier. These values were normalized to the mean score in response to 51 degrees C at site 4 during the three phases of thermal testing. The cumulative normalized pain score of each subject was determined as the sum of the four interval pain scores in that subject. Beginning 2 min after injection of saline or ketorolac at a given site, the effect of pretreatment on thermal testing was determined by repeating the thermal testing sequence at each site (Figure 1).

Phase 3: Injury and Postinjury Thermal Testing. Sites 2 and 3 were injected with 10 nmol of bradykinin in 10 micro liter normal saline; site 1, which did not receive pretreatment with ketorolac or saline, was injected with 10 nmol bradykinin in 20 micro liter saline (to achieve same total volume of injectant as sites 2 and 3). Subjects were asked to rate the pain experienced at the time of injection, and every 5 s thereafter for 2 min. As for the pain of ketorolac and saline injections, these values were normalized to the mean score in response to 51 degrees C at site 4 during the three phases of study.

On the burn model arm, a mild burn injury was produced at sites 5 and 6 by applying the thermal stimulator at 50 degrees C for 100 s. Site 4 was left as an untreated, unburned control site; burning at this site would have induced an injury different than that at a site pretreated with an injection of liquid. Subjects were asked to rate the pain of the 100-s burn injury at 20-s intervals throughout the burn period. The cumulative normalized pain score in each subject was determined as the sum of the normalized serial pain scores for the 100-s injury in that subject.

Postinjury sensitivity after bradykinin injection at sites 1–3 was evaluated by thermal testing 3–5 min after injection (Figure 1). The brief interval was chosen in light of the rapid onset and brief duration of bradykinin-induced hyperalgesia. [3,4,6,11]Postinjury sensitivity after the local burn at sites 5 and 6 was evaluated by thermal testing 10–15 min after injury, consistent with the interval that has been shown to allow postburn hyperalgesia to develop. [3,4,8].

Analysis of variance and repeated-measures analysis of variance were used to compare the different sites with respect to: the threshold temperatures for pain during thermal testing, the normalized pain scores over the entire range of temperatures during thermal testing, and the normalized pain scores over the anticipated range of maximum postinjury hyperalgesia (i.e., for temperatures 1–4 degrees C, inclusively, above the postinjury pain threshold). [3,4,8]Unless otherwise stated, the P values reported were provided by these tests. A paired t test was used for comparison of the cumulative scores at ketorolac- and saline-treated sites during injection of pretreatment and during injury; the values for these comparisons were expressed as mean difference and 95% confidence intervals. A P value less than 0.05 was considered to be statistically significant for all analyses.

All subjects completed the entire study. There were no petechiae or hematomata at the sites of injection, and no blister formation or scarring at sites of burn injury.

During baseline thermal testing, the three sites on each arm responded similarly with respect to threshold and overall pain response (P > 0.2 for differences among sites on the bradykinin-model arm, P > 0.5 on the burn-model arm); this is illustrated for the burn model arm in Figure 2. Although the pretreatment injection with ketorolac was significantly more painful than the injection with saline on each arm (P < 0.01 by paired t test;-1.06, -1.63 to -0.49), pretreatment with ketorolac did not alter the pain evoked by subsequent testing with thermal stimuli (phase 2) as compared to the response at the corresponding saline-treated site (P = 0.3 on the bradykinin model arm, P = 0.9 on the burn model arm by analysis of variance); nor was the response to phase 2 thermal testing at the ketorolac-treated site significantly different from the response at the untreated control site (P = 0.7 on bradykinin model arm, P = 0.2 on the burn model arm;Figure 3).

Burning caused acute pain at both ketorolac sites and saline sites. The cumulative pain score during burning at the ketorolac site differed from that during burning at the saline site by less than 3%(P = 0.8 by paired t test; 0.13, -1.02 to 1.29), consistent with reports that NSAIDs do not effectively ameliorate the acute pain of burn injury. [10]Burning also induced postinjury changes in pain threshold and pain intensity in response to thermal testing. As detailed later, pretreatment with ketorolac did not prevent the decline in threshold but did decrease the intensity of the hyperalgesic response (compared to pretreatment with saline).

After burn injury, the reduction in pain threshold in response to the successive thermal stimuli was similar at both ketorolac sites and saline sites. The postinjury (phase 3) thresholds at ketorolac sites and saline sites were 44.5+/-0.8 degree C and 44.8+/- 2.0 degrees C, respectively; they did not differ significantly from one another, but each was significantly lower than the threshold at the untreated, unburned control site (47.2+/-2.0 degrees C).

The most dramatic effect of pretreatment with ketorolac was that it blunted the increase in pain intensity after burn injury. As illustrated in Figure 4, postburn hyperalgesia was more pronounced at saline sites. The overall pain in response to postinjury thermal testing (across all temperatures) at the ketorolac sites and saline sites did not differ significantly (P = 0.08). However, for the range encompassing the anticipated hyperalgesic response (46–49 degrees C, inclusively), the pain scores at the ketorolac site were significantly lower than at the saline site (P < 0.05, analysis of variance). At the more extreme temperatures (e.g., 50 degrees C and 51 degrees C), there was no hyperalgesia and a benefit of pretreatment was not apparent.

The cumulative pain during and for 120 s after bradykinin injection was 24% less at the ketorolac-treated site than at the saline-treated site (P = 0.2 by paired t test;-1.6, -4.2 to 1.0). However, the bradykinin-model arm was not useful for testing the effect of ketorolac on postinjury hyperalgesia: although bradykinin induced significant hyperalgesia when injected into a site without pretreatment (site 1), it did not cause hyperalgesia at sites pretreated with ketorolac or saline. Over the entire range of temperatures, the pain in response to postinjury thermal testing (phase 3) at the saline site was only 3% greater than that during baseline and did not differ significantly from the pain induced by phase 3 testing at the ketorolac-treated site (P = 0.25) or at the untreated, uninjured control site (P = 0.46).

A major objective of the current study was to adapt models that induce hyperalgesia after bradykinin or burn injury for testing the effects of pretreatment by intradermal injection. To our knowledge, this is the first investigation in which a local intradermal injection was administered before injury with either the bradykinin injury or burn injury model. Adaptation of a burn injury model provided hyperalgesia, which enabled documentation of a beneficial effect of pretreatment with an intradermal injection of as little as 10 micro liter ketorolac (0.3 mg). Adaptation of the bradykinin injury model for this purpose was unsuccessful: the dose and volume of bradykinin used in the current study, while capable of inducing hyperalgesia when used alone, failed to induce hyperalgesia after pretreatment with either ketorolac or saline. This may have been caused by dilution of the bradykinin by the injectant in the pretreatment bleb, as well as by the local reactions that may have been initiated by the pretreatment injections. Fortunately, this did not bias our comparison of ketorolac versus saline (because both sites received a 10-micro liter injection). Further pursuance of the bradykinin model should include the evaluation of increased doses and volumes of bradykinin or the use of a route of pretreatment that does not dilute the subsequent injection.

To the best of our knowledge, the current findings are the first to document the local effect of a low dose of an NSAID in humans. The effects of ketorolac were selective, in that ketorolac significantly reduced postinjury hyperalgesia in response to the mid-range stimuli where hyperalgesia was most pronounced. It did not affect the postinjury response to the stimuli that were most painful before burning; nor did it lessen the response to thermal testing in the absence of burning or the acute pain that accompanied the burn injury. In addition, the data suggest that the two components of postburn hyperalgesia may occur via two different mechanisms, in that pretreatment with ketorolac significantly reduced the postburn increase in pain intensity but did not reduce the postburn decline in pain threshold.

The incomplete analgesia after pretreatment with ketorolac in the current investigation is consistent with reports that NSAIDs do not necessarily provide complete analgesia, [15,25,26,31,41–46]regardless of the route of administration. Consistent with the belief that several mediators play a role in inflammatory hyperalgesia, even ketorolac's ability to prevent postinjury hyperalgesia was incomplete. However, it is likely that pretreatment with ketorolac reduced postburn hyperalgesia by more than the 15% suggested by comparison of the saline sites and ketorolac sites in Figure 4. The data represent the aggregate effect of hyperalgesia and the pain that thermal stimulation would have induced even without a prior burn injury. Subtracting the pain at the unburned control site from that at the burned sites (saline site and ketorolac site) would provide a better indication of the hyperalgesic component. When this is done, the pain scores at the ketorolac site become 22% less than those at the saline site over the range of temperatures between 46 degrees C and 49 degrees C. We believe it is likely that this still leads to underestimation of the effectiveness of ketorolac against the intensity of the hyperalgesic component, because a given temperature at a burned site would represent a greater increment above threshold than would the same temperature at the control site (and hence cause more pain). Thus, the full antihyperalgesic effect of ketorolac was difficult to quantify; nevertheless, the decrease was statistically significant.

The current study employed magnitude estimation of pain rather than a categorical or interval scale. Magnitude estimation differs from the more commonly used visual analog scale in that the former is a floating score not restricted by anchors such as a built-in ceiling. This is important when one is delivering a stimulus that is much more painful than a stimulus that had been assigned a high visual analog score earlier in a sequence (or delivering one that is much less painful than one which had been assigned a very low score). For example, if a subject reported a score of 6 on a 1–10 visual analog scale, then it would be impossible for that subject to accurately score a subsequent stimulus that was twice as painful. Therefore, methods such as magnitude estimation, which yield ratios rather than interval scales, are preferable for documenting gradations of sensation. [47]Magnitude estimation has been used to monitor the response not only to thermal stimuli, [5,7,48,49]but also to a wide variety of stimuli, including mechanical, [38,50]chemical, [39,51]and electrical [40,48]stimulation of the skin. In normal hairy skin, nociceptor discharge accounts for the magnitude of brief heat pain [52]; there is a linear correlation between mean nociceptor discharge and the corresponding pain rating. [52]Alternatively, the visual analog scale, which initially was validated by comparison with a scaling procedure such as magnitude estimation, [47]is a simpler test that is better suited for clinical settings that are less amenable to normalization to a control site and controlled comparisons of different stimuli.

The major potential benefit of local administration of ketorolac is that it may provide analgesia while minimizing the effects of NSAIDs on the renal, gastrointestinal, and coagulation systems. These effects are of greatest concern after long-term administration; however, they may occur even after relatively short-term perioperative use. [15,53,54]While the small doses employed in the current study should mitigate systemic effects, it remains to be confirmed that small doses can be used successfully for more widespread infiltration, without exacerbating the potential for postoperative bleeding at the site of infiltration. Evidence suggesting the safety of NSAID infiltration may be found in recent reports that noted no increase in bleeding in patients receiving typical systemic doses (e.g., 30–60 mg) of ketorolac for local ketorolac infusion after mastectomy, [34]local infiltration before gynecologic surgery, [33]intraarticular injection after arthroscopic knee surgery, [32]or intravenous regional anesthesia. [55,56]There was no evidence of petechiae or hematomata in the current investigation.

Although the results suggest that intradermal injection of ketorolac reduced postburn hyperalgesia, it should be noted that the only available preparation of ketorolac contains 10% alcohol and that injection of ketorolac was more painful than injection of saline. The alcohol-containing preparation theoretically could damage nerve endings in the skin. However, this is unlikely, because responses to thermal testing (phase 2) and to the burn itself were not diminished by pretreatment. In addition, the lack of a difference between ketorolac sites and saline sites during phase 3 testing after bradykinin injection argues against the likelihood that the alcohol had a significant effect.

Thus, the current investigation has successfully adapted a burn injury model to provide a minimally invasive, low morbidity means for evaluating the effects of intradermal pretreatment with an injectable analgesic. Using this model, we have confirmed that local infiltration with ketorolac, in a dose less than 1% of that used systemically, can blunt subsequent postinjury hyperalgesia. Consistent with the selective effect when ketorolac was administered intrathecally (in that it did not affect the acute pain of formalin injection into the paw but did reduce subsequent hyperalgesia), [31,43]the effect in the periphery was selective; pretreatment with ketorolac did not reduce the acute pain of burning but did reduce subsequent hyperalgesia. An appreciation of this selective action may assist in the selection of drugs and routes for use in combination therapy. In addition to providing the basis for a potentially safer and more effective route of NSAID administration, the current model also provides a means for distinguishing the local and systemic effects of NSAIDs and other analgesic agents.

*Maehara K: The behaviorof anti-inflammatory drugs on the pain induced by scald. J Osaka Dent Univ 1968; 2:41–68.