Background
Ketorolac tromethamine is a non-steroidal anti-inflammatory drug (NSAID) that received authorization for marketing at the beginning of the nineties for the short-term management of post-operative pain. Ketorolac is a derivative of pyrrolizine carboxylic acid and is structurally related to tolmetin and zomepirac. The tromethamine salt enhances its water solubility making the drug fairly appropriate for parenteral administration (1).
In its first years of marketing, ketorolac was aggressively promoted as a safer alternative to opioid analgesics, overlooking its NSAID nature (2). This led to a sharp increase in its use but reports describing serious NSAID-type adverse reactions did not take long to appear. At the end of 1992 the manufacturer presented a safety update report to the regulatory authorities which raised concerns about its gastrointestinal, renal and hematological toxicity under the actual conditions of use (3,4).
By that time, the interim results of a postmarketing surveillance study (henceforth referred to as the Strom study) were also known. The results of this study and the spontaneous reports prompted the Committee for Proprietary Medicinal Products (CPMP), the scientific body of the European Medicines Evaluation Agency, to re-evaluate the benefit-risk balance of the product. As a result, more restrictive conditions of use were set forth (5). Furthermore, the company was urged to undertake a study in Europe aimed to compare the safety profile of ketorolac with other NSAIDs used for the same indication (the so-called POINT study). Some countries, however, considered the restrictions insufficient and decided either to suspend the license or to stop the authorization process (5,6).
In 1994 the CPMP ratified the restrictions (6). Further to that decision two population-based case-control studies were published, both pointing to a much higher risk of GI toxicity of ketorolac than that of other NSAIDs. Meanwhile, in 1997 the Steering Committee of the POINT study asked the CPMP for permission to discontinue the study because the incidence rates of adverse events were much lower than expected. Indeed, the differences were so small that this was viewed by some as a suggestion that the safety hazards of ketorolac had been over-emphasized.
A reappraisal seems timely. Firstly, I will describe in some detail
the data that gave rise to the regulatory measures taken on ketorolac in
1993-1994. Secondly, I will analyze the epidemiological evidence that came
out afterwards. Finally, I will try to draw some conclusions.
Clinical pharmacology of ketorolac
Ketorolac, like other NSAIDs, has analgesic, antipyretic and anti-inflammatory activity. These three actions derive from a common mechanism: the inhibition of the cyclooxygenase, a critical enzyme in the biosynthesis of prostaglandins, prostacyclin and thromboxanes. Prostaglandins released in response to trauma mediate pain by sensitizing afferent nerve fibers. NSAIDs act as analgesics mainly by attenuating this peripheral sensitization, although other mechanisms including an action on central nervous system have also been suggested (7). In animal models ketorolac has exhibited an analgesic/anti-inflammatory ratio greater than other NSAIDs. For instance, weight by weight ketorolac proved to be 50 times more potent than naproxen in analgesia models but only 3 times more potent in inflammation models (1). This remarkable dissociation between analgesic and anti-inflammatory effects was the basis for the development of the drug as an analgesic in clinical settings (8).
In clinical trials carried out in an ample variety of post-operative situations, IM ketorolac 10 to 30 mg has convincingly showed an analgesic efficacy comparable to the standard doses of opiates (IM morphine 6-12 mg, pethidine 50-100 mg). In most studies the daily dose of 120 mg was not exceeded and the duration of treatment was usually five days or less (7). The major advantages of ketorolac over opiates in post-operative pain are the more rapid return to normal GI function, less sedation, nausea and vomiting, and a probable shorter hospital stay which, it has been claimed, may compensate for its greater cost (7). Moreover, the analgesic effect of ketorolac seems to last longer than that of opiates and the risk of respiratory depression is remarkably diminished. As with other peripheral analgesics, ketorolac may not adequately control severe pain. Ketorolac may have a most valuable use as a sparing-opioid therapy decreasing the requirements of opiates in 25-50%. The oral bioavailability has been reported to be between 80 to 100% with respect to IM or IV routes, making the oral formulations a reasonable alternative as long as the oral route is feasible. Lower doses appear to be recommended for oral formulations, which lacks a pharmacological basis unless their application is reserved for situations of less severe pain.
Although the use of ketorolac has been proposed for other indications (e.g. renal colic, musculo-skeletal pain) and other populations (e.g. children), approval has been hampered by the concerns about its safety.
Adverse effects in premarketing trials
Data from premarketing clinical trials in about 5,000 patients exposed to ketorolac showed the expected pattern of adverse events for an NSAID, including GI disorders, renal and liver dysfunction, anaphylactic reactions and operative-site bleeding. Overall, no indication of a higher frequency or severity of these events with ketorolac was appreciated at the time of authorization (7). Nonetheless, in isolated clinical trials ketorolac did show a worrisome pattern, though at large doses. In an endoscopic study carried out in healthy volunteers, Lanza et al (9) showed that an invasive gastric antral ulceration was apparent in 4 out of 5 subjects receiving IM ketorolac 90 mg 4 times daily for 5 days. The degree of mucosal injury was dose-dependent and at 10 to 30 mg 4 times a day ketorolac compared favorably with aspirin 650 mg 4 times a day.
Spontaneous reporting data
In December, 1992 the manufacturer informed that about 16 million people had received the drug and a total of 923 reports of adverse events had been collected, 73 of them resulting in death. Reports referred to gastrointestinal (n=203), hematological (n=181, including operative site bleeding) and renal (n=124) reactions. Thirty four deaths were associated with gastrointestinal bleeding and/or perforation; of these 21 received ketorolac by parenteral route and a daily dose of 120 mg had been used in 50%; three deaths occurred after a single dose, 3 after one day and 10 within the first 5 days of treatment.
By December, 1993 the total number of reports had increased to 1,718 and the number of fatalities to 143. The worldwide exposure was estimated to be about 41 million people. The adverse event pattern remained essentially the same as before. Most patients among those reported with GI bleeding, perforation or ulceration were elderly and similar proportions were on oral and parenteral route. Interestingly, about 16% of patients with GI bleeding had been concomitantly treated with other NSAIDs.
Postmarketing surveillance studies
In 1993, the European regulatory authorities were informed that three postmarketing surveillance (PMS) studies were under way. One was carrying out in Germany involving near 30,000 patients treated either with oral or parenteral ketorolac, but the method employed and data provided were both so limited that it was considered hard to draw any conclusion. The two other studies were under way in the US, one of them conducted in hospitals (the already mentioned Strom study) and the third one was using a Medicaid database. No information about the latter was further provided and the present analysis will be focused only in the hospital-based study.
The company presented the final report of the Strom study in January, 1994. Strikingly, it would not be published until two years later (10). According to the authors, the study was conceived before ketorolac was marketed and focused on two major outcomes: GI bleeding and operative site bleeding, which were expected to be the major risks of the drug as compared with opiates, the standard therapy for post-operative pain, and the natural control for this study.
By design it was a retrospective closed cohort study, conducted in 35
hospitals in the Philadelphia region. Data collection began on November
18, 1991 and ended on August 31, 1993. Patients receiving parenteral ketorolac
were identified by pharmacy records. As a reference group, patients treated
with parenteral opiates subsequently matched to ketorolac patients by hospital,
admitting service and date of initiation therapy were selected. Trained
nurses that were not blinded to the exposure of interest abstracted data
from medical records. A patient was considered exposed to ketorolac from
the first dose through the third day after the final dose. If more than
3 days elapsed between consecutive doses, then two different courses were
considered. Treatment courses and not patients or patient-days were considered
in this study as the unit of analysis for all calculations.
No operational definition of GI bleeding was given. The abstracters
were instructed to record “any occurrence of the adverse event of
interest” avoiding clinical judgment. A GI bleeding was considered as “clinically
serious” if caused death, residual disability, prolonged hospitalization
or was life-threatening (so defined as documented hypotension, hemoglobin
levels less than 80 g/L or requiring transfusion of at least 4 U within
1 week of the start of bleeding). An intermediate category, “clinically
relevant bleeding” was also defined but unfortunately results by this category
were not given.
Data were analyzed by logistic regression models adjusting for 17 potential confounders. Dose-response analyses were performed for average daily dose and duration of treatment taking the entire opiate group as reference. Results were shown for two broad categories of outcomes: all events (which included microscopic bleeding) and clinically serious events. This latter category was better defined and the results are thus more reliable. Therefore, only information on that category will be provided unless otherwise stated.
The adjusted OR (95% CI) for clinically serious GI bleeding was 1.44 (1.09-1.89). The risk increased with age, so that people older than 74 years old presented an adjusted OR (95%CI) of 1.72 (1.05-2.81). A clear relationship with dose was found, the highest risk was for daily doses greater than 120 mg (OR: 7.34, CI95%: 3.71-14.5), although an increased risk was already apparent with doses between 60 and 75 mg a day (OR: 1.79; 95%CI: 0.98-3.2). A relationship was also observed between duration of treatment and the risk of serious GI bleeding. After 5 days of treatment the OR (95%CI) was 2.72 (1.26-6.10) whereas within the first five days the OR (95%CI) was 1.26 (0.93-1.69). People younger than 65 years of age, using the drug for at most 5 days and in doses at most 105 mg/day had only a small increase in risk (OR: 1.22; 95%CI: 0.72-2.06).
In terms of attributable risk for the hospitalized population, ketorolac would be responsible for 11 additional cases of GI bleeding, 4 of them serious, for every 1,000 patients treated with ketorolac instead of opiates only. Most of these cases would be elderly or people treated with high doses (>105 mg) or prolonged duration (>5 days).
Despite its limitations (mainly the lack of operational definitions of outcomes, the unblinded technique used for abstracting the medical records, and the possibility of residual confounding by unknown factors) this study provided a scientific foundation for the actions that regulatory authorities were to take. It was also an example of how a PMS study should be done: anticipating potential problems and trying to answer specific and clinically relevant questions.
The 1993/1994 regulatory actions
On 5 February, 1993 the Netherlands requested an official opinion of the CPMP concerning the safety of ketorolac. On March 16th, a hearing with the company was held and on June 16th the opinion was released (5). It was acknowledged that a final conclusion would not be possible until the company provided complete information, in particular the final report of the Strom study. Meanwhile, the summary of product characteristics of ketorolac would be changed emphasizing its contraindications, restricting the indication to short-term management of post-operative pain and including restrictions on dose and duration of treatment. The starting dose for IM/IV administration was reduced to 10 mg and the maximum daily dose to 90 mg for non-elderly and 60 mg for the elderly. The duration of treatment for IM/IV ketorolac was reduced to 2 days. The oral doses remained unchanged with the same maximum daily dose and the maximum duration of treatment was now established in 7 days.
Despite the restrictions and the efforts to reach a common position in the European Union, a consensus among regulatory authorities proved elusive. The company ceased supply in Germany on June 14th, 1993, forced by the regulatory agency, and the license was suspended in France on December 12th, 1993. On 14-15 March, 1994 the CPMP reviewed the issue again. The company had provided additional information, including the final report of the Strom study. No new restrictions were considered necessary. In pharmacovigilance opinion no.15 the CPMP concluded: “1) The Committee considers that the medicinal product has a narrow therapeutic margin; 2) Belgium, Denmark, Spain, Italy, Luxembourg and the United Kingdom consider that the benefit/risk ratio remains favorable, under the restricted conditions of use as set out in the Committee’s opinion of 16.6.1993; 3) Ireland cannot authorize the product until the results of a comparative study on efficacy and safety has been submitted and reviewed; 4) Germany, Greece, France, the Netherlands and Portugal consider that the restrictions are not sufficient and that the benefit/risk ratio is unfavorable, based on currently available data; 5) The company has indicated an intention to perform additional comparative safety studies. The Committee requests the company to submit, for review, the protocol for these studies and to submit the results to all Member States” (6).
In the USA the reaction was more moderate. In October, 1993 the labeling of ketorolac was modified to include the new findings but the indications and dose remained unchanged and the duration of parenteral treatment was restricted to 5 days. Later on, in 1995 the maximum recommended dose be reduced to 120 mg for IM/IV and 40 mg for oral therapy. For patients 65 years or older, for those who weigh less than 110 lb. or have renal impairment, the maximum daily dose was reduced to 60 mg (11).
The new epidemiologic evidence
Two case-control studies specifically aimed at quantifying the risk of GI toxicity associated with ketorolac appeared in 1995 and 1998, respectively. Both had in common the country where they were undertaken: Italy, but otherwise differed in many aspects: design, region, period of time, outcomes and research teams. Both, however, captured essentially the same picture.
The first study was done by Traversa et al (12) in residents of the province of Rome during 1991-1992. Cases were selected from a consecutive listing of patients who had undergone esophagogastroduodenoscopy in a public hospital of the province, and were defined as outpatients with a first-time endoscopic diagnosis of an ulcer or erosion that could be unambiguously matched to the registry of individual beneficiary codes in order to retrieve data on prescriptions (this was possible in 74% of cases). Controls were selected in two stages: first a 10% random sample of all residents were drawn and then 10 controls matched on exact age and sex were drawn for each case. Exposure was imperfectly defined because in the computerized file only was recorded the month and not the day of prescriptions. Thus, current use was considered when a prescription was recorded in the same month of the endoscopy or in the month before. Non-use was defined when no prescription was recorded in the 6 months before the month of prescription. Information about exposure to analgesics, corticosteroids, cytostatic agents, antacids, antiulcer therapy and any other drug during the six months previous to the month of endoscopy were evaluated as potential confounders or effect modifiers.
The prescription of any NSAID during the current-user time window was associated with a 30% increase in the risk of GI lesions as compared to non-use (OR: 1.3; 95%CI: 0.98-1.8). Ketorolac showed a distinctly elevated risk (OR: 4.2; 95%CI: 1.9-9.4) as compared to that of all other NSAIDs combined (OR: 1.2; 95%CI: 0.9-1.7). The increase in risk associated with ketorolac was even more dramatic for ulcers (OR: 10.5; 95%CI: 3.0-37.1) than for erosions. In most cases ketorolac was dispensed for short duration (6 days or less).
The most apparent limitation of this study was the inability to specify the date of prescription. However, as the authors appropriately discussed, this could only lead to an underestimation of the true effect. A confounding by indication was also a reasonable concern if the physicians have prescribed ketorolac to a different kind of patients than other NSAIDs. However, for that confounding to explain the results it would have been necessary that the indication of ketorolac patients had a 10-fold elevation in ulcer risk and that ketorolac were only used in that condition (12). Neither of those assumptions seems reasonable. Hence, the results strongly support the hypothesis that ketorolac is a particularly toxic NSAID for the GI tract.
The second study was undertaken by García Rodríguez et al (13) in the region of Friuli-Venezia-Giulia, Italy. Its main objective was to quantify the risk of hospitalization due to Upper GI bleeding (UGIB) associated with NSAIDs. The source population included permanent residents aged 25 to 89 years, during the period 1991-1995, who had not had a previous hospitalization for conditions that may cause bleeding. Cases arising from this population were identified by specific ICD-9 codes and validated by manual review of hospital charts when ICD-9 codes proved not to have a high predictive value according to a previous validation study. The date of admission was taken as the index date. A total of 1,505 patients with UGIB were selected. Controls were a random sample of 20,000 persons drawn from the source population with a random date assigned to each one as the index date. A current NSAID user was defined when the supply of the most recent prescription lasted until the index date. A person was a non-user when no prescription of a NSAID was recorded in the 150 days before the index date. Estimates of rate ratios for individual NSAIDs were adjusted by age, sex and calendar year. In addition, the analysis for individual drugs was restricted to the subset of persons without ulcer antecedents (defined as past use of acid-suppressing drugs and/or hospitalization for ulcer disease).
The current use of any NSAID increased the UGIB risk by a factor of 4 as compared to non-use. By individual drugs the risk ranged from 0 for dipyrone to 24.7 for ketorolac (95%CI: 9.6-63.5). Only piroxicam, other than ketorolac, showed a point estimate higher than 6 (RR: 9.5; 95%CI: 6.5-13.8). As compared to other NSAIDs combined, ketorolac showed a striking RR of 5.5 (95%CI: 2.1-14.4). The risk was higher when ketorolac was used as IM formulation (RR: 28.3; 95%CI: 8.7-92.0), for periods longer than 7 days (RR: 30.3; 95%CI: 8.5-107.5) and on doses higher than 20 mg (RR: 28.1; 95%CI: 8.7-90.9), confirming those factors are important determinants of risk. Moreover, a new and worrisome finding was provided by this study: the excess of risk of UGIB associated with ketorolac was still higher when used in doses as low as 20 mg or less (RR: 20.0, 95%CI: 4.3-93.6) or for periods of time as short as 7 days or less (RR: 19.1; CI: 4.6-79.2), that is, when used under the restrictions of use adopted since 1993. Again, it can be speculated that patients treated with ketorolac might have been high-risk patients for UGIB, but the authors took special care in restricting the analysis to patients without ulcer antecedents and, in addition, they were able to confirm from the attending physicians of ketorolac users (10 cases and 7 controls) that the indication of ketorolac was indistinguishable from that of other NSAIDs (most osteoarthritis and chronic pain).
The POINT study
In the hearing held in March, 1993 the manufacturer of ketorolac committed to the CPMP to conduct a study in Europe aimed to compare the safety of ketorolac with that of other NSAIDs also used for short-term management of post-operative pain. The study (called Post-Operative Injectable NSAID Trial, POINT) was conceived as a randomized clinical trial with three arms, ketorolac, diclofenac and ketoprofen. The trial was to be done in hospital inpatients requiring analgesia after planned surgery. The treatment would be given initially by injections, and then continued, as necessary, by oral route. The main endpoint was a composite of death, post-operative bleeding and GI bleeding. The trial started in July, 1995, and intended to recruit 16,000 patients from 8 countries.
It was anticipated in the protocol of the study that an interim analysis would be performed at certain points of recruitment to examine if major deviations in the safety endpoints between the groups have been produced. In February, 1997 the company, following the recommendations of the Steering Committee of the study, requested permission from the CPMP to discontinue the trial. The argument was that in the interim analysis of the first 4,901 patients the incidence of the composite endpoint was 1.3%, much lower than the one anticipated, based on the Strom study (3.7%). With such incidence the power of the study to detect a RR of 1.3 had declined to 47%, and a sample of 23,500 would have been needed to maintain the power in 90%. This unexpectedly low incidence would indicate that either the previous studies had grossly over-estimated the risk of ketorolac or the POINT study had been undertaken in a highly selected population with a very low risk of GI bleeding.
In June, 1997 the CPMP gave the green light for study discontinuation. As of this writing, the Steering Committee is preparing the final report of the study with the results of 11,352 patients finally recruited.
Conclusions
The long chapter of GI toxicity of ketorolac has not concluded yet. The new epidemiological data has confirmed that the potential of this drug to cause GI lesions and bleeding is several times higher than that of other NSAIDs. The risk when ketorolac is used outside the currently authorized conditions is unacceptably high and other alternatives are clearly preferred. Nor is it clear that within the labeled conditions ketorolac is a safe drug. The information we have up to now is worrisome but probably insufficient to draw a conclusion with reasonable confidence and new studies are clearly warranted. The results of the POINT study, although reassuring at first glance, should be interpreted with caution.
References
1. Buckley MMT, Brogden RN. Ketorolac- a review of its pharmacodynamic and pharmacokinetic properties and therapeutic potential. Drugs 1990; 39: 86-109.
2. Ariano RE, Zelenitsky SA. Ketorolac (Toradol): a marketing phenomenon. Can Med Assoc J 1993; 148: 1686-1688.
3. Choo V, Lewis S. Ketorolac doses reduced. Lancet 1993; 342: 109.
4. Anonymous. New ketorolac dosage recommendations. Pharm J 1993; 251: 49.
5. Committee for Proprietary Medicinal Products. Opinion pursuant to article 12 Directive 75/319/EEC on ketorolac i.m./i.v. (Toradol), 16 June, 1993.
6. Committee for Proprietary Medicinal Products. Pharmacovigilance opinion no. 15 on ketorolac i.m./i.v. (Toradol), 16 march, 1994.
7. Gillis JC, Brogden RN. Ketorolac- a reappraisal of its pharmacodynamic and pharmacokinetic properties and therapeutic use in pain management. Drugs 1997; 53: 139-188.
8. Spickler W. Ketorolac (Toradol): a new analgesic or an old NSAID? Can Med Assoc J 1993; 148: 1693-1695.
9. Lanza FL, Karlin DA, Yee JP. A double-blind placebo controlled endoscopic study comparing the mucosal injury seen with an orally and parenterally administered new nonsteroidal analgesic ketorolac tromethamine at therapeutic and subtherapeutic doses. Abstract. Am J Gastroenterol 1987; 82: 939.
10. Strom BL, Berlin JA, Kinman JL et al. Parenteral ketorolac and risk of gastrointestinal and operative site bleeding. A postmarketing surveillance study. JAMA 1996; 275: 376-382.
11. Anonymous. New safety recommendations issued for ketorolac. Am J Health-Syst Pharm 1995; 52: 1035.
12. Traversa G, Walker AM, Ippolito FM, et al. Gastroduodenal toxicity of different nonsteroidal antiinflammatory drugs. Epidemiology 1995; 6: 49-54.
13. Garcia Rodriguez LA, Cattaruzzi C, Tronconi MG, Agostinis L. Risk
of hospitalizations for upper gastrointestinal tract bleeding associated
with ketorolac, other nonsteroidal anti-inflammatory drugs, calcium antagonists,
and other antihypertensive drugs. Arch Intern Med 1998; 158: 33-39.