Bowen ME, Poehling KA, Gunn VL, Carroll KN, Callahan ST, Arbogast PG, Cooper WO. Does the Use of Oxycodone Affect Diagnostic Accuracy in Children With Acute Abdominal Pain?. Arch Pediatr Adolesc Med. 2005;159(4):326-328. doi:10.1001/archpedi.159.4.326
DIMITRI A.CHRISTAKISMD, MPHHAROLD P.LEHMANNMD, PhD
Historically, opioid analgesia has been thought to mask symptoms, alter physical exam findings, delay diagnosis, and increase morbidity and mortality for patients with acute abdominal pain of unclear etiology.1 However, studies in children and adults have demonstrated that administering intravenous opioids to patients with acute abdominal pain does not delay diagnosis or adversely affect diagnostic accuracy.1,2 In this issue of ARCHIVES, Kokki et al3 present a randomized, double-blind, placebo-controlled clinical trial examining the consequences of buccal delivery of oxycodone on pain relief and diagnostic accuracy in children evaluated in the emergency department (ED) with moderate to severe acute abdominal pain. Using the Users’ Guide to the Medical Literature,4,5 we reviewed this study with regard to the validity of results, size and precision of the treatment effect, and generalizability and applicability of outcomes.
Was the Study Randomized? Yes, but more details about randomization would allow the reader to determine whether each participant had an equal chance of being assigned to the oxycodone or placebo group. The goal of randomization is to produce comparable study groups with respect to known and unknown risk factors.6 Valid randomization requires that an independent person or center develop the randomization process and allocate participants to the appropriate study group. Although the article says, “Participants were randomized into 2 groups by a computer-generated allocation sequence,”3 it provides insufficient details on the timing and logistics of randomization. In particular, it is unclear when randomization occurred in relation to arrival at the ED and baseline surgical assessment.
To assess randomization, we compared the study groups with respect to important observed patient characteristics provided in Table 1. We found that the oxycodone and placebo groups were comparable based on these characteristics. Thus, the 2 groups were relatively balanced with respect to known prognostic factors.
Were All Patients Who Entered the Trial Accounted for and Attributed at Its Conclusion? It appears so. The inclusion criteria were clearly defined and uniformly applied to eligible children prior to randomization. All 63 enrolled patients contributed to analysis of sensitivity, specificity, and diagnostic accuracy. However, it is unclear whether all patients contributed a predose and postdose pain measurement in the calculation of pain scores. This lack of clarity is important because excluding observed outcomes from data analysis can lead to biased results of unknown magnitude or direction.
Were Patients Analyzed in the Groups to Which They Were Randomized? Yes. An intention to treat analysis compares patients according to original randomization assignments regardless of satisfaction of entry criteria, treatment received, withdrawal from the study, or deviation from the protocol.
This approach minimized overestimation of clinical effectiveness and maintained randomization by assuring that the 2 groups remained similar apart from random variation.7 Figure 2 and Table 2 indicate that patients were appropriately analyzed with respect to randomization.
Was Follow-up Complete? No. In all clinical trials, loss to follow-up and missing data are inevitable. Because both can limit the validity and generalizability of the study, researchers should specify measures a priori to minimize expected loss to follow-up and clearly outline a strategy to deal with missing data.8 A detailed description of the outcome of interest and a clearly defined follow-up period are essential. The article specifically describes a fixed follow-up period for pain assessment, beginning with the baseline assessment and ending 3.5 hours after the first dose of study medication. However, it does not specify outcomes such as surgery that would result in termination of follow-up prior to 3.5 hours. The article does not describe an analytic strategy to account for these missing data points. The follow-up period for determining clinical outcomes extended up to 4 weeks after enrollment but only for those discharged without a surgical diagnosis. Failure to explicitly define the time frame for diagnosing surgical and nonsurgical diseases and failure to outline criteria for terminating the study precluded complete follow-up.
Aside From the Experimental Intervention, Were the Groups Treated Equally? Yes. Accurate assessment of an intervention requires that all patients enrolled in the treatment and control group be treated identically aside from the intervention being tested. The authors used uniform methods to obtain provisional diagnoses, surgical dispositions, and pain assessments in patients assigned to oxycodone and placebo groups.
Were Patients, Clinicians, and Study Personnel “Blind” to Treatment? Perhaps. To minimize bias, participants and investigators should be blinded to the identity of the assigned intervention group. In this study, a nurse who did not assess participant outcomes assigned patients to oxycodone or placebo groups in accordance with the randomization scheme and prepared oxycodone and placebo solutions with similar physical characteristics. Because analgesic effect was a primary outcome, the potential lack of blinding must be considered. The study did not assess the ability of enrolled children to identify their treatment group. However, surgeons were asked to guess the treatment group of a subset of patients. This procedure, although failing to demonstrate a significant difference in the surgeons’ abilities to discern group assignment, demonstrated incomplete surgeon blinding because the surgeons correctly identified placebo recipients 75% of the time.
How Large Was the Treatment Effect? The analgesic effectiveness of buccal oxycodone and placebo was measured by comparing predose and postdose pain scores using a visual analog scale.
To account for differences in baseline pain, pain scores were converted into maximal pain intensity difference (PID) and summed pain intensity difference (SPID). The PID reflects peak analgesic effect, and the SPID is the area under the time-pain reduction curve.9 This study found that buccal oxycodone, but not placebo, was associated with a significant decrease in SPID 3.5 hours after the first administered dose. In contrast, PID scores for oxycodone and placebo were not significantly different. However, the clinically significant level of pain reduction, as measured by SPID or PID, was not discussed. Lee et al10 reported that a 3-cm reduction in PID, as measured on a visual analog scale, is the minimum clinically significant difference in pain control. Farrar et al11 indicated that a 33% reduction in percent SPID (SPID/maximum possible SPID) was clinically relevant. However, these calculations cannot be performed with the data provided. At both predose and postdose assessments, no significant differences in sensitivity, specificity, and diagnostic accuracy were found between oxycodone and placebo groups. We were unable to reproduce Table 2 from the data provided in Figure 2. It is not clearly stated (1) when patients were assigned to the observation or laparotomy group and (2) when they were classified as having a surgical or nonsurgical disease. However, calculations of sensitivity, specificity, and diagnostic accuracy based on postdose assessments are not significantly altered by reassigning patients using different time frames for observation vs laparotomy group assignment or surgical vs nonsurgical disease group assignment.
How Precise Was the Estimate of the Treatment Effect? The precision of pain scores, sensitivity, specificity, and diagnostic accuracy can be assessed by examining the 95% confidence intervals around the estimated value.
The 95% confidence interval for the difference in mean PID score between the 2 groups is relatively narrow, indicating a small degree of variation between the 2 groups. Although there is a significant difference in the mean differences of SPID score between the 2 groups, the 95% confidence interval around the estimate is large, indicating considerable variation in SPID between the 2 groups. Based on the confidence interval, the true mean difference in SPID could be as little as 2 cm. If such is the case, the clinician must determine whether this small amount of pain reduction after 3.5 hours is clinically significant and worth the potential risks of giving oxycodone.
Were All Clinically Important Outcomes Considered? During study design, it is important to specify a priori the clinically important outcomes of interest, the parameters that define them, and the manner in which the outcomes will be measured. The more specific the outcome, definition, and measurement technique, the more accurate the findings are. This study examined multiple outcomes, including the effect of buccal oxycodone vs placebo on (1) pain reduction, (2) physical examination findings, (3) diagnostic accuracy, and (4) clinical outcomes.
Previous research has demonstrated that opioids are effective in reducing acute abdominal pain.1,2,12 In evaluating the effectiveness of buccal oxycodone to relieve acute abdominal pain, investigators needed to demonstrate that the medication had a clinically significant reduction in pain score. Although the article reports that the study was designed to have 80% power“ to detect a 2-fold difference in SPID between the 2 groups at a 0.05 level of significance,”3 insufficient data are provided to reproduce this power calculation.
Among the outcomes investigated, the effect of oxycodone on diagnostic accuracy is the most clinically relevant end point. Despite the title, the study was unfortunately not powered to detect differences in sensitivity, specificity, or diagnostic accuracy between the 2 treatment arms. Using the postdose diagnostic accuracy data (88% for oxycodone, 84% for placebo, type I error = 0.05, power = 80%), 1180 patients in each treatment group would be required to detect a 4% difference in diagnostic accuracy. A power analysis using the same data and a 1-arm sample size of 32 demonstrated that the study had only 7% power to detect a 4% difference in diagnostic accuracy. Thus, the study appears to have been insufficiently powered to allow an accurate interpretation of results with respect to sensitivity, specificity, and diagnostic accuracy.
The adverse effect profile of oxycodone in acute abdominal pain also merits further investigation. Although only 2 adverse effects—1 headache and 1 urticarial reaction—were observed, no conclusions on the adverse effect or safety profile can be drawn from the data provided.
Are the Likely Treatment Benefits Worth the Potential Harm and Costs? Uncertain. The study concluded that buccal oxycodone provides pain relief in children with acute abdominal pain. However, because the study was underpowered, it did not provide enough evidence to evaluate whether or not children are harmed by changes in diagnostic accuracy resulting from oxycodone administration.
Are the Results Applicable to My Practice? Before the results of this study can be safely applied to clinical practice, further investigation is needed. The management of acute abdominal pain varies between institutions based on treatment protocols, the availability of diagnostic imaging, the expertise of radiologists, and the level of coordination between surgeons and ED physicians. Many institutions routinely use imaging in their assessment of the acute abdomen when the clinical signs, symptoms, and exam are equivocal. Consequently, many of these institutions routinely administer opioid analgesia to children with acute abdominal pain after the initial evaluation, even if the diagnosis remains in question. In children, abdominal ultrasound can have greater than 95% sensitivity, at least 85% specificity, and a positive predictive value and overall accuracy in excess of 90% in diagnosing acute appendicitis.13 In children, computed tomography scans have greater than 90% sensitivity, 85% to 90% specificity, and a positive predictive value and overall accuracy in excess of 95% with either a standard or appendiceal protocol to diagnose acute appendicitis.13
The generalizability of this study was further challenged by a heterogeneous study population. Because it defined acute abdominal pain as “pain of less than 7 days’ duration,”3 patients with both acute and subacute or resolving abdominal pain were included in the study. Furthermore, “most of the children had a primary care referral.”3 Patients who are referred to the ED are systematically different from those who come to the ED without referral. Approximately 50% of children came to the ED with surgical abdomens, which enriched the study population with truly acute abdomens and potentially introduced bias because surgeons might have been aware of the referral status of a patient prior toexamination.
This randomized, double-blind, placebo-controlled clinical trial investigating the efficacy of buccal oxycodone on pain management and its impact on diagnostic accuracy in acute abdominal pain did not provide sufficient evidence to change current clinical practice, but it did help in identifying opportunities for further study. Adequately powered studies are needed to examine the effects of analgesia on clinical outcomes in children with acute abdominal pain. Further investigations should also address the most appropriate analgesic agent, dose of medication, and route of administration.
Correspondence: Mr Bowen, Child and Adolescent Health Research Unit, Department of Pediatrics, Vanderbilt University School of Medicine, AA-0216 MCN, Nashville, TN 37232 (firstname.lastname@example.org).
Acknowledgment: We are grateful to Timothy Givens, MD, Division of Pediatric Emergency Medicine, Vanderbilt Children’s Hospital (Nashville, Tenn), who provided clinical insight into pediatric emergency medicine practice as we prepared the journal club discussion.