Wednesday, January 16, 2013

Opioid Dose & Automobile Driver Injuries

Data MiningThe latest data-mining study from Canadian researchers claims there is a significant relationship between prescribed opioid analgesic dose and the risk of injurious automobile crashes among drivers taking these medications. However, this may be another example of low-quality evidence designed to further disparage opioid prescribing for chronic noncancer pain.

Effects of prescription medications, and particularly opioid analgesics, possibly associated with motor vehicle accidents have been of much concern. Opioids may interfere with attention and delay reaction time, impairing driving performance; however, in controlled-study settings these effects have not been demonstrated as significant, and influences of particular opioid doses have not been adequately examined.

Tara Gomes, MHSc and colleagues at the University of Toronto, Ontario, Canada, constructed a retrospective investigation to assess the relationship between opioid dose and risk of “road trauma” among patients in a community-based population [Gomes et al. 2013]. In this context, road trauma broadly refers to patients who suffered injuries due to automobile mishaps as drivers, passengers, pedestrians, or other circumstances (eg, bicyclists or unknown). However, a primary focus of the study was on drivers taking opioids and injured in crashes.

The research consisted of a case-control study of patients aged 18 to 64 years old who were eligible for prescription drug coverage under the Ontario Provincial Public Drug Program from April 1, 2003 through March 31, 2011 and were prescribed opioid analgesics. Relevant records retrieved from the database included patients who had received at least 1 publicly-funded prescription for oral codeine, morphine, oxycodone, or hydromorphone, or transdermal fentanyl during the 8-year timeframe.

Cases were defined as opioid-prescribed patients who had an emergency department visit related to road trauma. Serving as controls were patients who had been prescribed opioids, but did not experience a road trauma incident — they were closely matched 1-to-1 to cases by age, sex, index year of starting opioids, road trauma incidents prior to the study timeframe, and a disease risk index (taking into account medical or psychiatric conditions).

All opioid dosages were converted to daily morphine equivalent (MEQ) dose amounts, and the researchers compared the risk of road trauma among patients receiving very low dose (<20 MEQ) amounts to those in 4 other opioid-dose categories: low (20-49 MEQ), moderate (50-99 MEQ), high (100-199 MEQ), and very high (≥200 MEQ).

Writing in an early online edition of JAMA Internal Medicine (formerly Archives of Internal Medicine), the researchers report that among 549,878 patients in the database receiving opioids during the 8-year period there were about 85,440 (15.5%) who had emergency department visits for road trauma. From this sample, the researchers identified 5,300 eligible cases for their study, who were matched to an equal number of controls. Data for all comparisons were statistically adjusted to account for a number of potential confounders, such as age, prior treatment for alcoholism, past medication use (including CNS depressant drugs), total number of drugs prescribed, and frequency of physician or emergency department visits in the past year.

Among all cases — eg, drivers, passengers, pedestrians, and others — analyses yielded no significant association between escalating opioid dose and odds of road trauma (adjusted Odds Ratios, or ORs, ranged between 1.00 and 1.09). Similarly, among new users of opioids (starting medication within 2 weeks prior to road trauma incident) there were no significant effects of opioid dose, as compared with longer term opioid users (adjusted OR 1.33; 95% Confidence Interval, 0.84-2.12).

However, looking specifically at automobile drivers — accounting for 2,428 incidents, or 46% of all cases — the researchers reported there were significant associations between opioid dose and road trauma. Compared with very low opioid doses, drivers prescribed low doses had a 21% increased odds of road trauma (adjusted OR 1.21; 95% CI, 1.02-1.42); for those prescribed moderate doses, there was a 29% increased odds (OR 1.29; 95% CI, 1.06-1.57); high doses had a 42% increased odds (OR 1.42; 95% CI, 1.15-1.76); and for very high doses a 23% increased odds (OR 1.23; 1.02-1.49).

Based on these data, Gomes et al. conclude that, among drivers prescribed opioids, a significant relationship exists between medication dose and risk of injurious accidents. They suggest that increasing opioid doses can impair drivers and amplify risks of road trauma, and public policy makers “could consider restricted drivers’ licenses for patients treatment with high-dose opioids.” However, there are many aspects of this research requiring closer scrutiny before those conclusions are accepted as valid.

COMMENTARY: As might be expected, the news media touted this study with headlines like, “Taking Opioids Linked to Increased Risk of Car Crashes” [news item here]. Fully accepting the premise of the journal article, news reporters noted that taking even low doses of opioids increase the risks of being involved in a “traumatic motor vehicle accident,” and the risk escalates as people take “higher amounts of the powerful pain killers.”

Many healthcare providers, and their patients, have wondered about the safety of driving while taking opioid medications at any dose. This study by Gomes and colleagues questions the prudence of driving while taking anything but very low opioid doses — which likely would be subtherapeutic for almost all patients with pain requiring opioid therapy in the first place — and appears to be further evidence against the appropriateness of long-term pain management with opioids.

A recently published survey of pain specialists by Benzon et al. [2013], described in an UPDATE [here], found that nearly 70% of those practitioners allowed their patients to drive automobiles when the opioid dose was stable. Furthermore, 85% of responding specialists believed there was no maximum opioid dose that would prohibit safe driving. However, the study authors recognized that this is a complex issue and a multitude of factors might affect driving competence in these patients, include the distraction of unresolved pain due to inadequate analgesia.

Benzon and colleagues discuss research finding that opioids do not normally impair psychomotor performance, and that driving abilities may improve once pain is properly managed with opioids. Although, patients first starting opioids or those receiving dose escalations of greater than 30% should be cautioned against driving until the dose is stable. Some research has found that patients taking daily opioid doses of 290 MEQ or more for at least 2 weeks do not have impaired driving ability; yet, unlike with alcohol, there is no standard for what levels of opioids in the blood might impair driving.

Unfortunately, when considering the daily MEQ of opioids that might affect safe driving, the study by Gomes et al. is unenlightening and even misleading for a number of reasons that are worth noting:

  • For starters, the sample population for this study might have been convenient, since records could be dredged from a large and readily-available electronic data warehouse, but the Ontario Drug Benefit database includes only adults younger than age 65, who are economically disadvantaged and/or unemployed, and may have disabilities and/or receive home care services. Surely this would not represent a sampling of typical drivers in the overall population and, in fact, the researchers could not determine the extent to which these patients had access to automobiles or their driving frequency.

  • For this study, qualifying cases of drivers involved in automobile accidents represented only 0.4% of all persons prescribed opioids during the 8-year study timeframe — a 0.05% annual incidence rate — although the authors do not indicate if this small accident rate is high or low compared with the rate in the general population. However, persons ≥65 years of age were excluded in the study, and this could be a population of great concern when it comes to the influence of any medication on driving safety.

  • In a complex algorithm, a patient’s total opioid exposure was defined as the sum of all opioid prescriptions — calculated as daily MEQ (morphine equivalent) doses — overlapping the date when the patient visited the emergency department for “road trauma.” However, there was no way of knowing if the patient was actually taking the opioid medication(s), or taking more or less of it than prescribed.

  • Additionally, the researchers could not determine from the records why the opioid medication was prescribed or whether the dose was adequate to relieve distracting pain that could have affected driving safety.

  • Gomes2013Of special importance, the data do not support a strong effect of opioids on injurious accident rates among drivers or differences in risk across dosing levels (see figure, adapted from Gomes et al. 2013).
     
    As the figure illustrates, the Odds Ratios for the 4 dosing levels of interest represent small effect sizes of questionable clinical significance [see prior UPDATE on effect sizes here]. Also, it is important to remember that the increased odds — ranging from 21% to 42% — are relative to those of very low dose opioids (<20 MEQ), and not in direct comparison with no opioids at all.

    Furthermore, the Confidence Intervals surrounding the 4 OR point estimates overlap to a large extent suggesting that there are no statistically significant differences between dose groups. Plus, the lower limits of the CIs hover perilously close to 1.0, the point of no effect that would denote statistical non-significance.

  • Paradoxically, the Odds Ratio and Confidence Interval for the highest-dose group (≥200 MEQ) are almost identical to those of the lowest dose group (20-49 MEQ). Gomes et al. try to explain this as possibly due to patients in the highest-dose group diverting their prescribed opioids, thus not actually belonging in that dose category; however, they also concede that physiologic tolerance in persons on long-term opioid therapy at a fixed high-dose “may offset the detrimental effects of these drugs on driver performance.” Of course, this latter possibility contradicts the whole premise of their study, which claims that higher doses are detrimental to safe driving.

  • It also seems of some consequence that among all new opioid users — drivers and others — there were no significant effects of opioid dose on their involvement in accidents. The first few weeks of opioid therapy, particularly at higher doses, might be of special concern regarding safety, but this was unverified in this study.

  • Finally, in their discussion of results, the researchers introduce “adverse selection” as possibly having influenced outcomes. This concept proposes that greater numbers of patients with substance abuse or mental health disorders are selected to receive long-term opioid therapy at higher doses and, presumably, they are more prone to driving accidents. The authors offer no data in support of this contention and mentioning adverse selection in this context appears to reflect some bias.

Assuredly, cautioning patients about safe driving practices while taking opioid medications — at whatever dose or duration of time — would be prudent medical practice. And, patients first starting opioids or increasing their dose might be advised to refrain from driving until they are stabilized on the dose and their pain is better relieved, which would depend on dose, type of opioid, and individual patient factors.

However, in addressing these concerns, this study by Gomes and colleagues represents a low quality of evidence that, at best, suggests hypotheses for further study — which is typical of data-mining research. Of some concern is the inherent bias implied by using evidence of dubious reliability and validity to challenge the effectiveness and safety of opioids for chronic pain.

Indeed, two of the article coauthors (Dhalla and Juurlink) were signers of a petition to the U.S. FDA from a group, Physicians for Responsible Opioid Prescribing (PROP), seeking changes to opioid labeling that would restrict drug companies from promoting chronic opioid therapy [see UPDATE on this here]. This was not disclosed in the journal article, and whether authors should reveal such an affiliation as a potential conflict of interest that could bias their perspectives might be considered.

It also is disconcerting that a highly-respected top-tier journal like JAMA Internal Medicine (formerly Archives of Internal Medicine) would seem to convey bias by not using its editorial pages to present a fairly-balanced critical appraisal of this research. Instead, an accompanying editorial, by Mitchell H. Katz, MD, lauds the Gomes et al. 2013 study as presenting “another good reason to avoid the prescription of high-dose opioids” [Katz 2013].

Katz also denounces “the epidemic of deaths due to prescription opioids” in his editorial and makes favorable reference to an earlier article by Gomes and essentially this same Canadian research team, appearing in 2011 in this same journal, which purported to demonstrate that increasing daily doses of opioids for noncancer pain results in higher mortality rates. This is interesting because an earlier Pain-Topics UPDATE [here] discussing that 2011 Gomes et al. study found many of the same flaws and fallacies as in this new study by these researchers.

For one thing, in their 2011 publication Gomes et al. had used the exact same Ontario Drug Benefit database, with all of its limitations, for their data-mining analyses. Opioid-related mortality incidents were poorly defined and confounded in many cases by the presence of other drugs. The researchers used the same opioid-dose categories as in this present study, but computations of dosing, without access to actual patient charts, was complex and imprecise; also, the same very low dose (<20 MEQ) was used as the reference point for comparisons.

Gomes2011Particularly noteworthy, the presentation of Odds Ratios and Confidence Intervals demonstrating increased mortality with increasing dose was remarkably similar to trends in the current study [see figure]. That is, there was considerable overlap in the CIs, suggesting a lack of statistically significant differences between groups and, in this case, the CI for the lowest dose group (20-49 mg) crosses the line of no effect (1.00) and was not statistically significant.

The relatively large OR for the highest-dose group (≥200 mg) appears to be of considerable consequence. However, our calculation from raw data of the absolute risk difference between cases and controls for this group was 0.10, representing a small effect size that challenges the influence of dose. Hence, it appears for all groups that, at the least, mortality was not associated with opioid dose alone and other factors may have been of more critical importance [see full discussion in the UPDATE here].

Regrettably, this earlier study by Gomes et al. (2011) presenting a low quality of evidence has been widely cited in arguments favoring more restricted prescribing of opioids for moderate to severe chronic noncancer pain — as in the Katz editorial and also in the PROP petition. The current article, addressing driving safety, will no doubt be added to the list of evidence, even though its reliability and validity are questionable.

Lastly, by way of full disclosure we offer the following:

  1. Our criticisms of the above research studies should not be construed as advocating for unbridled opioid prescribing for any pain condition, at any dose, or for any duration of time. Rather, our concern is with the interpretation of research from evidence-based medicine perspectives and the use of highest-quality evidence as proof in arriving at judgments regarding the effectiveness and safety of any therapy. Thus far, in our opinion, the quality of evidence and the standard of proof in the debate over opioid therapy for chronic noncancer pain have been inadequately low — more and better research are needed [see a discussion of proof in pain research in an UPDATE here].

  2. Pain Treatment Topics and these UPDATES are supported in part by educational grants from opioid analgesic manufacturers. In accordance with the strictest standards, these organizations have absolutely no input regarding the selection, development, review, or approval of any contents at Pain-Topics.org or these UPDATES. All perspectives and opinions expressed here are exclusively those of the author.

REFERENCES:
> Gomes T, Redelmeier DA, Juurlink DN, Dhalla IA, et al. Opioid Dose and Risk of Road Trauma in Canada: A Population-Based Study. JAMA Intern Med. 2013; online ahead of print [
abstract].
> Katz MH. Opioid Prescriptions for Chronic Nonmalignant Pain: Driving on a Dangerous Road [editorial]. JAMA Intern Med. 2013; online ahead of print [
access here].

eNotifications Don’t Miss Out. Stay Up-to-Date on Pain-Topics UPDATES!
Register [here] to receive a once-weekly e-Notification of new postings.

5 comments:

Anonymous said...

Well, at least they admitted there were study flaws. Although, they did downplay them. It definitely was not a rigorous or well designed experiment. They might as well have a test subject flip a coin and then correlate that to driving. I can see it now. People who flip and get heads are more likely to get in an accident. Everyday I am more and more shocked at the people who become doctors. It seems like they are just letting anybody in and handing them a degree if they have money. If they performed this experiment in medical school for a class, then they probably would have received an "F," from the professor. This "experiment," is garbage. The only thing it's fit for is toilet paper.

Anonymous said...

Here is a link for one particular article which is based on this study:

http://www.msnbc.msn.com/id/50473644/#.UP2gZCc8CSo

There are so many problems with this study. It is not even an experiment. It is just correlating two random events. It breaks one of the basic axioms of science; correlation does not equal causation. That is middle or high school level science. I don't know how some of these people became doctors or scientists.

The sad part is the media accepts everything they say without any substantial solid indisputable proof or asking questions about the design flaws of the "experiment."

It is possible for some scientists to be stupid or incompetent. The people who designed this study could have passed all their classes with "C's" or "D's" and we wouldn't know. It certainly seems like these "experiments," were designed by people who graduated at the bottom of their class.

Anonymous said...

I have quite a few questions about this dubious "experiment," or "study."

1) How do the researchers know who caused the accident? What if a person being treated with opioids was hit from behind? Or run over by a car in a cross walk?

2) Why did the study target opioids in particular? We know that there are many other medicines which can interfere with driving. Look at the people who take sleeping medicines like ambien. They cause a lot of accidents and nobody is creating hysteria over that.

3) How can the observers be absolutely certain that opioids caused these accidents or played a significant role in it? What if someone can drive perfectly while using opioid medicines, but gets distracted by a cell phone or didn't get enough sleep and then causes an accident?

4) There are too many variables to be accounted for. The "scientists," performing this study have made a lot of assumptions. In fact, I would say they made too many.

I can't believe a medical journal actually ran this article. The editor of JAMA should be fired. It would be pretty hard to get away with making this many assumptions in a math, chemistry , or biology journal.

SB. Leavitt, MA, PhD said...

Thank you, for the comments above. News reporters rarely have the time, inclination, or skills to assess the quality of the research evidence they are writing about. Sadly, they just want to attract readers, and sensational subject matter does that --- even when it presents misinformation.

As I noted in the UPDATE, there are many unanswered questions about the reasons for this study, its design and data interpretation, and its subsequent validity. However, that is not at all the same as questioning the competence or intelligence of the researchers themselves -- which I did NOT intend to imply.

Anonymous said...

I am aware of that and your article is clear. You never mentioned that nor was it implied. It is just my opinion.

If these doctors and scientists can retroactively correlate opioid medicines to car accidents in the manner that they did, then I can correlate their intelligence/competence with the flaws in their experiment.Without the basic axioms of logic and science holding me back, I can correlate almost any event with another. If these scientists and doctors are going to disregard these axioms, then so can I.

So, based on my experiment and its results, I conclude that the authors of the JAMA article are incompetent and not as intelligent as previously thought.