Red Flags Uber Alles
Red Flags Uber Alles; Alex DeLuca; War on Doctors/Pain Crisis blog of the Pain Relief Network; 2007-08-06.
See also:
Red Flags and the Standard of Care; DeLuca; 2007
Interpretation of Aberrant Drug-Related Behaviors; Fisher; 2004
Reuters UK reports on recent research on substance use disorders in a population of chronic pain patients receiving daily opioid analgesic medications for at least 3 months. The subjects were 801 patients from 235 different family practices with “severe, incapacitating, [non-cancer] pain.” The study was NIDA supported and the research team lead by Dr. Michael Fleming.
Thirty patients (3.8 percent) met the clinical criteria for opioid use disorder, compared with a rate of 0.9 percent reported in the general population, the researchers report in The Journal of Pain.
The most significant predictor of abuse was aberrant drug behavior, including deliberate oversedation, feeling intoxicated by the medication, using opioids for reasons other than pain, and raising the dose without authorization.1
I have not been able to get a hold of full text of the published study yet, but how was it established that they “met criteria” for opioid use disorder (aka “addiction”)? This was, after all, a multi-center study; Dr. Fleming did not himself do diagnostic interviews on all 800 patients. So how was addiction status determined? Purely on the basis of Aberrant Drug-Related Behaviors (ADRBs)? Or on the basis of the family doc’s diagnosis in the chart? Or did all the patients get addiction medicine consults and a time-consuming Addiction Severity Index test? (I will make the full text of the Fleming study available here when I obtain it).
The elephant-in-the-living-room question is “what percent met clinical criteria for successful titration of medication to satisfactory analgesic effect,” but NIDA et al haven’t ever bothered to define the phenomena of “good pain relief” nor invented a method of measuring it.
Further, how can you measure adverse consequences of, say, drug seeking behavior, when the consequences of not seeking drugs is not examined or taken into account?
I fear this is one of those, frankly poorly designed, studies that don’t actually tell us very much of anything useful, but which will be proclaimed far and wide as “proving” that “addiction” is “far more common in chronic pain patients that previously realized.”
Footnotes
Tags: addiction, adrb, analgesic, behavior, chronic pain, disorder, drug abuse, drug-related, michael fleming, opioid, pain, pain management, pain relief, prescription drug abuse, red flag, research, statistics, study, undertreated, use
Comment by Bill
3.8% still seems very low in the event this study is reliable. The drug warriors would tell you, behind closed doors, that all 801 patients are “addicts” or “abusers” just because they want opioids.
Comment by James Stacks
Proportion tests against a population estimate are easy to bump with low base rates, because you can’t show that the population estimate is valid for the sample (e.g. you can’t show that the sample is a random sample of the population the estimate came from — the “generals”). Were these patients a random sample of those generals? Did this include five-star generals, or just one-star and two-star generals? How can chronic pain patients be compared to a general population? If the underlying suggestion is that the opioid therapy caused the “higher” addiction measures, then we need to look real close at that “general population” estimate that looks like it may be serving in place of controls. Maybe the full text will tell us more.
Comment by James Stacks
I am sure such a journal would not allow ex post facto data to be interpreted as causal, but I am not sure I see the relevance of the result to the question of opioid pain therapy unless the underlying suggestion of causality is swallowed. If there are more “addicts” among pain patients (but no causal link), then what does that have to do with pain treatment for the other 771 people? For that matter, what does the fact that a pain patient is an addict have to do with whether their pain is treated or not? If the causal link is not assumed, then are we afraid that they will become addicted addicts? If the causal link is assumed, that is bad science.
Comment by James Stacks
http://tinyurl.com/ynsns7
“considering the potential benefit to improving the lives of patients with chronic pain, a 3.8-percent rate of opioid addiction is a small risk compared with the alternative of continuous pain and suffering.” ( from http://tinyurl.com/yqh8ga )
Apparently, causality is assumed because there is a “risk” from therapy, although the authors apparently are on our side, and do not intend the result to be used for opioid fear mongering.
Comment by James Stacks
“This study found that the frequency of opioid use disorders was 4 times higher in patients receiving opioid therapy compared with general population samples (3.8% vs 0.9%). The study also provides quantitative data linking aberrant drug behaviors to opioid use disorders.” ( from: http://tinyurl.com/yrocew)
It’s one for you, thirty-four for me! Remember, the generals have one. And I’ll advocate for that one. Their pain should be treated. But the news is spreading fast: Now the 3.8% is a “High Frequency of Opioid Use Disorders…” ( http://tinyurl.com/2f9pde ). Indeed…absolutely astronomical! Did anyone mention that “The rate of toxicology tests positive for illicit drug use was 24%”?
The wonders of statistics abound: “…and 4 aberrant drug behaviors (OR = 11.48; 6.13 to 21.48). The final model for opioid use disorders was limited to aberrant behaviors (OR = 48.27; 13.63 to 171.04) as the other variables dropped out of the model.”
( from http://tinyurl.com/yrocew )
It would be funny if the “4 aberrant behaviors” turned out to actually be the DSM-IV criteria. Surely not. The abstract doesn’t say how the logistic regression model was developed. Usually, it is either sequential or statistical, but since they use the term “dropped out” it may have been some variation on statistical (stepwise). Hopefully not (see Thompson, B. (1989). Why won’t stepwise methods die? Measurement and Evaluation in Counseling and Development, 21(4), 146-148). It very well could have been a more advanced method for generating models. The full text should illuminate. If it was stepwise, that would mean the other predictors overlapped with the aberrant behaviors, but the aberrant behaviors were favored because of higher raw initial association, which would make a lot of sense if the behaviors were tied closely to the DSM-IV criteria. That fact in no way makes them better explanatory variables, though. If it was sequential, then that means the researcher just liked the behaviors better than the other variables. The competing variables that “dropped out” were the usual suspects — that would be 18-30 years old, positive for cocaine or marijuana, psychiatric disorder history, etc. That doesn’t even address living in Wisconsin, or what kinds of generals there are in Wisconsin. Are there military bases there?
I can’t get to this on any of my databases, so I’ll reserve further comment until I can read it!
Comment by doctordeluca
Thank you very much, Bill and James for the discussion of this study — even in the absence of much information about it. I would love to join this Commentary, and I will, but tonight I’m busy ‘formatting the wealth’, so to speak.
Because our cup runneth over. Not only has our small journal club, here, finally gotten a copy of the paper to study, but I’ve also discovered that a Continuing Medical Education (CME) course has already, in August, been created and put on sale based on a paper published in July. Wow! Weird. More about the CME in my next blog post. For now, here is the cite, link to abstract html page which has link to full text pdf:
http://www.doctordeluca.com/Library/Pain/ SudInChronicPainOpioidRx07.htm
Substance Use Disorders in a Primary Care Sample Receiving Daily Opioid Therapy. Fleming, M.F.; Balousek, S.L.; Klessig, C.L.; Mundt, M.P.; Brown, D.P.; Journal of Pain; 8(7): 573-582; 2007-07.
Dig in, Friends, enjoy! I’ll give it a careful read after I finish posting all this stuff. I think there are interesting flaws and some interesting data and I will be interested to get your takes on the ‘instant CME’ phenomenon, too.
..alex…