“…Subtle choices have big consequences..”


Subtle choices have big consequences!!! The below link hosted by Stanford Medical School November 2022 with Dr. Naburun Dasgupta informative lecture on the falsehood of the Morphine Milligram Equivalent and calls into question the Prosecution of healthcare providers by the US Attorney General and the Department of Justice and DEA in the use of Narcotic Analgesic Medications (Opioids)



  • All right. Here are some results. So here are the California results: First, same data, same conversion factors, but definitions D, One and D. Four, which are both used in clinical practice showed a threefold difference in who is perceived to be high dose,
  • adding in the forward results, we see a similar story with D four really identifying a lot more patients as high dose than the other definite other three definitions at some point i’m going to get tired of using the air quotes, but I hope i’m in a friendly audience that understands that. I’m not saying that ninety Mme. Is like a high dose threshold, but just using it, because that’s what the standard has been in the literature
  • it’s worth, pointing out that the definitions perform differently in each State, while D three returned the fewest high dose patients in California. It was d one that was the lowest in Florida.

“...Overall, the new recommendations sacrifice accuracy for a fabricated sense of clarity…”

Stanford Pain Relief Innovations Lab Speaker Series – Dr. Nabarun Dasgupta


“..that subtle choices have big consequences..”

Stanford Pain Relief Innovations Lab Speaker Series – Dr. Nabarun Dasgupta


  • user avatarDr. Nabarun Dasgupta02:50Thank you very much, Dr. Darnell, and thank you so much for having me. I’m sorry I couldn’t be there in person today. But i’m glad you all are on the zoom with me.
  • So uh, as Dr. Darnell said, uh, I work at the opioid Data lab at Unc. And it’s a collaboration between Unc. The University of Kentucky and the University of Florida, and we have our studies are divided into three groups. So we have theory, practice, and lived experience, theory uh studies are Ep. Methods and bio statistic methods uh practice our practice of medicine, pharmacy, and medical examiners,
  • and then um! The lived experience. Ones are most fun which involve uh research that is initiated by community members, whether they are pain patients or people, use drugs as well as qualitative and other kind of field studies. So today i’m going to focus on one of our theory projects which is looking at the definitions of how many?
  • So what if we have been unwittingly calculating daily me in different ways, but never realized it. And what happens if those differences are not based on conversion tables. But the way we calculate, mmes
  • I’ve been doing. I’ve been using mmes and research for about sixteen years, and during that like decade and half. I’ve had a really hard time convincing colleagues that subtle choices have big consequences. So we initiated this research to set things straight today. I’m going to show you something that once you see it, you can never unsee it,
  • something that will fundamentally change how you view daily Mme. And the ninety Mme. Per day threshold
  • uh funding disclosures are here. Uh no industry. Funding is probably the most important thing here.
  • Uh and this uh this research was funded by Fda and the Department of Justice. Uh, but the views expressed are not necessarily theirs.
  • Uh, the all the data that have that are being presented have been published. And so you can look up this paper. The Doi is there. Uh, we also have code, statistical code and software code equations, all of that to uh to really nerd out to when you feel like it.
  • Uh, I want to point out, to begin with, that we think about Mme. Conversion tables as like these pharmacological truths, but in reality the you know, there’s a lot of people who’ve pointed out the difficulties in understanding Mme. And some of the myths around them, including the late Dr. Feuden.
  • And uh, I think these are some of the better uh sources for this, so please go look them up. But the truth is is that enemies were devised like decades ago.
  • Ah, ah! Out of a series of small randomized trials in a hospital setting where people were being converted from one opioid analogy to another. They were not. They were based on the outcomes and the the conversion factors were not based on pharmacology, but they were based on analgesic effect.
  • What has happened in the interceding few decades is that the same concept of of of what was causing equal analgesic effect has been adopted to also represent the toxicological phenomenon associated with larger doses of these medications, so that even though we think of them as pharmacological properties, they’re not. They’re based on pain. And so when we use them in studies of overdose mortality and other
  • toxicology, there’s a big assumption that’s being made. That’s all not always made public, and i’m guilty of using them in those studies as well, and I will fully cop to that, and show you more about that in just a moment
  • all right. The purpose of this presentation today is to show you something new. We reveal that there are even more
  • uh
  • influential is shilly andversion factors. The problem is that there are actually four different ways to calculate enemy per day. But these differences have been entirely overlooked in the scientific literature.
  • A solution we offer is a clearer understanding of how to calculate daily Mme. And how to choose between the definitions
  • daily. Mme. Has been enshrined to law in at least fourteen States, which makes the assumption that daily Mme. Is a standardized clinical metric. But Mme. Is not a standardized clinical metric.
  • None of the laws that have been published. Define how to calculate Mme. Per day, because they already assume that this is a clinically standardized metric which it’s not.
  • Um so. But there is some face validity to the metric uh to the conversion factors. So we did a study years ago. On looking at the street prices for prescription drugs we pulled from a bunch of different sources. I’ll summarize briefly here in the in the second, in the first column there with the means. That’s the price per milligram. So if we took those prices per milligram and standardized to morphine at at one, the predictive relative potencies of base just on street price
  • are very close to what you would get from a equ analogies at conversion table. I think we were using the Vas one in this, but the other, you know you can see that they’re pretty similar. Um! In terms of what the street price demand for diverted opioids is,
  • you know. Does that mean that it has analysic effect of at these relative levels. Does it mean that there’s toxicology at these relative levels? That’s a different story. But the point is is that one of the reasons why the the concept of Mmes has endured is because there is some face validity in the clinic on the street and maybe in the toxicology.
  • Okay, to start with, let’s imagine that we give four analysts the same data set and ask them to identify, which are the quote high-dose, opioid-handledes of patients the same time period, same location, same patient same prescription same identical data set.
  • We can even specify which conversion factors to use. So they’re using all the same ones and tell them to use ninety Mme. Per day to define, quote high dose, and we can tell them to use only the Mme. Definitions that were vetted by and in the Cdc guidelines Sounds like a pretty boring experiment. Right?
  • Well, this is how it plays out. Each of these four analysts identified a different set of patients who are considered high dose. They agree at the extreme top end of the range, but in the space where most long term patients fall. There was simply no consensus.
  • What happened? How is there any room for variability when we’re using the same conversion factors. The same data set. Something beyond pharmacology is happening here
  • to get to the root of the problem. Let’s give the same four analysts. A simple scenario. Here is a patient with two prescriptions. We modify this example that appears in the Cdc Cme. For the for the pain guidelines the original ones, and we added one more prescriptions to make it more illustrative.
  • So if you’re reading a following along a home, you may want to screen Cap aside if you’re used to doing. Mme. Uh calculations, I promise this won’t be more complicated than basic arithmetic. So the first script is thirty Megs of er oxycodone twice a day for around the clock pane
  • sixty tablets for a total uh, mm, Mm. Mm. Of two thousand seven hundred. The second script is five Meg’s oxycod own twice a day. Pr. And that’s uh fourteen tablets, and it comes to one hundred and five milligrams of marking equivalents, and the total Mme. Then would be
  • twenty, eight o five. And look, there’s there might be somebody wants. And how you calculate this. That’s fine for the purposes of this analysis. Let’s just assume that everyone is using the same uh same numerators for the study.
  • So two thousand seven hundred, divided by thirty is ninety mm. Right? So before the second script for the breakthrough pane. We’re already hovering right at the ninety Mme. Threshold. So this should be easy to identify which ones are the high. You know whether this is a high dose, Patient or not,
  • you might have, so you may want to drop down to eight hundred and five, because we’ll see that number again. In a moment
  • the four analysts would actually disagree on how much daily Mme. This single patient is getting. Their calculations will range from thirty, one to one hundred and five Mme. Per day, half of them saying that this is a high dose patient, and the other half saying it’s not
  • the same could happen across doctors within the same clinical practice, or between a prescriber and an insurance company.
  • So we’ve we’ve so far concentrated on the Mme. Part of the Mme. Per day metric. What we Haven’t considered is the denominator the word day. So to to borrow a line from the musical rent, the fundamental question is, How do we measure a day in the life of a patient?
  • I’ll spread you by not singing the song, but, uh, that’s kind of the concept we’re getting in, so i’ll. Next. I’ll show you exactly how these four analysts arrived at these four very different Mme. Per day Calculations for the same patient to Script scenario,
  • We took a careful look at all of the studies cited in the Cdc guideline, the original guideline. To justify the ninety Mme. Threshold of these we found eighteen that use daily me.
  • We combed over the methods and appendices and reverse engineered underlying equations which none of the papers explicitly included. Look, my own paper is on here, and I’m just as guilty for not including those equations. And I apologize. I’ve learned better after doing this project. So this is something i’m here to rectify today
  • in the accepted paper accompanying this presentation, we reproduce the verbatim extract from these eighteen papers, and i’ll show you some of them in a moment. I know It’s a little hard to believe, but the eighteen studies used to establish the ninety, the ninety daily Mme. Threshold, and the Cdc. Guideline
  • silently used four different definitions
  • all centered around. How you interpret the word day. In some cases the same authors use different definitions between studies without commenting on why they switched it up.
  • So if people are using all these definitions, then it can’t be that big a deal right to to choose between these definitions
  • I don’t know. Let’s find out
  • to a tone for the lack of detail in my published studies I worked with Allen kin Lot Unc. To reverse engineer the equations that underlie these four definitions at its heart. The four definitions are measuring something different. What we are building to is that ninety Mme. Is not a standardized clinical metric.
  • We face this challenge in other parts of medicine like prostate-specific antigen tests and Bmi, but we haven’t noticed it in pain medicine in the same way.
  • Next, I’m going to show you some equations, but if you don’t speak scientific, nerd like, I do just focus on the numbers. This slide is here for completeness.
  • We’ve heard lots of amazing discussions in the Pain medicine field about the pharmacology of opioids,
  • but that’s something that most patients and even most doctors don’t feel comfortable challenging as because they’re kind of held up to be these fundamental molecular truths. In contrast, what i’m about to show you is nothing more basic more than is nothing more than basic arithmetic. You can do this at home.
  • Okay? So i’m going to briskly show you how the four equations apply to that two prescription scenario you saw earlier
  • Mit. Ctl, and the first definition we call total day supply. In this definition the numerator is still to eight hundred and five to eight hundred and five mm. That you saw earlier. The denominator is thirty-seven days adding up the the day supply for the two prescriptions thirty days and seven days. Dividing these numbers, we get seventy-five point eight daily mme one hundred and eight.
  • This is by far the most common definition used in the literature,
  • and this paper by Von Korf, is by far the most commonly cited four definitions of Mme.
  • Um.
  • This is the metric that often shows up in clinical data dashboards like in Pdm. Piece.
  • Two things to note: First, the day supply can exceed the number of calendar days. Second. This definition definition is perverse. By adding a second script for breakthrough, you actually get a lower daily Mme. On a clinical basis. This measure does not make any sense at all, but on a population level. I see this citation and measure use all the time in the literature, in part, because it’s so easy to calculate at large scale.
  • The second definition is similar, but takes into account the overlapping days. So the denominator is thirty calendar days in our two Script scenario dividing, we get ninety, three point five daily. Mme. So just with this subtle change in denominator, we end up on either side of the ninety M Mme. Threshold.
  • This probably makes sense to most clinicians, but it was only used in two out of the eighteen studies cited in the Cdc Guideline.
  • This is also the method that the Hhs office of the Inspector General recommends, and they provide a handy set of tools to calculate it in Sas R. And Sql.
  • The third definition uses a fixed number of days for the denominator. Clinicians may be baffled, but this is actually the method used in a lot of the papers that are cited in the Cdc guideline, and it is a single most, and in the single, most cited paper, on the risk of dose and overdose mortality.
  • Uh, which is this paper by Kate Dun and Colleagues. This definite, they use this definition. So this definition definition has also been used in Cdc. Published studies, So going back to two thousand and five for the numerator. When we put ninety days into the denominator, we get thirty one point two daily. Mm-hmm
  • This method again gets used a lot in research because it’s so easy to implement, but it’s a bit unclear If those findings really have clinical relevance
  • in the study study in the guideline ninety days was the most common denominator, but some studies also use longer times, further shrinking the daily Mme. Measure and inpatient Medicare studies often use thirteen days, because that’s the reimbursement cliff.
  • So we went with. So here in this example we just went with ninety, because it’s the most common,
  • all right. Finally, the fourth definition D four is something we call maximum daily dose. This is the definition used in the mobile app for clinicians that accompanies the Cdc guideline
  • they call it total daily dose. But we think maximum is a better word, because when you look at the equation you see what’s actually happening here. Um! It assumes that the max is the maximum prescribed dose on any one day ignoring any uh large dosing from intentional cell phone or suicide.
  • Using this definition ninety Mme. Plus fifteen. Mme. Gives us one hundred and five. Not surprisingly, Max Daily dose returns the highest measurements. This definition is actually a bit tricky to implement with staggered start and overlapping scripts. But we’ve shared our code to help you uh kind of work through those. If you’re interested in this particular definition.
  • It’s worth noting that this definition is different from the one suggested by the Hhs office of the Inspector General. So we have two Federal agencies telling us to calculate the same metric in different ways again. Um, Mme. Is not a standardized clinical metric.
  • One of the analysts on our team yanning uh works with software engineers who who process Pdmp data and other large databases. She’s been frustrated that most software vendors won’t even give you enough detail to know how the Mme. Per day was calculated, let alone explain in terms that clinical decision makers can actually use. This is a real world practical problem that impacts clinical decision making and clinical support tools, including in pdmps.
  • Okay, So we’ve established that there are four ways to calculate daily. Mme. But Hey, this is all just academic right? How much relevance does this have in the real world? It’s a fair question. So let me show you
  • we did a controlled experiment. Imagine we have two places, and we observe that one place has a higher rate of opioid prescribing than the other eight point seven versus seven point nine per one hundred adults.
  • This is exactly the set up that’s used in a lot of policy and intervention, evaluation and in epidemiology. For our purposes the cause is less important. Let’s just agree that a difference exists between these two places, and that we have a cause that we have a reason to measure it.
  • Given the mild and balance of rates on this slide, we may want to know if there is a difference in the proportion of high dose unquote patients between these two locations. So we did a study comparing two locations like we would in a policy or intervention analysis.
  • We conceptualize this as if they were four different papers, using the exact same data set, evaluating the same exact intervention or policy or clinical behavior.
  • The key thing to remember here is that the only source of variation, the only source comes from the four definitions, and in particular, how the denominator for daily Mme. Is used.
  • Here are the methods we use outpatient dispensing data from pdmps in California and Florida
  • We chose a short three-month period to avoid secular time trans and seasonality
  • we define high dose as greater than ninety mm, because that’s what is in the guidelines, and it’s not some. It’s not a number that we are going to be obsessing about, uh in our clinical practice, or in our regular day to day science. But this is the kind of threshold that’s in the literature, and we wanted to make an example out of
  • um. So we defined the doses as ninety. A high high dose quote unquote is ninety mm, and we looked only at the solid oral and trans normal opioids uh patches should have been mentioned on the on the previous slide, too.
  • Um, And so we use the Cdc conversion tables for equality as a potency. The equations allow you to substitute other potency factors, but we held them constant here because we were focused on the denominator of the word day.
  • For the statistical analysis. We did three main things. First, we compared the percent of high dose patients between Florida and California, bearing only the daily Mme. Second, we quantified the milligram difference in average opioid dose per day again varying only the daily Mme.
  • Third, we conducted a meta-analysis, seeing if four different studies using the same data set the same conversions, factors would have statistically agreed with each other.
  • This method is used a lot in observational studies and clinical trials to evaluate. If a set of studies can be, or even comparable, and if they can be combined for a Meta analysis, and if they’re measuring the same thing,
  • all right. So for the sample size, we have about nine and one over two million prescriptions representing about four million patients.
  • The numbers you saw earlier were real. Here’s the three month dispensing rates in Florida and California for the third quarter of two thousand and eighteen. We chose these places, because these are two of the three most populous states in the country, and i’ll show you some preliminary data from Texas as well.
  • All right. Here are some results. So here are the California results: First, same data, same conversion factors, but definitions D, One and D. Four, which are both used in clinical practice showed a threefold difference in who is perceived to be high dose,
  • adding in the forward results, we see a similar story with D four really identifying a lot more patients as high dose than the other definite other three definitions at some point i’m going to get tired of using the air quotes, but I hope i’m in a friendly audience that understands that. I’m not saying that ninety Mme. Is like a high dose threshold, but just using it, because that’s what the standard has been in the literature
  • it’s worth, pointing out that the definitions perform differently in each State, while D three returned the fewest high dose patients in California. It was d one that was the lowest in Florida.
  • We also have primary data from the Texas Pdmp. For the first quarter of two thousand and twenty.
  • At the outside of the pandemic uh of the Covid pandemic you see a similar pattern across the four definitions, but the overall overall proportions are lower. This could be due to time trends, but also that Texas tends to prefer ir hydrocodone way more than most other states.
  • The takeaway across the three largest states in the country. The definitions disagree, whether hundreds of thousands of patients should be classified as high dose or not.
  • Taking the same data. If we were doing a policy or intervention analysis comparing the two places Florida and Texas. I’m sorry Florida and California.
  • Four studies on the same data set wouldn’t even come close to agreeing with each other. Was there thirty, nine percent more high dose patients in Florida? Or was it eighty? Four? Those are big big differences in terms of policy implementation.
  • In fact, the heterogeneity from definition alone is so high that we wouldn’t be able to combine these into a summary measure for a meta-analysis. This really calls into question a lot of how Mme. Is used in the scientific literature.
  • So instead of percents. We may want to know how much higher doses were given, how much higher doses were in one place versus the other.
  • The four definitions, don’t even agree. If the average er only pain. Patient is getting a high dose, so you can see that, based on just the definition, the average dose will bounce higher and lower than this artificial ninety Mme. Per day threshold.
  • So let’s look at this a little bit more closely. On a technical level.
  • We’re still comparing a milligram difference between Florida and California. The vertical axis is average Mme. Per day. Each blue bar is a different definition. The way to read this chart is at the bottom of the bars, California and the top of the bars. Florida is always higher.
  • The height of the bar is how different the States are in terms of the milligrams of morphine equivalence, using maximum daily dose t four. You get a really big difference. But these numbers are all over The place
  • ready. It gets worse. The means were highly, right-skewed, meaning ultra-high- dose Outliers were driving the averages to be unnaturally high in these situations we turned to the Medians in orange to do studies more accurately.
  • Here’s where things get even more interesting. D Four, which exaggerated the differences between States using the arithmetic average returns much less variation between states with the Median D three median shows the least difference at zero point, nine milligrams.
  • The message here is that subtle choices have major consequences for policy and intervention interpretation
  • which one of these is correct, I honestly don’t know. It depends on the research question. I guess it’s not as simple as choosing something in the middle, though that’s our natural cognitive tendency.
  • Each bar on this plot could legitimately have been justified in an observational study and glossed over in the method. Section. It’s a mess, while they all point to doses being higher in Florida, they really call into question what is being measured
  • in policy and intervention studies even small effects with large data sets can carry a lot of inferential validity because of p-values
  • My recommendation is to let go of the ninety Mm. A threshold and odds ratio in these types of studies, and instead treat Mme. As continuous, and use multiple metrics, Median and mean at the very least. Please please please state the equation or definition you’re using,
  • because otherwise I can’t interpret any of the papers that fail to put those in. And if you look at the all, all the dozens of papers that are published every week, using Mme. Per day as a metric. You’ll find nearly none of them reference with adequate detail which of the definitions were used and which equations would be relevant.
  • So how do these definitions impact our interpretation?
  • While using D Three and Medians? You could conclude that there are a lot more high-dose patients in Florida, but they’re only getting a milligram more or You could conclude that there are definitely more high-dose patients in Florida. But on average they’re getting a lot more thirteen milligrams per person
  • So if you’re evaluating an intervention or policy, these subtle choices have huge consequences Again, which of these is correct is not clear. But we do know that what I both kind of describe which one of these definitions, I think is better for the research setting at least one.
  • When you break this analysis out by Ir and er opioids, you see a lot of heterogeneity so much so that a Meta analysis would conclude that these studies were so heterogeneous that you couldn’t combine them to summarize in a Meta analysis, summary statistic.
  • The other incredible thing is patient selection
  • in pharma-sponsored observational studies. You sometimes look at people who are getting er who are only getting er opioids to make the cleanest possible comparisons without confounding from additional ir opioids. If we did the study on er only patients, we would actually conclude that California had higher opioid doses. Again subtle choices, major consequences.
  • And this is why I love epidemiology. Because devil is really in all these details.
  • So, despite variation in the underlying definition, the studies cited in the Cdc. Guideline, consistently found an increased risk of fatal overdose above ninety Mme. Per day. The simplest explanation is that this observation is the artifact of turning a continuous metric into one that is categorical,
  • all except for two of the studies categorized Mme. Using ninety to one hundred and twenty milligrams as a lower bound for the highest stratum, and i’ll show you this visually in just a few minutes.
  • Our study supports Fda’s contention that overdose risk with opioid analges is a continuous function and rare.
  • A big part of the issue is overlapping scripts. These really impact how the definitions perform. So how common are overlapping day supply forty, two percent of prescriptions overlapped with another script in our sample. It affected one out of every four patients, including most long term opioid patients.
  • Let me address the epidemiologists and methods folks in their audience for a second. Here, on the technical. On this technical level it may seem that, choosing one definition and applying it over time would be less worrisome right
  • in most other countries. This is true, however, when you look at the equations carefully, two specific Time-trans scenarios emerge as problematic in the United States. First, if overlapping scripts decrease over time choosing definition. One total day supply will attenuate the intervention effects compared to definition two, which is on therapy days,
  • so they will. They will attenuate the effects differentially at ah earlier time points. Second, if ir and er opioid, dispensing Don’t both decline at the same rate over time. So it’s the assumption of non-parallel linearity.
  • The same thing will happen both of these prescribing trends happened in the Us. Over the last decade, seeing faster declines in er versus I are opioids.
  • So we’re doing simulation studies to quantify this. But in preliminary work the amount of discrepancy that you would get by using the same, you know, by saying like, okay, I can just going to use one definition over time in the United States over the last um ten, fifteen years would lead to an artificial thirty-three percent discrepancy i’m happy to go into more details about the uh longitudinal aspect of this. But, uh, those are some of the kind of broad uh points for
  • epidemiologists trying to operationalize these definitions.
  • We also explored what happens at the threshold boundary, comparing ninety point oh, versus ninety point nine. So like, where precisely do you draw the line for a high dose? Published studies often Don’t make clear whether their ninety Mme. Threshold is ninety or ninety point nine.
  • These are like small methods, details that get left out of papers, but we really should be paying attention to them.
  • If you shift the high-dose threshold from ninety point nine down to ninety you increase the number of high-dose patients by fifteen. That means there are a lot of patients who are being held at this artificial threshold
  • mit ctl. And despite the definitions not being clinically standardized. I found that to be astounding that um! That this is happening right at that threshold, and even a small difference in where you round and how you round can have a fifteen percent difference on whether you consider a patient to be high dose or not. One hundred
  • June and analysts on our team thought this was unexpectedly, unexpectedly huge, too, and I agree, he points out, that these little changes in studies can lead to a lot of misclassification.
  • So we quantified the extent of that Miss classification. Think of this like a doctor and an insurance company said, setting the threshold on either side the boundary, either ninety point zero or ninety point nine. So within rounding error, right
  • using definition, one the insurance company and the physician would disagree one out of fifty-six patients, whether the beneficiary was a high dose, patient or not with definition in four it’d be one out of thirty. It’s worth noting. That definition three is the most robust to this kind of discrepancy.
  • Florida using definition, One, eighteen out of every one thousand opioid patients are getting exactly ninety Mme. Per day,
  • and using definition for thirty, four out of every one thousand opioid patients are getting exactly an idea at a per day.
  • So which
  • so? Here’s some all you can prescribe, not to decide. Uh, and we do not consider other sources, whether pharmaceutical or unregulated opioids, we do not try to differentiate cancer from non-cancer, pain um and we do not consider a typical me opioid receptor agonism for respiratory depression, such as the pens it all or buprenorphine,
  • and we do not consider pharmacist-based variations in day supply calculations which is part of some other research studies we are doing.
  • And finally, we did not consider any social and structural determinants that would change um the Mme. Calculations. But at the end of the day what we’re really trying to do is figure out what the definitional aspect and how the word day is operationalized in Mme. Per day.
  • user avatarUnknown Speaker33:26So which definition should we use?
  • user avatarDr. Nabarun Dasgupta33:29Um else? Uh, So this is so. This is Tosca. Uh from our team, and she is assisting to reply to that. There’s no one size fits, but at least we need to be showing our work, and I think that’s really an important part of what we’re doing here.
  • Uh, we’ve heard lots of reasons for clinical caution for using Mmes. So in terms of research, studies and policy and evaluation intervention. Here’s kind of the big. Here’s kind of the the where the rubber meets the road.
  • I don’t think d one should be used period. It’s not a real representation of calendar time. D two is the version that I use D three is the most robust to misclassification. So if you have messy data. This one might be attractive.
  • D. Four could be useful, maybe, in short-term toxicity studies in opioid knife patients without risk of suicide, but definitely, should not be used for long term situations.
  • The inaccuracy built inherent in that definition, grows with time, and as you can into it, from staring at these equations long enough.
  • But candidly, if a paper doesn’t sufficiently define how they calculate a daily Mme. I just don’t read it anymore. I just can’t make sense of it after having seen how much variation comes just from the definitions.
  • So here’s a here’s the fundamental here’s another fundamental question: right? Why do the studies that have been published consistently show increased risk at higher than ninety Mme. Per day.
  • I think we can take a lesson here from Bmi body mass index,
  • which I assume everyone’s familiar with where there is, you know it used to be. There were kind of three classes or four classes underway, normal, overweight and obese, and as the population has grown older and larger, now we have all these other of these categories. So this is kind of the distribution. I think these day are actually from Australia. But distribution looks similar. I like the pretty colors. This, so I chose this one. Um.
  • This is from a study that we did in North Carolina about a decade ago.
  • But we’re looking at the uh one year. Risk of overdose death among patients prescribed uh opioid analges. The first the top graph is the absolute risk. So instance rate per per uh ten thousand person years, and the bottom is the instant rate ratio uh pinning it at the lowest uh the lowest band of opioid prescribing. So. Um! I hope you guys have seen that paper before, but the concept here is what i’m trying to get at get at is, if we did the same thing for Bmi,
  • we would treat this risk scale right. This like this. This the dose dependent, overdose risk as a continuous function. We would, we could theoretically categorize it at different cut points. But we would definitely not be treating everybody as underweight or not underway that gives you. That’s kind of nonsensible when you look at continuous data.
  • So this is the same graph kind of with those color bands, and what most of the studies that have been published to date, showing a ninety ninety Mme. Inflection point for uh for overdose risk with prescription by do by prescription, opioid dose have done it the top the way that it looks like on the top that use ninety as a cut point, and then put everything else above that in one bucket. But we know that there are a lot of patients who are
  • getting all you know, doses that are above one hundred, and the risk is not linear And so, instead of treating this as a can as a dichotomous, high, low, or even like slower gradations below ninety, I think we should be taking a much more nuanced approach, and looking at risk over a spectrum of actual uh clinical, clinically relevant dosing
  • on top of that kind of going back into the studies that have shown a relationship between opioid prescribing and overdose mortality. There’s two specific assumptions that have gone silently unaddressed
  • under address, unaddressed. Um, and these This is the subject of a newspaper that we have coming out in the next couple of months. It’s just about to go into review. Um! But I wanted to flag these for folks who are doing these types of studies.
  • So the two assumptions are that first, while drug exposure was measured with product levellevel specificity. So you can say, you know oxycodone patients hydro-codone patients, whatever it is,
  • the outcome for all the big studies that have been published on this association was overdose from any substance. So someone who died uh who received hydro more phone, but subsequently died of a heroin overdose would be treated the same way as someone who was prescribed hydro, more phone and then died of hydromorphone toxicity. And I know medical examiner data uh are easy to beat up on in terms of how the uh you know what the causal relationship is, and we know there’s a lot of variation
  • from state to state, and how accurately that’s measured, and how much causality is put into it. Let’s just assume for, uh that we’re looking at states where there is a standard protocol in the Medical Examiner, or Corners offices for um for saying It’s not sense of an of an opioid that’s detected and postmortem talks. But rather the the levels were high enough, and there’s corroborating evidence that that particular medication was involved in that particular substance
  • was involved in their fatal respiratory depression.
  • The second assumption that we don’t talk about is that the time between dispensed exposure and fatal over those are generally treated as uninformative with the common assumption that once exposed, always exposed to the analgesic. So this is kind of taking an int intention intention to treat a approach which we do often in safety studies.
  • But that time period there’s a lot of other things that could be happening. Doses could be fluctuating. Um, so it’s and people could be coming off that medication. So I think these are like two really important assumptions that a lot of the studies almost all the studies to date have made. But we Haven’t made these explicitly clear. So this is where we’re going. Um! We’re. This is where we’re headed with our next uh set of research, which will be which i’ll be able to present on very soon. Once that paper gets
  • published or reviewed.
  • The fundamental thing that you need to understand here, from what I’m saying is that these two assumptions conflate distinct causal pathways that are separated in time,
  • and by this i’m talking about four possible causal pathways, and i’m generalizing here. But this is kind of where we think you know the pathways, we think, where uh medical opioid exposure may subsequently be associated with overdose uh overdose deaths, and this is
  • again not a high um. It’s not a high frequency event. In our studies. In North Carolina we found the rate the one year of of overdose among paying patients to be like zero point zero.
  • Sorry zero point zero two. So this is a very, These are very rare events. But This is a important question that a lot of people are trying to study. So we’re trying to make this make the assumptions more clear, so that we can have better data.
  • So I think there are and feel free to disagree with this. But there are four generally general causal pathways, and the reason we break these out is because the kind of statistical methods you need as an Ep. Study designs you would use to measure each of these causal causal pathways is different. But this is something we don’t really take in where this the published studies have really not taken into account, so there could be patient behavior taking more than prescribed
  • uh, and a subset of that could be a people who had a previous history of substance, use disorders, or pace placed on opioid analogy at therapy with limited guard rails and end up having trouble with those medications
  • there’s, also prescriber behavior medication errors, two high-starting doses for opioid naive patients medication interactions, and those are a lot of the things that are controllable within a medical system. Three.
  • There’s also this popular concept of Irogenic exposure. Whether this is real or not. Um, that that patients get predisposed to opioid disuse disorders from getting prescribed a medication, and then eventually overdose, either from leftover diverted or illicitly manufactured opioids,
  • and then the fourth possible causal pathways, abrupt disc discontinuation, where patients on long term opioid therapy have their doses abruptly reduced or terminated resulting in either suicide, because their pain is in adequately managed, or ah having to supplement from getting opioids on the street.
  • So when we combine these all these causal pathways and try to create like a omnibus um on the bus. You know, association between prescribed dose and overdose death. We’re actually having a lot of heterogeneity, heterogeneity in what we’re actually measuring, and what the research question is.
  • Um. So again, we will come back. I will hopefully, in a future seminar, be able to give you the results from this. But the short of it is that these are things that we should be considering.
  • So, going back to the definitions. What do doctors and patients think so first of all. So here’s an example of where policy matters when it comes to how you define Mme. Um.
  • And so Arkansas Medicaid required beneficiaries with greater than two hundred and fifty Mme. Per day, to be paper tapirts, ninety milligrams per day. During an eighteen month. Period. But we saw how d four is not ideal for patients already on opioid therapy,
  • but when using the Cdc mobile app, this could easily be the definition that gets applied clinically,
  • while another prescriber may choose on therapy days, because that definition, two which makes a lot more clinical sense.
  • The bottom line here is that Mme. Cannot be considered a hard threshold because it’s not a standardized clinical metrics. And uh, you know, interventions like this, and policy documents, cl clinical guidelines that Don’t define how Mme. Per day, especially the word day is handled, are uh inherently problematic in my mind.
  • Uh Doctor uh, Dr. Tiji, who’s one of uh, one of the Uh Pain physicians and the head of pain. Medicine here at Unc. Um had this to say, while pairs insist they’re not dictating care because the patient can still pay out of pocket for the medication
  • for most patients. This is not financially feasible, so the way the Mm. Is calculated, and Mme. Restrictions that are put on. Ah! Put on patients by insurance companies have a tangible effect on patient care
  • and and the definitions that this, that the definitions we use are central to calculating those numbers.
  • We also have a quote here from a pain patient who we work with, Who’s also a sociologist on our team, and she says, far too often we are the victims of good intentions, of those wanting to do something about the opioid, overdose epidemic.
  • But that something that is done oversimplifies the problem, and pushes cookbook medicine upon those of us. With complicated medical situations.
  • We wait and we suffer, and we hope it will all get sorted, so we can get the care that we need. And I think that’s the That’s the sentiment here that we need to remember of patients being affected by these arbitrary definitions.
  • Finally, Chris Delcher, Ah, who was who was part of the science, points out that this work is an example of how we can put Pdmp data to work positively for patient care. And there are a lot of reasons why pdmps are
  • harmful to patients, and uh kind of have forced physicians into different in in, you know, box positions into corners in terms of how they handle patients, and we heard uh about that from on end during the last uh seminar series with Dr. Olivia.
  • Um. So in this particular instance we were able to use the Pdmps Pdp. Data for something useful. Um! And we were able to educate the Pdmp at administrators on the importance of the Mme. Calculation. So that’s the um. So that’s the
  • That’s the so. The key message here that is, subtle. Choices have big consequences. What we’ve seen today, and reinforced by a lot of the presentations you’ve seen in this seminar series is that is proof that daily Mme. Is not a standardized clinical metric,
  • and
  • so the question is like, Why has this not been detected sooner. Here’s my take on it
  • the computational ease and the evocative lure of molecular fundamentals collide in an optimal level of cognitive complexity to engender mmes with an unsubstantiated aura of immutability. They’re not immutable. These are not standardized clinical definitions.
  • Thank you,
  • user avatarBeth Darnall46:58Dr. Des Scoop, but thank you for a terrific lecture. I want to invite everyone um on the Webinar. You can chat in your questions. Um, so please chat, chat those in. We want you to be part of the discussion. Um, we have one here. Um! Would you comment on Gaba Penton? Combination concerns regarding overdose?
  • user avatarDr. Nabarun Dasgupta47:22Yes, so Gaba Pens and Pre Gavelin are have been promoted as a way as an opioid sparing uh option uh, often given, like during surgery or right before surgery, and the hope that uh people will need less opioid analgesics uh post off
  • Uh. What we have found is that that does not change. Uh, that is not actually change. How much uh pain really people get afterwards, and it’s probably more of a marketing thing. Um, in terms of overdose risk.
  • It’s. I think it’s hard to say. I don’t see a lot of pharmacological mechanism that would lead to combined respiratory depression when it comes to uh to gab a pension plus opioids. Um, But
  • it does show up on toxicology. I don’t think it’s a major driver of overdose, even though it does show up
  • user avatarBeth Darnall48:17terrific. Thank you. I You know I was so struck by how you were talking about how software vendors don’t need to disclose how they’re calculating. Um, me, I I just love to hear more your thoughts on that. You you described a Cdc mobile app. Is that
  • true for the Cdc mobile app as well
  • user avatarDr. Nabarun Dasgupta48:39the see? Yeah, the Cdc mobile app doesn’t like doesn’t break it out with enough detail, but we
  • we reverse engineered it. We figured it out. Um, there was like, there’s like using kind of other Cdc documents. Um, But Yeah, none of the clinical tools actually get down to this level. What you would think is a fundamental thing, but
  • it’s just not something that has been
  • user avatarBeth Darnall49:04publicly disclosed in any of the clinical support tools that we looked at.
  • user avatarDr. Nabarun Dasgupta49:21Yeah, I agree. So like the you know, it’s like we’re We’re often talking to cedar right? Which is the drug division, And then Cdr. H. Which is the deviation. Yeah, I think it’s time for them to get it together and like, really understand. You know, I think that there’s a great like regulatory framework for mobile apps. I think these should. These definition issues with Mme. Calculators really need to be
  • user avatarBeth Darnall49:44uh, you know, really need to be looked at with a careful lens. Yeah, yeah, well stated um. Next question, Steve Arians, this all seems to ignore the elephant in the room. Pharmacogenomics, Dna testing.
  • user avatarDr. Nabarun Dasgupta49:59Sure, I think you know, genomics and metabolism uh of opioids and uh of different kind of pain markers are really important for a clinical decision making.
  • And uh, again, we’re not we weren’t. Looking at a outcome of pain or of overdose in this study, right? We were looking specifically at the way that we measure opioid Mme. A. Daily Mme.
  • Because that was that was our intent. So I agree that if you’re looking at something more nuanced with like more hard outcomes than pharmacogenomics, would be an important part of that
  • user avatarBeth Darnall50:35terrific um. Dr. William Rose, given that standard of care is determined by the practice of the prudent physician, and that the Medical Board accusations against physicians are in large part based on the application of State endorsed Pdmp data. How can one defend the standard? Recognizing the room for these errors?
  • user avatarDr. Nabarun Dasgupta50:59I think this is a this. This is kind of like the part that I didn’t say out loud right and um! I think that some of the prosecutions against physicians need to be re-looked at in light of
  • uh in in light of these findings. Right there is. There are legitimate reasons why you would use one definition versus another. And if the you know, if that definition, if what the standard of care definition is
  • has not been specified. Um, I I don’t even know that it should be specified. But there is enough. There is enough variation here that I think. Um. You know defendants who are physicians being brought up in front of their medical boards, might want to look at some of this variation to push back against
  • the you know, an arbitrary standard of care based on Mme.
  • user avatarBeth Darnall51:52Terrific. You know. Um Dr. Deskuca, you made some really great
  • concrete recommendations. Some of them I was live tweeting as you’re speaking. Um, but i’m curious. Do you have a single publication or resource hosted anywhere That would be a quick reference. Guide for your recommendations for how to minimize error in Mme. Calculations
  • user avatarDr. Nabarun Dasgupta52:20cool. I probably should do that. But I think, as
  • I mean
  • user avatarBeth Darnall52:26for clinicians, or some or researchers right like maybe both. I was just like, because i’m thinking, Gosh! You know, if if people just want to take away summary, because I think from your talk, what we what we can appreciate is like. Look, there’s so much variability. How do we minimize that? And And what what do you recommend for clinicians? What do you recommend for researchers? But I mean you’re the guy, so I I’d love to have those resources.
  • user avatarDr. Nabarun Dasgupta52:54I would say, go with definition, too. I’ll make it very simple, you know. Make it face valid right like Don’t. Make the number of days longer than the number of days in a month, and don’t
  • uh assume that patients are getting one script, you know, for an Ir opioid and taking them over ninety days. I think that’s kind of the bottom line. Um, I mean. There’s There’s a lot of nuance and Epi studies that get that gets, you know,
  • swept under the rug, and there are a lot of, I will admit there’s a lot of epidemiologists in my profession who don’t pay attention to these details, either. But there are some of us who are interested in
  • the best types of outcomes for for patients, and I think these are a part of being a conscientious and thoughtful scientist.
  • user avatarBeth Darnall53:43Thank you. Um, Roger nags hopefully. I’m pronouncing your name right. Does the type of pain acute versus chronic pain make a difference?
  • user avatarDr. Nabarun Dasgupta53:53It does. And if you look back at the Meta analysis slide Um! That was where we broke it out By er and I are opioids. We were trying to get at that indirectly, because we couldn’t get a cute versus chronic in the data sets we had.
  • Um, but I think the more chronic pain patients you have in your,
  • for example, especially if you’re in a practice setting where uh people are allowed to get uh I our opio is for breakthrough. More, there’s overlapping scripts. Or if you’re in a setting with where you’re using multimodal, you know different different opioid molecules. At the same time,
  • when you’re over when you any time when you have overlapping scripts? What happens more in chronic pain than an acute pain,
  • user avatarBeth Darnall54:39you will have the the the definition choices will get magnified so uh, definitely more of a concern with the chronic pain. Perfect? Thank you. Well, we’re just about it. Time. I want to thank you again. Dr. Nabaroon des scooped off for um a truly informative lecture giving uh clinicians, researchers, policymakers a lot to think about. Thank you so much for your important work in this space.
  • Um, I want to thank Ashley Gomez for both organizing and carefully curating our lecture series and making it easy for people to register to um join the the lectures and have a good experience. Um. Ashley mentioned that after uh today, within a couple of hours, those of you who want Cme will receive an email with instructions for how to uh claim that Cme. And uh for the rest of you. Please
  • join us in February for Dr. Mark Mcgovern and his uh his lecture on Implementation Science, and we will be having a speaker in January. That will be Tv. So please stay tuned, and we hope to see you on the next Stanford Pain Relief Innovations Lab Speaker Series, Webinar. Thank you. Everyone and Happy Thanksgiving.

The authors thank Bonny P. McClain MSc, of Data and Donuts for her graphical contributions.

“..that subtle choices have major consequences..”


$10, $15, $20, $25,$50, $75, 175, $500. OR MORE TO CASH APP:$docnorm

ZELLE: 3135103378


So, Donate to the “Pharmacist For Healthcare Legal Defense Fund,



Neat, Plausible, and Generally Wrong: 
A Response to the CDC Recommendations for Chronic Opioid Use






1 Comment

    Author: maryw M (IP address:,

    U know,,when u gotta change long standing definition and long standing units of measure,to trump up your lieing data on addiction claims for $$$,our government should of smelled a RAT,,Furthermore,was it not their own Krebbs study that used this forced mme,that PROVED at these low doses our medicine is NOT as effective??!!!!!!!!!!!!!!Again proof,,he/prop/cdc guidelines made torture in the healthcare setting,,,legal,,,,on purpose,,and now,10,000 if not more are in torturous treatable physical pain from their medical conditions ,and 1000’s are dead,if not more,by using death as their ONLY means to stop treatable physical pain from a medical conditions,,,Are we not theee ONLY country who changed all this???maryw

    You can see all comments on this post here:

Leave a Reply