Tag: analysis

Are doctors bribed by pharma? An analysis of data

By Rafael Fonseca MD & John A Tucker MBA, PhD

A Critical Analysis of a Recent Study by Hadland and colleagues

Association studies which draw correlations between medication company-provided foods and physician prescribing behaviour have come to be a favourite genre one of advocates of higher separation between drug makers and doctors. Recent studies have shown correlations between approval of medication maker payments and undesirable doctor behaviours, for example increased prescription of encouraged drugs. The writers of these posts are often careful to avoid making immediate promises of a cause-effect relationship because their observations have been based on significance alone. Nonetheless, this type of connection can be indicated by conjecture. What’s more, the high number of books in high profile books on this topic could only be justified by concerns that this type of rapid-fire connection is different and is prevalent and nefarious. Within the following report, we’ll analyze a recent newspaper by Hadland et al. which investigates correlational information relating same-sex to manufacturer obligations and where the writers imply the occurrence of a cause-and-effect relationship.1

We suggest the Connection between trades between the private industry (e.g., foods supplied, consulting obligations ) and prescribing habits may drop into one of 3 classes:

Type Effect Comments
0 There is no cause-effect relationship between these transactions and prescribing habits. Correlative observations may merely be reflections of practice patterns, and likelihood to use a drug category. No harm exists.
Ia There is a demonstrable cause-effect for transactions and prescribing patterns. However, this relationship is associated with increased use of drugs that have been proven to be an improvement over the current standard. The effect is beneficial for patients. “Beneficial marketing.”
Ib An adverse causative effect is documented with establishment of causation. There is a possibility of patient harm. Patient harm occurs because the wrong medication is administered and contravenes medical standards. A minor damage is done but arguably exists, if a physician prescribes a more expensive medication when a cheaper alternative exists.

“Nefarious marketing.”

Hadland et al.: Opioid Prescriptions and Manufacturer Payments to Physicians

The writers of the newspaper linked physician-level data in the 2014 CMS Open Payments Document into 2015 opioid prescribing behavior described from the Medicare Opioid Prescribing Database. They researched the theory that foods and other obligations increase doctor opioid prescribing by analyzing the association between reception of foods and other fiscal benefits together with the amount of esophageal prescriptions written[1]. Specifically, they discovered the next:

  1. A nearly linear relationship between the number of opioid manufacturer-provided meals accepted by a prescriber and the number of opioid prescriptions written. The relevant data is provided in Figure 1 below. Prescribers who received nine meals from opioid manufacturers in 2014 prescribed opioid analgesics at slightly more than 3x the rate of those who accepted only one meal.
  2. When broken down by physician specialty, those who accepted any payment from opioid manufacturers wrote between 1.2% more and 11% more opioid prescriptions as those who did not accept any such payments (Table 1).

Figure 1.

Figure reproduced from JAMA Internal Medicine 2018, volume 178, 861-3 under the Fair Use provisions of Section 107 of the U.S. Copyright Act.

Table 1.

Table replicated from JAMA Internal Medicine 2018, volume 178, 861-3 below the Fair Use provisions of Section 107 of the U.S. Act.

Hadland et al. finish that

Amidst federal attempts to suppress the overprescribing of both opioids, our findings imply that manufacturers must think about a voluntary reduction or total cessation of advertising to doctors. Federal and state authorities should also think of legal limitations on the quantity and volume of obligations.

While no cause-and-effect connection between obligations and prescribing habits was shown with this correlative studythat the consequence that one is made apparent in the writers’ guidelines. In our analysis belowwe try a deeper dip to find out if such a cause-and-effect connection exists.

Our View: It is More Complicated than That….

To better comprehend the topics introduced from the Hadland’s correlative analysis we undertook a different evaluation of the very exact data. We replicated that the Hadland data extraction in the CMS sources mentioned in the newspaper. We associated obligations with prescribing behavior utilizing doctor name and geographic information as clarified by Hadland. Regardless of the absence of detail supplied in the novel, we carefully replicated the amount of opioid prescribers, the amount of esophageal prescribers accepting obligations, and also the entire number of obligations described from the Hadland newspaper. The sole discrepancy we discovered between our data and that being mentioned by Hadland is we discovered that a substantial total payment sum of $13.1M compared to the $9.1M mentioned by Hadland et al.. We found no easy explanation for this discrepancy, because the entire payment amount has been always about 50 percent greater than that explained by Hadland if stratified by origin or by payment form. While we’re unable to rigorously evaluate the origin of the difference given the absence of a thorough protocol at the newspaper, we think that portion of this gap could have arisen by adding a broader selection of opiate goods in our investigation relative to that utilized by Hadland.

How Big is your Association Between Manufacturer Upgrades and Prescribing Volume?

Our very first criticism of this Hadland evaluation is directed in the non-standard demonstration of this information at Figure 1. The most commonly accepted method to demonstrate the connection between two constant factors such as obligations along with also the prescription count is really that a correlation diagram. We provide the information in this way in Figure two (Notice that the logarithmic Y axis). Doctors who admitted no free foods from psychiatric manufacturers composed between two and 1000 opioid prescriptions in 2015. As did people who admitted 50 or more.

Figure 2. Correlation Diagram Relating Number of Opioid Prescriptions Written to Number of Drug Maker Meals Accepted

This chart provides a very different opinion than the demonstration of exactly the very exact information in Figure 1. Why is this? Here we’ve proven every single data point, although some are difficult to see as there are a lot of these (345K to be precise ). In Hadland’s demonstration of this information they grouped the prescribers into groups depending on the amount of foods they admitted. They calculated that the mean for each category, that hides the enormous variation in prescribing behavior within every category. The error bars have been displayed in Hadland’s figure aren’t normal deviations (a step of within-group variant ) but normal mistakes (An example of just how the mean was anticipated ). The latter value comes from the prior dividing the square root of the amount of information points, which ranges as large as 8468 for a number of the groups in Hadland’s figure. Therefore a very clear representation of within-group variant would reveal error bars up to 92-fold bigger than those exhibited.

A similar criticism could be directed in the demonstration of this information at Table 1. Assessing mean prescribing rates between people who admitted any payment and people who accepted not provides a non-representative image since the distributions are highly skewed. Picture a cancer trial where 5 patients reside two, 3, 4, 3, or 20 weeks. Reporting the normal survival was 7.5 weeks along with the standard deviation was 8.3 months actually doesn’t offer an extremely meaningful image of what took place from the trial. In the same way, Hadland et al. report which doctors who admitted obligations in 2014 composed 539 +/945 prescriptions in 2015, although people who didn’t composed 134 +/281. Who are the doctors who composed less than prescriptions in 2015, and also exactly what exactly does a negative prescription seem like? This sort of bizarre effect arises from using statistical procedures appropriate to a standard distribution of values into a data collection that’s decidedly non-normal.

The issues become even more evident when we compare those amounts to the writers’ statement in the text which people who accepted obligations in 2014 improved their prescription count at 2015 from 1.6, while people who didn’t accept obligations in 2014 decreased their prescription count from 0.8. What’s the gap (2.4 prescriptions) equivalent to 9.3percent of 134 prescriptions (Table 1)? ) And doe a comparative growth of 2.4 prescriptions each year in the foundation of 539 prescriptions warrant publication in JAMA Internal Medicine along with also a call for laws?

Are Drug Companies Paying Doctors to Write Prescriptions?

Though the correlation between foods and opioid prescriptions is a lot poorer than suggested by the statistics introduced in Hadland et al., a sensible person may still thing that ANY market where prescriptions result by a conscious or subconscious quid pro quo free of lunches is unacceptable (Type Ib). We’d definitely take that place. So let us examine whether the connection is causative or simply correlative. Hadland’s implied theory is that physicians are composing opioid prescriptions at”market for pizza” Another explanation may be that attending maker informational sessions where foods are served along with prescribing opioids may both be pushed by with a practice that entails treating several pain sufferers. Let us examine the information and determine if we could differentiate between these choices.

  • If Doctors are writing prescriptions in exchange for payments, one would expect that the number of prescriptions would rise predictably with the payment amount.

In practice, we find this is not the case.

Regressing the amount of opioid prescriptions written by full payments obtained, we locate r2 for the significance is 0.01. So only 1 percent of the entire variation amongst prescribers is related to variation in the sum of payment received. (The gap from the chart between $0 and $10 appears because CMS will not require coverage of payments under $10).

Figure 3. Relationship Between the Number of Opioid Prescriptions Written and Total Payments Received

  • If doctors are writing prescriptions as quid pro quo for industry payments, one would expect that non-meal payments would show a correlation with prescribing similar to the correlation with meals shown in Figure 1.

Alternatively, if both attendance at educational sessions at which meals are served and opioid prescribing are driven by having a practice that involves treating many pain patients, one might expect a very modest or no correlation of prescribing with non-meal payments.

In practice, we see the latter (Figure 4).

Figure 4 was attracted with Hadland’s categorical fashion of demonstration to permit direct contrast to Figure 1. While Hadland discovered that same-sex tripled as the amount of industry-sponsored meals rose by one to eight, we detect no fad in toward improved prescribing among individuals who obtained between $0.01 and $65,536 from non-meal obligations from opioid manfacturers. In reality, the geometric mean speed was almost identical for people getting greater than $1 at non-meal obligations (711 prescriptions) and also for individuals getting $32,000 to $64,000 (718 prescriptions). For its 58 doctors who obtained over 65,536, the speed of alcoholism has been raised by almost twofold relative to individuals getting under a buck, but because of substantial in group sections, this difference wasn’t statistically significant.

The simple fact that same-sex correlates with the amount of foods approved but not using the entire number of non-meal payments obtained indicates that presence at educational events where foods are served along with opioid prescribing are equally driven by clinic characteristics. By comparison, these statistics are hard to accommodate within the concept that the institution of prescribing levels with foods approved is because of quid pro quo, or which firms will be bribing doctors to prescribe their products.

Figure 4. Geometric Mean Prescribing Rates by Total Non-Meal Payments Received

  • If doctors are writing prescriptions in exchange for free meals, one would not expect meals provided by the manufacturer of non-opioid pain treatment to be associated with increased opioid prescribing. If doctors with large pain practices are more likely to attend informational lunches about pain products, such an association is expected and natural.

In practicewe find the institution of raised opioid prescribing using attendance at informational lunches supplied by the producers of pain therapeutics is different from if the pain merchandise is a opioid!

St. Jude Medical is a medical device company that sells neuromodulation apparatus for treating chronic pain. Individuals who attended St. Jude lunches prescribed opioids in precisely exactly the exact identical speed as physicians who attended an equivalent quantity of lunches governed by opioid makers. This observation holds up nicely when looking just at people that attended St. Jude lunches but didn’t attend any psychiatric lunches. We found similar relationships with lunches supplied by producers of additional non-opioids goods (information not shown).

Figure 5. Relationship Between Attendance at Industry-Sponsored Lunches and Opioid Prescribing: St. Jude vs. Opioid Manufacturers


Correlation isn’t causation. When many advocates of decreased interactions involving”commercial” interests and doctors have indicated or directly indicated that a quid pro quo between business meals and other fiscal interactions and limiting customs, significance alone doesn’t establish that a quid pro quo connection. In the instance of same-sex, we think that we’ve presented a solid case that 1) the association between business obligations and prescribing is considerably weaker than was introduced in the literature( and two ) that provokes and presence in manufacturer-sponsored informational lunches will be equally driven by clinic characteristics, as opposed to the foods themselves forcing prescriptions (Type 0 connection ).

We think that a lot of what’s been released concerning the significance of prescribing with business payments and foods that are sponsored suffers from the shortcomings explained within this brief note. Specifically, a number of these newspapers conflate causation with significance. In scenarios where quite straightforward and obvious investigations would function to distinguish between the writers’ preconceptions and other interpretations of their data, these investigations haven’t been performed. We advocate having an interest in this region to strategy these data together using the greatest possible amount of objectivity, since is our duty as scientists. We’ve done our very best to do this here, and dedicate to doing this in our proposed investigations of different newspapers in this region.

We anticipate some stimulating debate with individuals who have additional information bearing on this problem, or alternative interpretations of the information presented herein.


Hadland SE, Cerdá M, Li Y, Krieger MS, Marshall BL. Association of pharmaceutical industry marketing of opioid products to physicians with subsequent opioid prescribing. JAMA internal medicine. 2018

[1]. This analysis, as well as alternative analyses performed by the present authors, was limited to the prescribing behavior of those who wrote at least ten opioid prescriptions in 2015 due to redaction of counts between 1 and ten by CMS.

About the authors

Rafael Fonseca is a hematologist at the Mayo Clinic in Arizona and John is a medicinal chemist residing in Northern California.


Dr. Fonseca – Consulting: AMGEN, BMS, Celgene, Takeda, Bayer, Jansen, AbbVie,
Pharmacyclics, Merck, Sanofi, Kite, and Juno.

Scientific Advisory Board: Adaptive Biotechnologies