Association between industry payments and prescriptions of long-acting insulin: An observational study with propensity score matching.

BACKGROUND: The rapidly increased spending on insulin is a major public health issue in the United States. Industry marketing might be one of the upstream determinants of physicians' prescription of long-acting insulin-the most commonly used and costly type of insulin, but the evidence is lacking. We therefore aimed to investigate the association between industry payments to physicians and subsequent prescriptions of long-acting insulin. METHODS AND
FINDINGS: Using the databases of Open Payments and Medicare Part D, we examined the association between the receipt of industry payments for long-acting insulin in 2016 and (1) the number of claims; (2) the costs paid for all claims; and (3) the costs per claim of long-acting insulin in 2017. We also examined the association between the receipt of payments and the change in these outcomes from 2016 to 2017. We employed propensity score matching to adjust for the physician-level characteristics (sex, years in practice, specialty, and medical school attended). Among 145,587 eligible physicians treating Medicare beneficiaries, 51,851 physicians received industry payments for long-acting insulin worth $22.3 million. In the propensity score-matched analysis including 102,590 physicians, we found that physicians who received the payments prescribed a higher number of claims (adjusted difference, 57.8; 95% CI, 55.8 to 59.7), higher costs for total claims (adjusted difference, +$22,111; 95% CI, $21,387 to $22,836), and higher costs per claim (adjusted difference, +$71.1; 95% CI, $69.0 to $73.2) of long-acting insulin, compared with physicians who did not receive the payments. The association was also found for changes in these outcomes from 2016 to 2017. Limitations to our study include limited generalizability, confounding, and possible reverse causation.
CONCLUSIONS: Industry marketing payments to physicians for long-acting insulin were associated with the physicians' prescriptions and costs of long-acting insulin in the subsequent year. Future research is needed to assess whether policy interventions on physician-industry financial relationships will help to ensure appropriate prescriptions and limit overall costs of this essential drug for diabetes care.

Introduction

Financial relationships between physicians and the pharmaceutical companies have received increased scrutiny, because while industry marketing often provides physicians with important updates on new research evidence, clinical guidelines, and new treatments, such relationships may also influence physicians’ prescription practices [1,2]. As a part of the Affordable Care Act, since 2013, the Open Payments program has mandated that medical product manufacturers publicly report all payments to physicians and teaching hospitals [3]. Previous studies have documented associations between industry payments and prescriptions for specific drugs including opioids [4,5] and cardiovascular drugs [6,7]. Some studies also showed that even the receipt of meals—the most frequent and inexpensive type of industry payments [8]—was associated with an increased rate of prescriptions of drugs from the sponsoring manufacturer [7]. Improving awareness about such associations for long-acting insulin analogs is needed because they are more expensive than other oral antihyperglycemic therapies and human insulin, which could be alternative options for some patients with type 2 diabetes [9,10], and unnecessary prescriptions of long-acting insulins that are not supported by clinical guidelines [11] should be avoided to reduce the risk of hypoglycemia and save costs. However, little is known as to whether the financial relationship between physicians and industry is associated with an increased rate of physicians’ prescriptions of long-acting insulin.

A study specific to insulin is important because Medicare Part D spending on insulin increased 840% from $1.4 billion in 2007 to $13.3 billion in 2017 [12,13], mainly due to the rapidly increasing trajectory of insulin prices and the number of Medicare beneficiaries using insulin in the United States [13-16]. In response to the increased spending and limited affordability of this expensive but essential drug for diabetes care, the Centers for Medicare and Medicaid Services (CMS) recently announced a voluntary model under which some Part D plans and pharmaceutical manufacturers would come together to lower Medicare beneficiaries’ out-of-pocket cost for insulin to $35 per month, starting January 1, 2021 [17]. In addition, pharmaceutical companies have recently started marketing follow-on biologics that are less expensive and similar types to their brand-name insulin products [18]. Since physicians are the conduit between insulin suppliers and patients with diabetes, it is vital to understand the dynamics of insulin prescribing within a national sample of physicians. Specifically, it is critically important to investigate whether the financial relationship between physicians and pharmaceutical companies is one of the upstream determinants of physicians’ prescription of long-acting insulin—the most commonly marketed, commonly used, and most costly type of insulin [9,19,20].

Therefore, using physician databases from the Open Payments program and Medicare Part D, we examined the association between physicians’ receipt of industry payments and their subsequent prescription of long-acting insulin. To assess whether any association varies based on the number of industry marketing encounters, particularly in the form of meals, we also examined the dose–response relationship between the number of meals sponsored by the pharmaceutical companies and prescriptions of long-acting insulin.

Methods

Data sources

This study linked 4 publicly available databases: (i) the CMS Open Payments data for 2016 [21]; (ii) the CMS National Plan & Provider Enumeration System (NPPES) database [22]; (iii) the CMS Physician Compare database [23]; and (iv) the CMS Medicare Provider Utilization and Payment database for 2017 [24]. To address temporality, we used industry payments data in 2016 and prescription data in 2017. The prespecified analysis plan is described in .

First, using exact matching of physicians’ full name and zip code, we linked the Open Payments database that contains physician-level data on industry payments they received with the NPPES database that contains physicians’ National Provider Identifier (NPI) as previous studies did [8,25]. Under the Physician Payment Sunshine Act, biomedical companies are required to report the information about payments to physicians to CMS, and physicians are encouraged to review the information before publication and dispute submitted reports if correction is needed [21]. Using physicians’ NPI, the merged database of the Open Payments and NPPES databases was linked to the Physician Compare database that contains physician characteristics (i.e., sex, years in practice, specialty, and medical school attended) and the Medicare Provider Utilization and Payment database that contains physician-level data on drug utilization (e.g., the number of claims including refills, the aggregate cost paid for all claims, etc.). To ensure that physicians included in this study have the potential to prescribe long-acting insulin, we restricted to those who prescribed at least 1 antihyperglycemic therapy in both 2016 and 2017. Further details regarding the derivation of the study population are provided in .

Payment data

We identified and extracted all non-research payments for long-acting insulin in the Open Payments database in 2016. The categories of payments include food and beverage (i.e., meals), travel and lodging, speaker compensation or honoraria, and others (e.g., consulting fees, gifts, educational materials, etc.). When a single payment was associated with multiple drugs (up to 5), we divided the amount and number of the payment equally by the number of reported drugs.

Prescription data

Our primary outcomes were (i) the total claims of long-acting insulin in 2017 (30-day standardized, including refills); (ii) the total costs paid for claims of long-acting insulin in 2017; and (iii) the costs per claim of long-acting insulin in 2017, calculated by dividing total costs by the number of claims. Our secondary outcomes were changes in these 3 outcomes from 2016 to 2017. To protect the privacy of Medicare beneficiaries, records with fewer than 11 claims were not included in Medicare Provider Utilization and Payment Data [24].

Physician characteristics

We extracted data on physicians’ sex, years in practice (estimated from years since graduation of medical school), and the medical school graduated from the Physician Compare database. We categorized medical schools into 3 groups according to the 2017 U.S. News & World Report, which lists the research ranking of US medical schools in 2017 [26]: ranked 1 to 20, 21 to 50, and others (including all unranked and foreign medical schools). Physicians’ specialty was classified into primary care physicians, endocrinologists, cardiologists, nephrologists, pulmonologists, infectious disease physicians, allergy/immunology physicians or rheumatologists, gastroenterologists, surgeons, and other specialties, using taxonomy codes from the NPPES databases.

Propensity score matching

We employed nearest-neighbor propensity score matching without replacement to 1:1 match physicians who received industry payments for long-acting insulin with those who did not receive these payments. We used a logistic regression model to derive the propensity scores for the receipt of payments in 2016 that included physician characteristics (sex, years in practice, medical school attended, and specialty) which were considered to be associated with the receipt of industry payments and prescriptions. We used a caliper of 0.20 standard deviation of the logit of the propensity score and evaluated the balance between the 2 comparison groups using the standardized mean difference across covariates (<10% indicated successful balance) [27].

Statistical analysis

We described baseline physician characteristics before and after propensity score matching according to whether they received industry payments for long-acting insulin in 2017. Then, after confirming that the matching balanced potential confounders, we used a paired t test to compare matched pairs of physicians with regard to the total claims, the total costs, and the costs per claim for long-acting insulin in 2017. We also conducted a paired t test for the change in these outcomes from 2016 to 2017. Finally, because meals are the most common and inexpensive type of industry payments, we examined the association between the number of meals received in 2016 (as continuous; per 5 meals to calculate interpretable estimates) and the total claims, the total costs, and the costs per claim for long-acting insulin in 2017, using the entire cohort prior to propensity score matching. Given the possible nonlinear relationship, we reanalyzed the data using the 6 categories of the number of meals received in 2016 (0, 1, 2 to 5, 6 to 10, 11 to 15, and 16+) instead of continuous variables. In these analyses, we employed multivariable negative binomial regression models adjusting for the same covariates used in the propensity score matching. The trend test was conducted using the median number of meals received in each category as a continuous variable.

Sensitivity analyses

Given the possible bias due to residual confounding even after propensity score matching, we reanalyzed the propensity score–matched samples using ordinary least squares (OLS) regression models with Huber–White robust standard errors adjusting for the covariates used to estimate the propensity scores. In addition, to address the possible model misspecification (i.e., to account for the potential right-skewed distributions of residuals), we also conducted the analysis using negative binomial regression models adjusting for the same variables.

Additional analyses

Using the entire cohort prior to propensity score matching, we conducted the following 3 additional analyses. First, to address the potential reverse causality issue (e.g., pharmaceutical companies might have targeted physicians with a large number of prescriptions of antihyperglycemic therapies), we additionally adjusted for the total number of Medicare Part D claims of antihyperglycemic therapies in 2016. Second, as physicians who prescribed long-acting insulin before 2017 might have been more likely to receive industry payments for long-acting insulin in 2016, we reanalyzed the data restricting physicians to those who did not prescribe long-acting insulin in 2016. Third, given the potential different distribution of claims and costs across long-acting insulin products (i.e., Lantus, Levemir, Toujeo, Tresiba, and Basaglar), we investigated the association between the receipt of industry payments and prescriptions for each insulin product.

All analyses were performed using Stata, version 15 (Stata Corp, College Station, Texas, United States of America). After fitting regression models, we calculated the predictive value of each outcome under hypothetical exposure levels with the observed distribution of covariates using margins command in Stata [28]. The study was exempted by the institutional review board at the University of California, Los Angeles (IRB#18–001960).

Results

Physician characteristics are shown in . Physicians who received payments in 2016 were more likely to be male, endocrinologists, and have more years in practice. They were less likely to have graduated from 1 to 50 ranked US medical schools. The 2 treatment groups of physicians were well balanced on all baseline covariates after propensity score matching (i.e., standardized differences were <1% for all covariates). The distribution of the propensity scores across the 2 treatment groups is shown in .

Among 145,587 physicians who prescribed at least 1 antihyperglycemic therapy for Medicare beneficiaries in 2016 and 2017 and had physician-level information (i.e., our entire study sample before propensity score matching), industry payments for long-acting insulin worth $22.3 million were made to 51,851 physicians in 2016. Food and beverages were the most common type of payments (96%), and speaking fees constituted the largest value ($12,827,441) (). In 2017, 13.3 million claims with a total cost of $4.7 billion were filed for long-acting insulin by 102,960 physicians (71% of 145,587) caring for Medicare beneficiaries included in this study ().

Industry payments in 2016 and prescription in 2017

Among 102,590 propensity score–matched physicians, we found that physicians who received industry payments for long-acting insulin in 2016 were associated with a larger number of total claims in 2017 (adjusted difference, 57.8; 95% CI, 55.8 to 59.7), higher costs paid for total claims in 2017 (adjusted difference, +$22,111; 95% CI, +$21,387 to +$22,836), and higher costs per claim (adjusted difference, +$71.1; 95% CI, +$69.0 to +$73.2) of long-acting insulin in 2017, compared with physicians who did not receive payments in 2016 (). We also found the association between the receipt of payments and change in the following 3 outcomes from 2016 to 2017: change in claims (adjusted difference, 2.4; 95% CI, 1.8 to 3.0), change in costs paid for total claims (adjusted difference, +$1,209; 95% CI, +$976 to +$1,442), and change in costs per claim (adjusted difference, +$2.4; 95% CI, 0.6 to 4.1) (). We observed a reduction of costs per claim of long-acting insulin from 2016 to 2017 among both groups of physicians.

We found the association between the number of industry-sponsored meals for long-acting insulin received in 2016 (per 5 meals) and the total claims (adjusted relative ratio [95% CI] = 1.241 [1.235 to 1.246]), the total costs (adjusted relative ratio [95% CI] = 1.268 [1.262 to 1.273]), and the costs per claim (adjusted relative ratio [95% CI] = 1.094 [1.090 to 1.099]) for long-acting insulin in 2017, after adjusting for physician characteristics. We found the consistent results for all outcomes (p for trend <0.001) when using 6 categories of the number of industry-sponsored meals for long-acting insulin received in 2016 ().

Association between the receipt of industry payments and claims of long-acting insulin by the number of meals.

Negative binomial regression model was employed to adjust for physicians’ sex, years in practice, specialty, and medical school attended for (A) number of claims, (B) cost paid for all claims, and (C) costs per claim of long-acting insulin in 2017 by the number of meals received in 2016.

The results did not substantially change when we reanalyzed the propensity score–matched samples using the OLS regression model () and the negative binomial regression model (), adjusting for covariates used to estimate the propensity score.

The association was also observed when we additionally adjusted for the total claims of antihyperglycemic therapies (adjusted difference in claims, 3.9; 95% CI, 2.7 to 5.1; adjusted difference in costs, +$2,820; 95% CI, +$2,348 to +$3,294; adjusted difference in costs per claim, +$52.3; 95% CI, +$50.4 to +$54.3; ) and restricted the analysis to physicians who did not prescribe long-acting insulin in 2016 (adjusted difference in claims, 3.6; 95% CI, 3.1 to 4.2; adjusted difference in costs, +$1,347; 95% CI, +$1,147 to +$1,546; adjusted difference in costs per claim, +$33.1; 95% CI, +$28.8 to +$37.4; ). Across the brands of long-acting insulin, the association was consistently found with the largest absolute difference for Lantus followed by Levemir, and the largest relative ratio for Toujeo followed by Tresiba ().

Discussion

Using the US national database of physicians treating Medicare beneficiaries, we found that the receipt of industry payments for long-acting insulin in 2016 was associated with increased numbers of claims, costs paid for all claims, and costs per claim of long-acting insulin in 2017. We also found a consistent association for the change in these outcomes from 2016 to 2017. Meals constituted 96% of the total number of industry payments for long-acting insulin, and the number of meals received in 2016 had a dose–response relationship with the utilization of long-acting insulin in 2017.

Prior studies have demonstrated similar findings for antihyperglycemic medications other than insulin. A previous ecological cross-sectional study showed an association between industry payments and regional prescription of marketed non-insulin antihyperglycemic therapies [29]. Another cross-sectional study showed that the receipt of industry payments for the combination drug of saxagliptin and metformin was associated with the increased odds of prescribing this drug [30]. However, these studies used Open Payments data in 2014. The Open Payments program started in 2013, and its data accuracy has improved since 2014 [3], suggesting the need to establish these relationships in updated data. Furthermore, whether these findings also apply to long-acting insulin has not previously been established. In this context, our findings using more recent databases with well-defined temporality (i.e., payments in 2016 and prescription in 2017) and physician-level characteristics, along with propensity score matching, provide new robust evidence about the potential impact of industry marketing on prescriptions of long-acting insulin—a key driver of increased spending in the healthcare system in the US [20].

The influence of the physician–industry financial relationship on clinical practice for diabetes care involving insulin has been of great concern under the regulating system for pharmaceutical manufactures setting insulin prices and direct competition in the insulin market [16,31]. In this context, our findings focusing on long-acting insulin (which constitutes a large proportion of insulin prescriptions and expenditures in the US [20]) generate a hypothesis that physician–industry financial relationship may increase the overall costs of long-acting insulin for patients with diabetes, which works against the national effort to reduce the patients’ financial burden of using insulin. Although our data do not adjudicate the appropriateness of the prescription, it may also be possible that industry marketing payments introduce physicians’ preference of prescribing long-acting insulin rather than cheaper, oral antihyperglycemic therapies that may be equally effective in some clinical situations, given the prior literature on this topic [4-7]. More investigations are warranted for long-acting insulins to avoid unnecessary prescriptions of this costly and common drugs that affect millions of Americans with diabetes.

Of note, costs per claim paid for long-acting insulin generally decreased from 2016 to 2017 with the larger reduction among physicians who did not receive industry payments. This indicates that there is still an opportunity for physicians whose behavior may be influenced by industry payments to reduce costs for long-acting insulin or at least help hold the line against additional cost increases. In December 2015, a follow-on insulin glargine (Basaglar [Lilly]) received US Food and Drug Administration approval [32]. Although our data in 2017 included only a few physicians who prescribed Basaglar and was not sufficient to make an argument specific to this drug, future studies are needed to understand whether programs without conflicts of interest such as academic detailing [33] or other educational programs focusing on the follow-on insulin are effective to reduce the overall insulin costs by providing unbiased third-party information on long-acting insulins to physicians.

It is noteworthy that we found the significant association between industry payments and utilization of long-acting insulin even when additionally adjusting for total claims of antihyperglycemic therapies or restricting the physician sample to those who did not prescribe long-acting insulin in 2016. Although the magnitude of the association was smaller than our main findings, these results, along with the results from the analyses for change in outcomes from 2016 to 2017, support the presence of direct impact of industry payments on prescriptions and costs of long-acting insulin.

Our findings also suggest that the frequency of industry marketing encounters, even with inexpensive payments such as meals, may contribute to the increased prescriptions and spending for long-acting insulin. A previous study showed that physicians who specifically received meals were more likely to prescribe the marketed brand-name medications of the pharmaceutical manufacturer that provided the meals [7]. Given that 96% of industry payments for long-acting insulin were meals and almost all physicians who received other types of payments also received meals in the present study, these findings indicate the importance of interventions that refocus even industry marketing encounters with small payments to promote appropriate prescribing that reduces unnecessary spending on long-acting insulin.

Limitations of the study

Several limitations of our analysis should be considered. As we used the Medicare Part D database, our findings may not be generalizable to non-Medicare beneficiaries. Given the observational study design, it is possible that our models had residual confounding even though we adjusted for several important physician-level variables. In addition, as the databases did not contain patient-level information, it was possible that patients’ characteristics such as severity of diabetes and complications might be different between groups of physicians. We did not have access to individual-level data on other forms of marketing that potentially impact physicians’ prescribing behavior [34] including funded research, ownership interests, investments in start-up entities, and other physician–industry interactions without direct payment transfers [35]. In the Open Payments database, each payment listed up to 5 products related to the payment, and, thus, we had a possibility of misclassification due to payments related to more than 5 products including long-acting insulin. However, such misclassification, if any, would not change our findings because 99.9% of payments listed less than 5 products in our final analytical sample. Lastly, even though we clarified the temporality by using payments data in 2016 and prescription data in 2017, some physicians might have attended industry events where information about long-acting insulin was provided because they were interested in prescribing specific long-acting insulin medications. In such a scenario, we might have the risk of reverse causation.

Conclusions

Among physicians treating Medicare beneficiaries in the US, industry marketing related to long-acting insulin in the prior year was associated with higher utilization of long-acting insulin in the subsequent year. The number of meals had a dose–response relationship with prescriptions for long-acting insulin, suggesting that industry marketing encounters, even those at relatively low dollar amounts, may influence physicians’ use of long-acting insulin in clinical practice. Future research is needed to investigate whether policy interventions on the financial relationships between physicians and industry will be an effective strategy to ensure appropriate prescriptions of long-acting insulin.

Ethical approval

The study was exempted from human subjects review by the institutional review board at University of California, Los Angeles.

Transparency statement

The corresponding author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

STROBE checklist.

(DOC)

Click here for additional data file.

Data analysis plan.

(DOCX)

Click here for additional data file.

Flow diagram showing derivation of the study population.

(TIF)

Click here for additional data file.

The distribution of the propensity scores across the 2 treatment groups.

(TIF)

Click here for additional data file.

Association between the receipt of industry payments in 2016 and (A) the number of claims or (B) the cost in 2017 for each long-acting insulin.

(TIF)

Click here for additional data file.

Number (%) of physicians who prescribed long-acting insulin in 2016 and 2017.

(DOCX)

Click here for additional data file.

Association between the receipt of industry payments for long-acting insulin in 2016 and claims of long-acting insulin in 2017 using OLS regression model adjusting for physician characteristics. OLS, ordinary least squares.

(DOCX)

Click here for additional data file.

Association between the receipt of industry payments for long-acting insulin in 2016 and claims of long-acting insulin in 2017 using negative binomial regression model adjusting for physician characteristics.

(DOCX)

Click here for additional data file.

Association between the receipt of industry payments for long-acting insulin in 2016 and claims of long-acting insulin in 2017 using OLS regression model additionally adjusting for total claims of antihyperglycemic therapy in 2016. OLS, ordinary least squares.

(DOCX)

Click here for additional data file.

Association between the receipt of industry payments for long-acting insulin in 2016 and claims of long-acting insulin in 2017 using OLS regression model adjusting for physician characteristics, restricting physicians who did not prescribe long-acting insulin in 2016.

OLS, ordinary least squares.

(DOCX)

Click here for additional data file.

2 Sep 2020

Dear Dr Inoue,

Thank you for submitting your manuscript entitled "Association Between Industry Payments and Prescriptions of Long-acting Insulin: An Observational Study with Propensity-Score Matching" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff as well as by an academic editor with relevant expertise, and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by .

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

7 Dec 2020

Dear Dr. Inoue,

Thank you very much for submitting your manuscript "Association Between Industry Payments and Prescriptions of Long-acting Insulin: An Observational Study with Propensity-Score Matching" (PMEDICINE-D-20-04080R1) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer (r#1). The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Dec 28 2020 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Emma Veitch, PhD

PLOS Medicine

On behalf of Caitlin Moyer, PhD, Associate Editor,

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

*At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

*In the last sentence of the Abstract Methods and Findings section, please include a note about any key limitation(s) of the study's methodology.

*Please clarify whether the analytical approach followed here corresponds to one laid out in a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

*Because papers are repaginated for publication, please update the STROBE checklist using section and paragraph numbers, rather than page numbers. The authors might also want to consider using instead the RECORD guideline, if they feel that is appropriate (https://www.equator-network.org/reporting-guidelines/record/) as this was designed to support reporting of studies done using routinely collected health data (and/or registry linkage methods, which seem to have been used in your study).

*Due to PLOS Medicine policy/style, please remove trademark symbols from the manuscript and supporting information files.

-----------------------------------------------------------

Comments from the reviewers:

Reviewer #1:

Thanks for the opportunity to review your manuscript. My role is as a statistical reviewer so my comments are focused on the study design, data, and analysis. This is an interesting manuscript that has made good use of some routinely collected healthcare data. I have put overall comments and queries first and then followed these by queries relating to specific parts of the manuscript (with a page and paragraph reference).

I didn't feel as though there was enough background given to clarify the ramifactions of a switch to long-term insulin for both patients and health systems. My understanding (as a non-expert who has worked with similar data) is that long-term insulin could in fact translate to better outcomes for patients (and also potentially worse outcomes). Is a switch to long-term insulation always inappropriate with current guidelines? Is it possible that although costs are higher, this translates into appropriate expenditure for the improvement in patient outcomes?

My other main query was the interpretation of the results - in the main results the difference in claims (134 vs 77) and costs ($48k vs $26k) between the two groups is quite stark. When adjusted for total claims of antihyperglycemic medicines the difference diminishes (93.4 vs 89.5; and $33k vs $31k). This needs further discussion - this would seem to imply that the main mechanism for the increases in claims and costs comes from the industry funding influencing overall activity around antihyperglyemic therapy which doesn't seem to make sense on face value. Could patient load (i.e. number of patients with diabetes per physician) be on the causal pathway here and not adjusted for during the matching process and leading to the discrepancy between the main results and this particular sensitivity analysis?

P6. Paragraph 2. It wasn't clear that direct matching (i.e. exact matching of the names and ZIP) was used even after checking the two references - could you just add whether it was exact matching or a details of statistical linkage key if the info was used this way?

Does the MPUP database record medication dispensings or prescriptions?

P7, Paragraph 1. What proportion of the payments were single-drug, and what proportion associated with a multi-drug record? Were payments for more than five drugs excluded at this step or was 5 the limit that payments were adjusted for?

P8. Did you use the 0.2 SD of the logit of the PS, or PS on the original scale? Where is the balance information? Did you include interactions between these variables?

What was the process of considering which physician characteristics were considered to be important confounders of the association between industry payments and long-term insulin prescribing?

Was there any missing data in the variables used to create the propensity scores? If so, how was this dealt with?

What was the criteria used for the categories in the trend test? Why was this used over the original value of the meals received variable?

P9. Sensitivity analyses. I assume that the negative binomial sensitivity analysis was used as cost data often has an inconvenient distribution (and this should be clarified in the methods), was the glm with sandwich errors (H-W) a way to estimate a model with a normal distribution while accounting for the matching?

P10, Methods. Was psmatch2 or teffects used for the matching?

Was common support assessed? A figure of PS distribution by exposure group would be a helpful addition to the appendix.

What method was used to get the estimates of claims, costs etc. per exposure group? Margins in Stata? Was this set at particular levels of the covariates, or at the average for the entire cohort?

P10, Results, Paragraph 1. Is it worth adding that the standardised differences after matching were below the target (and in fact are all below 1% different?)

P11/P12. Is it possible (given space limitations) to report the key effect estimates from the sensivity analyses in the results? There is still a difference between the two exposures groups when there is adjustment for total claims of antihyperglycemic therapies but the magnitude of the difference is greatly reduced.

S1. Figure 1. This is an important diagram understanding how the data for the study was created. To clarify, there were ~5,000 physicians in the open payments database who weren't able to be matched to the NPPES data? Similarly, ~43,000 providers who had prescribed at least one antihyperglycaemic in 2016/2017 weren't matched to the physician Compare database? Also, there were ~60228 physicians who had received payments but then only 51,851 that had been classed as having done this in the final database - were these physicians that couldn't be linked with the NPI?

I appreciate that this is already a detailed diagram but if there is a way to include information on the numbers excluded at each step this would be very helpful, with linked administrative data the quality of the linkage process is very important for study validity and this would help me assess this aspect of the study.

-----------------------------------------------------------

Reviewer #2:

This manuscript investigates the association between payments to physicians and prescribing of long-acting insulin. It adds to the growing literature showing that industry payments have an influence on the way that doctors prescribe. My comments are of a relatively minor nature:

1. Besides the increase in the price of insulin Medicare spending could also have increased because of more patients using insulin.

2. Page 4, 2nd paragraph, 6th line: The correct word should be "companies" not "industries".

3. Rather than using references 13-15 for the influence of promotion on physicians' prescribing behaviour, the authors should cite the systematic review by Spurling et al in PLoS Medicine that was published in 2010.

4. Besides academic detailing what other types of policy interventions are the authors proposing to regulate financial interactions between physicians and the pharmaceutical industry?

Joel Lexchin

-----------------------------------------------------------

Reviewer #3:

This study examines a potentially interesting issue—the association between industry payments and prescription of long-acting insulin. The authors do a good job of situating the study in relation to prior work in this area. As they rightly note, while there have been a number of studies that explore and find associations between industry payments and prescribing in other areas of medicine, relatively little is known about these dynamics in diabetes care. A few similar studies of diabetes care have relied on older (2014) data. Overall, then, the study has the potential to make an incremental but meaningful addition to the literature on financial COI and their implications for medicine.

However, I have two broad concerns about the framing and conceptualization of the study.

First, the authors introduce and motivate the paper with a concern about the rising cost of insulin in the US. This is well-established and a genuine policy problem. But they seem to suggest—implicitly throughout and explicitly at times—that the rising cost of insulin is influenced by physician-industry relationships and the lack of transparency around these relationships. For instance, in the Discussion, they write that "Key factors related to the rise in insulin prices over the past two decades include…the lack of transparency in such financial relationships between physicians and industry for insulin." On the next page, they appear to fault the ADA Working Group for failing to mention "the industry marketing payments asfactors in the escalating cost of insulin."

However, the authors do not propose a mechanism by which industry payments to physicians, or the higher rates of prescribing that they might induce, would lead to higher drug prices. As such, the findings presented in the study, which show an association between industry payments and prescribing, do not appear to be plausibly linked to the policy problem of interest, namely, the rising cost of insulin. In this sense, the bookends of the paper seem to be dealing with a different problem (i.e., the cost of insulin) than the heart of the paper (i.e., determinants of insulin prescribing). The author should either reframe the Introduction and Discussion sections, or else specify the mechanism by which they believe industry payments to physicians affect insulin prices.

Second, and somewhat relatedly, it isn't clear how we ought to interpret the implications of higher rates of long-acting insulin prescribing. Typically, in these sorts of studies, investigators are interested in the associations between industry payments and prescribing because of an underlying concern that payments might influence physicians to prescribe in ways that are clinically suboptimal or clinically equivalent but costlier. So, for instance, in DeJong et al. (JAMA IM 2016)—which this manuscript cites—the authors find that industry payments are associated with higher rates of prescribing brand name drugs when generics are available.

Do higher rates of long-acting insulin prescribing raise similar concerns? It isn't clear from the manuscript. One concern could be that physicians are prescribing an expensive drug, long-acting insulin, when a cheaper alternative would do. But in the Discussion, the authors say that "long-acting insulin is an essential drug and cannot be easily replaced for many patients" so that would seem not to be the case. Later, however, they say that detailing programs "focusing on a less expensive version of long-acting insulin" may be appropriate. If such alternatives exist, that should be stated earlier in the paper. More importantly, if there are both expensive and cheap versions of long-acting insulin, then this study's design would seem to be seriously flawed. In that case, the research question should be whether receipt of industry payments is associated with prescription of the more expensive type of long-acting insulin. But the question that this actually asked here is whether receipt of industry payments is associated with higher rates of prescribing any long-acting insulin.

I suppose another reason to be concerned about higher rates of long-acting insulin prescribing could be that physicians are overprescribing insulin. This is the underlying concern, for example, in recent studies of the association between industry payments and opioid prescribing. But that seems implausible in the case of insulin, and the authors provide no evidence to that effect.

Perhaps some of these issues involving the availability of clinical alternatives and the implications of higher rates of long-acting insulin prescribing are obvious to specialists, but as this is a general medical journal, they would need to be more clearly spelled out.

Overall, these two issues—1) that the authors don't provide a plausible mechanism by which industry payments to physicians drive higher insulin prices, and 2) that the author don't otherwise clarify why higher rates of long-acting insulin are problematic—raise questions about the policy or clinical relevance of the study's findings.

-----------------------------------------------------------

Any attachments provided with reviews can be seen via the following link:

[LINK]

10 Jan 2021

Submitted filename: Response letter PLOSMED final.docx

Click here for additional data file.

4 May 2021

Dear Dr Inoue,

On behalf of my colleagues and the Academic Editor, Aaron S Kesselheim, I am pleased to inform you that we have agreed to publish your manuscript "Association Between Industry Payments and Prescriptions of Long-acting Insulin: An Observational Study with Propensity-Score Matching" (PMEDICINE-D-20-04080R2) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Dr Raffaella Bosurgi

Executive Editor

PLOS Medicine

Table 1

Physician characteristics who prescribed antihyperglycemic therapies in 2016 and 2017 according to whether they received industry payments for long-acting insulin in 2016.

	Cohort before propensity matchinga			Cohort after propensity matchinga
	Physicians who received payments	Physicians who did not receive payments	Standardized difference, %	Physicians who received payments	Physicians who did not receive payments	Standardized difference, %
Number of physicians	51,851	93,736		51,295	51,295
Sex, N (%)
Female	32.0	39.1	−14.9	32.4	32.1	0.7
Male	68.0	60.9	14.9	67.6	67.9	−0.7
Years in practice, mean	25.2	23.8	11.2	25.0	25.1	−0.6
Specialty, %
Primary care	92.2	87.0	16.9	93.2	93.2	−0.0
Endocrinology	4.6	2.0	14.5	3.5	3.5	0.0
Cardiology	0.8	2.8	−14.9	0.8	0.8	0.1
Nephrology	0.7	2.1	−12.2	0.7	0.7	−0.1
Pulmonology	0.3	0.7	−5.2	0.3	0.3	0.2
Infectious disease	0.2	0.8	−8.2	0.2	0.2	0.1
Allergy/immunology and rheumatology	0.2	0.5	−6.3	0.2	0.2	0.0
Gastroenterology	0.2	0.5	−5.5	0.2	0.2	0.0
Surgery	0.3	1.0	−9.0	0.3	0.3	−0.1
Other specialtiesb	0.6	2.6	−15.8	0.6	0.6	0.0
Medical school attendedc, %
Top 20 schools in the US	4.2	8.2	−16.7	4.3	4.3	−0.1
US schools ranked 21 to 50	13.1	16.7	−10.0	13.3	13.3	0.1
Other schools	82.7	75.1	18.6	82.5	82.5	0.0

aOne-to-one pair matching was performed by nearest-neighbor matching without replacement, with the use of a caliper width equal to 0.2 of the standard deviation of the logit of the propensity score. A standardized difference of less than 10.0% was considered to indicate a negligible imbalance between the 2 groups.

bOther specialties include hematology/oncology, neurology, ophthalmology, dermatology, radiology, anesthesiology, etc.

cTop 20 schools and Top 21 to 50 schools in US were defined based on the U.S. News & World Report ranking (research ranking) in 2017. Other schools include foreign medical schools.

Table 2

Characteristics of payments involving long-acting insulin in 2016.

Nature of payments	Number of paymentsa	Number of physicians	Total payment amount, $	Median value of the payment, $ (IQR)
Food and beverages	427,841	51,851	$7,030,045	$15 (12 to 18)
Travel and lodging	6,398	579	$1,653,082	$197 (102 to 320)
Speaking fees or honoraria	6,458	645	$12,827,441	$1,880 (1,449 to 2,227)
Othersb	4,093	2,726	$801,289	$5 (4 to 14)

aThe proportion of payments for single drug and multiple drugs were 67% and 33%, respectively. Only 0.1% of payments were related to 5 products.

bOthers include education, grant, gifts, entertainment, and space rental or facility fees.

IQR, interquartile range.

Table 3

Association between the receipt of industry payments for long-acting insulin in 2016 and claims of long-acting insulin in 2017 adjusting for physician characteristics.

	Physicians who received industry payments for long-acting insulin in 2016	Physicians who did not receive industry payments for long-acting insulin in 2016	P value
Claims of long-acting insulin in 2017
Mean (95% CI)	134.5 (133.0 to 136.1)	76.8 (75.6 to 77.9)	<0.001
Difference (95% CI)	57.8 (55.8 to 59.7)
Costs paid for all claims of long-acting insulin in 2017
Mean (95% CI)	$48,201 (47,606 to 48,797)	$26,090 (25,677 to 26,502)	<0.001
Difference (95% CI)	$22,111 (21,387 to 22,836)
Costs per claim of long-acting insulin in 2017a
Mean (95% CI)	$300.4 (299.0 to 301.8)	$229.3 (227.8 to 230.9)	<0.001
Difference (95% CI)	$71.1 (69.0 to 73.2)

aEstimated by [costs paid for all claims of long-acting insulin in 2017]/[number of all claims of long-acting insulin in 2017]. No claims were replaced as zero.

CI, confidence interval.

Table 4

Association between the receipt of industry payments for long-acting insulin in 2016 and change in claims of long-acting insulin from 2016 to 2017 adjusting for physician characteristics.

	Physicians who received industry payments for long-acting insulin in 2016	Physicians who did not receive industry payments for long-acting insulin in 2016	P value
Change in claims of long-acting insulin from 2016 to 2017
Mean (95% CI)	4.4 (3.9 to 4.9)	2.0 (1.6 to 2.4)	<0.001
Difference (95% CI)	2.4 (1.8 to 3.0)
Change in costs paid for all claims of long-acting insulin from 2016 to 2017
Mean (95% CI)	$1,707 (1,520 to 1,894)	$498 (358 to 637)	<0.001
Difference (95% CI)	$1,209 (976 to 1,442)
Change in costs per claim of long-acting insulin from 2016 to 2017a
Mean (95% CI)	−$1.2 (−2.4 to −0.6)	−$3.6 (−4.9 to −2.3)	0.008
Difference (95% CI)	$2.4 (0.6 to 4.1)

aEstimated by [costs paid for all claims of long-acting insulin in 2017]/[number of all claims of long-acting insulin in 2017] − [costs paid for all claims of long-acting insulin in 2016]/[number of all claims of long-acting insulin in 2016]. No claims were replaced as zero.

CI, confidence interval.