Medical-Claims Database Analysis of Off-Label Prescribing

June 27, 2010
By Jill Van Den Bos

Daniel Perlman, Milliman Inc

Pharmaceutical Commerce, Pharmaceutical Commerce - May/June 2010,

Examining off-label use of highly prescribed drugs reveals factors that differ from the usual criticism of such usage

Off-label prescribing has made a splash in the news in recent months. A September 2009 settlement of a lawsuit against pharmaceutical giant Pfizer is once again bringing attention to the practice of off-label prescribing and marketing of pharmaceuticals. The settlement resulted from allegations that Pfizer improperly marketed its drug Bextra, in addition to several others, for indications other than those approved by the FDA.

There has been much recent research focused on finding drugs used most frequently for off-label uses. We wondered instead how often the drugs with the highest claim costs were used for apparent off-label indications. Our informal research presented below provides some interesting insight into the clinical characteristics of patients who receive drugs associated with high claim costs.


Off-label use is not illegal. Once a drug has been approved by the FDA, physicians are free to prescribe it as they see fit, even for non-approved indications. And there are often good reasons to do so. For example:

• Rare diseases may not have many good therapeutic options on the market, making drugs for similar conditions the best option in some cases. • Older drugs that have lost patent protection may have been demonstrated to be effective at treating a condition other than its official indication. But without patent protection it is unlikely that its maker could recoup the costs of the clinical trials needed to gain a new approved indication. • If a patient has not responded to conventional therapy, there may be little to lose from trying a drug that is being prescribed off-label. • Some demographic groups, especially children and pregnant women, are frequently not included in clinical trials. Often, there are few approved therapies available for such populations, and prescribing drugs for non-approved patients is considered an off-label use.

While physicians are legally permitted to prescribe a drug for conditions for which it has not been approved, pharmaceutical companies are not permitted to market a drug for unapproved uses. Allegations of off-label marketing have proved costly to a number of pharmaceutical companies in recent years, even before Pfizer’s recent settlement over Bextra. Other drugs that have recently been the subject of litigation over off-label marketing include Zyprexa, Actiq and Provigil.

While physicians are legally permitted to prescribe a drug for conditions for which it has not been approved, pharmaceutical companies are not permitted to market a drug for unapproved uses. Allegations of off-label marketing have proved costly to a number of pharmaceutical companies in recent years, even before Pfizer’s recent settlement over Bextra. Other drugs that have recently been the subject of litigation over off-label marketing include Zyprexa, Actiq and Provigil.

Research has suggested that, especially for some types of drugs, off-label prescribing is very common. A 2006 study published in the Archives of Internal Medicine analyzed physician-reported prescription data and found that 21% of the prescriptions studied were for off-label uses.[1] A study of Georgia Medicaid patients who received antidepressant, anticonvulsant, or antipsychotic medications found that 75%, 80%, and 64%, respectively, had at least one claim for apparent off-label use.[2] A study by the New England Research Institutes estimated antidepressant use for off-label indications at 42% using a commercially insured population.[3]

Researchers often criticize off-label prescribing, questioning the thought put into the decision to choose off-label therapy. For example, a survey conducted among physicians determined that they could correctly identify actual FDA-approved diagnoses for a set of commonly prescribed drugs only a little more than half the time. In that survey, only 32% of physicians said that FDA labeling is “very important” in guiding prescribing patterns, with 47% saying it was “somewhat important.”[4]

A 2008 study of off-label prescribing attempted to identify those drugs most widely prescribed off-label with an eye toward creating a list of the top drugs in need of additional research on their off-label uses.[5] Criteria for inclusion on the list included high volume off-label use with inadequate evidence for these apparent indications, drug safety, and cost/marketing considerations. Using these criteria, researchers identified the 14 drugs listed in Fig. 1.

Rather than looking for drugs with indications of high percentage off-label use, and then looking for high-frequency prescribing among those drugs, we will do the opposite: look first for highest claim-cost options among all drugs, then look for what appears to be off-label use among these drugs. This, we reason, will lead us to drugs likely to interest health plans in terms of their off-label use because it occurs among their high-cost drugs and therefore potentially affects a substantial portion of their covered populations.

Data Analysis

To look for evidence of how often high-cost drugs are prescribed for off-label uses, we examined the 2007 Medstat MarketScan commercial and Medicare claim databases. We selected the 10 highest-cost drugs for both commercially insured patients and Medicare patients, measured by per-member per-month (PMPM) claim costs to health plans according to the Milliman Health Cost Guidelines™, along with a few additional drugs that have been the subject of litigation over off-label marketing. For each drug, we identified patients with at least one prescription in 2007, and then examined all medical claims for ICD-9 codes for the FDA-approved indications.

Fig. 2 presents all of the drugs we queried. The left column presents the 10 highest-cost drugs in a commercially insured population on a PMPM basis in descending order of cost. The middle column presents the parallel information for a Medicare population. The right column lists the additional drugs we added to our query based on their having been the subject of recent lawsuits mentioned above. Note that we did not include Bextra in our analysis because it was pulled from the market prior to 2007.

We would not expect to see an indicated condition appear in medical claims for every patient. Especially with chronic conditions, it is possible for a patient to have a condition without it being noted on a medical insurance claim form during a particular year. There also may be some imprecision in mapping some approved indications to the ICD-9 diagnosis codes actually recorded by healthcare providers. However, one possible explanation for finding a low frequency of the FDA-indicated condition among patients prescribed a given drug is that the drug is used off-label.

Our analysis showed that there is much variation in the apparent prevalence of approved diagnoses among patients receiving these drugs. Fig. 3 presents selected drugs from the analysis. Also presented is the following information: approved indication(s), the percentage of users having a diagnosis of an approved indication, and other conditions common among users of the drug.

The drugs tended to fall into one of five categories:

1. Drugs almost always used on-label

For some drugs, patients prescribed it very commonly also have a medical claim with a diagnosis for one of the approved indications. For example, more than 92% of Enbrel users and 89% of Actos users have a medical claim with a diagnosis that is approved for the drug. There is little evidence that these drugs are widely used off-label.

2. Drugs for which users frequently have less specific, symptomatic diagnoses

Nexium and Prevacid, two commonly used proton pump inhibitors, are FDA-approved for certain specified disorders such as gastroesophageal reflux disease (GERD) and peptic ulcers. Only a small percentage of users of these drugs had medical claims with those specific diagnoses, but users had very high rates of associated symptoms such as abdominal pain and chest pain. Some of these patients may have had the approved indications but did not have them coded specifically on their insurance claims; others may have been prescribed the drugs off-label to relieve symptoms that are similar to those of the approved conditions. Using claim data to infer medical information is not a precise science; these cases demonstrate where making inferences of off-label use might be more difficult.

3. Possible undercoding of approved indications

Users of three common antidepressants had an apparently low prevalence rate of depression. This is consistent with the under-coding of depression diagnoses observed in claim data in previous Milliman research.[6] While we suspect there is off-label use of these drugs, the low rate of diagnosed depression among antidepressant users may be misleading in terms of the rate of off-label antidepressant use.

4. High prevalence of related conditions

For many drugs, the prevalence rate of the approved indication is low, but there is a high prevalence rate of similar conditions. Many of these prescriptions likely represent off-label use. For example, Actiq is approved specifically for breakthrough cancer pain. Yet only one in five Actiq users in the data had any diagnosis of cancer during the time period examined. However, diagnoses related to pain more generally were common. Another example is Singulair, which is approved to treat asthma and allergies. While slightly more than half of Singulair users did have one of those diagnoses, the prevalence of other types of upper respiratory conditions was also high and may account for many Singulair prescriptions in patients without asthma or allergies.

5. Other potential off-label use

Provigil users have low prevalence for the approved indications and also did not have an obviously closely related condition or symptom explaining a large portion of prescriptions. This suggests that these drugs may be frequently used off-label.


Analyses of the form above provide useful information to pharmaceutical companies, but also suggest that apparent off-label use should be examined in some detail to ensure that the implications are thought through carefully. A finding that a drug is frequently used off-label still does not permit a manufacturer to market the drug for that condition; however, drug-makers can benefit from this information in several ways:

This type of analysis can suggest whether it might be worthwhile to seek FDA approval for a currently non-approved indication. Although it might be tempting to forgo further testing if the drug is already being used off-label, recent events might change this dynamic.

If additional clinical trials are contemplated, a preliminary analysis based on retrospective claim data may provide clues as to the likelihood of success.

Drugs may be found to be frequently prescribed in connection with an unexpected medical condition; this may suggest a disease area for future research and development efforts.

With looming reforms in the health insurance industry, there is likely to be greater focus in coming years on evidence-based medicine, with health plans potentially basing copay structures or reimbursement decisions on research into the relative effectiveness of alternative therapies.

Even if physicians are free to prescribe FDA-approved drugs as they see fit, having FDA approval for the indications for which a drug is most widely used in practice may be an important first step in demonstrating effectiveness. Drugmakers must therefore know how their products are actually being prescribed.

These results highlight the importance of continued post-approval research on the safety and efficacy of pharmaceutical products. Clinical trials are clearly of great importance, but the extent of off-label use of the high-utilization drugs described above is a good illustration of one of the shortcomings of clinical trials; because they are conducted in a highly controlled environment, they may miss important facts about how drugs are used in the real world. As shown, the medical profiles of patients taking the drugs may differ significantly from what was anticipated in clinical trials. It is important for manufacturers to study real-world adherence patterns, concomitant medications, and comorbidities of actual patients once a drug is on the market. PC


1. Radley, D., Finkelstein, S., & Stafford, R. (2006). Off-label prescribing among office-based physicians. Archives of Internal Medicine 166:1021-26.

2. Chen, H., Reeves, J., Fincham, J., Kennedy, W., Dorfman, J., & Martin, B. (2006). Off-label use of antidepressant, anticonvulsant, and antipsychotic medications among Georgia Medicaid enrollees in 2001.The Journal of Clinical Psychiatry 67(6):972-82.

3. Larson, M., Miller, K., & Fleming, K. (2007). Treatment with antidepressant medications in private health plans. Administration and Policy in Mental Health and Mental Health Services Research 34(2):116-26.

4. Chen, D., Wynia, M., Moloney, R., & Alexander, G. (2009). U.S. physician knowledge of the FDA-approved indications and evidence base for commonly prescribed drugs: Results of a national survey. Pharmacoepidemiology and Drug Safety.18(11):1094-1100.

5. Walton, S., Schumock, G., Lee, K., Alexander, G., Meltzer, D., & Stafford, R. (2008). Prioritizing future research on off-label prescribing: Results of a quantitative evaluation. Pharmacotherapy 28(12):1443-52.

6. Melek, S, & Norris, D.. Chronic conditions and comorbid psychological disorders. Retrieved Oct. 26, 2009, from