Establishment of the Canadian Patient Safety Institute followed by release of the Canadian Adverse Events Study drew well-deserved attention to the issue of patient safety in Canada.¹ With respect to drug safety, two major concerns are adverse reactions and adverse interactions.² Risk of either of these is correlated with the number of diseases patients suffer from and the number of medications they are taking concurrently.³ Consequently, drug interactions have become an increasing problem for seniors.⁴

Pharmacists now use software programs that electronically screen for drug interactions. The effectiveness of these programs is limited because the software can check only for interactions among drugs (mostly prescription drugs) patients obtain from a particular pharmacy. Pharmacists are alerted to all potential drug interactions, regardless of significance, which results in a tendency to ignore or override the constant alerts.⁵ Electronic screening for drug interactions is one of the components being incorporated into electronic medical records. The clinical relevance of alerts generated by these systems has been identified as an issue,⁶ as has the propensity for physicians to ignore or override them.⁷ In fact, a recent Canadian study evaluating the effectiveness of a computerized decision-making support system limited the system's interaction alerts to those that were agreed upon as clinically important.⁸

In 2002, we conducted a survey of community-living seniors with osteoarthritis (OA) regarding how they managed their treatment of OA.⁹ As part of that survey, we asked participants to list all medications and supplements they were taking for any reason. The purpose of this article is to document the potential for drug interactions among the seniors in our study and to consider the usefulness of computerized support for detecting clinically important interactions. This is the first known study on this topic in Canadian primary care.

Clinical staff initially sought subjects for the survey from electronic billing records where patients had on record at least one of the following International Classification of Diseases (ICD) diagnostic codes between 1999 and 2001: OA (ICD-9 715, 721), joint pain or stiffness (ICD-9 719.5, 719.5), or other types of arthritis and diffuse connective tissue disorders (ICD-9 710-713, 719.2, 719.3, 725). Diagnosis of OA was confirmed by each subject's family physician. These family physicians were asked to exclude patients they believed were too mentally, physically, or emotionally disabled to participate. In total, 244 seniors (65 years or older) were asked to complete the self-administered mailed survey. We used a modified Dillman method to obtain responses.

The "Living with Arthritis" survey used in this study was developed specifically to gather information on medications patients were using to manage OA and was created by the investigative team and others versed in pharmacy and survey design. Face and content validity were assessed by experienced physicians and pharmacists. The survey was pilot-tested in a volunteer sample of seniors. For this article, we focus on the question that asked respondents to list all medications they had taken in the last 7 days. The question explicitly asked respondents to report over-the-counter medicines, including supplements (such as vitamins, natural or herbal products, or painkillers), which we refer to as "self-care products," along with medicines prescribed by their doctors. They were asked to refer to labels on pill containers to ensure correct information on the product name, strength, and daily dose was given. Full details of the survey method have been published by Lawson et al,⁹ and a copy of the survey is available from the authors on request.

For each respondent, all prescribed medications and self-care products listed were entered into the interaction search software, eFacts from http://www.pubmedcentral.nih.gov/redirect3.cgi?&&reftype=extlink&artid=1479712&iid=131742&jid=303&&www.drugfacts.com.¹⁰ We elected to use this on-line subscription service in lieu of a CD-ROM version to ensure that we accessed an up-to-date resource. Identified interactions were tabulated by level of clinical significance on a scale of 1 to 5.¹⁰ We considered levels 1 (potentially severe or life threatening) and 2 (might cause deterioration in a patient's clinical status) to be of clinical significance and important to physicians and pharmacists at the point of care. The other levels were either "minor," based on "very limited" data, or lacking "good evidence."

Seventy-one of the self-care products mentioned by seniors were not found in the database. We elected not to investigate Aspercreme, "protein," "minerals," or vitamin B of unspecified type. The remaining 67 products were sought in one of seven on-line or print sources and compared manually.^11-18 Based on the information provided by the source, a pharmacist (D.F.) judged it best to categorize any interaction she found by the levels of significance described by http://www.pubmedcentral.nih.gov/redirect3.cgi?&&reftype=extlink&artid=1479712&iid=131742&jid=303&&www.drugfacts.com.¹⁰ Interactions were counted and summarized by significance level and clinically significant combinations.

The study was approved by the Research Ethics Board of the Capital District Health Authority in Halifax, NS.

Of 244 patients with a physician-confirmed diagnosis of OA who were mailed the survey, 191 (78%) responded. Average age was 76.5 years (range 65 to 97 years), 66.7% were women, and 94% reported having some form of drug insurance. Details on products taken were provided by 174 respondents (91.6%) who took an average of 4.7 (standard deviation (SD) 3.2, range 0 to 15) drugs, of which 2.8 were prescription drugs (SD 2.5, range 0 to 10) and 1.9 were self-care products (SD 1.9, range 0 to 10).

Table 1 shows the number of potential interactions by significance level for prescription-prescription, self-care --prescription, and self-care --self-care combinations. Thirty interactions (14%) were level 1 or 2, indicating potential clinical significance. Most interactions (187, 87.4%) were discovered by the eFacts program, including all 30 of the clinically significant ones.

The specific drug (or therapeutic group) and self-care product combinations involved in the 30 clinically significant interactions (level 1 and 2) are listed in Table 2. Eighteen interactions (60%) involved self-care products, most frequently ASA and angiotensin-converting enzyme inhibitor (ACEI) combinations. Because the hypotensive and vasodilator effects of ACEIs can be reduced by ASA, eFacts¹⁰ recommends monitoring blood pressure and hemodynamic parameters. The action recommended in almost all interactions (29/30) was cautionary, such as closer monitoring of blood tests, observation for toxic effects, or advising patients to be aware of side effects. For example, taking warfarin and vitamin E together required "close observation for signs of excessive hypoprothrombinemic response" although that was unlikely to occur at the 400-IU daily dose our respondent was taking. One patient reported taking acetaminophen in combination with warfarin, which might increase the latter's antithrombotic effect. Only one recommendation suggested avoidance, because the combination of diltiazem and atorvastatin leads to risk of increased statin toxicity.

Seven herbal and natural health products had been taken by our respondents in the previous week: glucosamine by 21, garlic by six, methylsulfonylmethane by four, Ginkgo biloba by two, and St John's wort, chondroitin, and echinacea by one each. The eFacts interaction software included Ginkgo biloba and St John's wort and reported only one level-4 interaction (based on limited data) between ASA and Ginkgo biloba with a recommendation "to avoid concurrent use of aspirin and Ginkgo biloba because of the potential for serious bleeding complications."¹⁰ Four interactions found through a manual search of the Pharmacist's Letter¹³ (for glucosamine, echinacea, and garlic) were all clinically insignificant.

There were a large number of potential interactions, 214 in 190 respondents, but only a small proportion of these (14%) were clinically significant. Equally important, almost all recommendations for action were cautionary rather than suggesting that the combinations be avoided completely. The largest number of potential interactions was between ACEIs and anti-inflammatory medications (11 with ASA and one with indomethacin). Physicians who care for patients taking ACEIs need to be aware that they might need to increase doses if patients begin taking commonly used anti-inflammatory medications. The authors suggest that many patients taking ASA are doing so at a physicians' request for ischemic heart disease. More troublesome is the potential for an enhanced antithrombotic effect of warfarin if a patient starts taking either acetaminophen, a very commonly used nonprescription medication, or large doses of vitamin E.

Physicians and pharmacists both need to be alert to the fact that most clinically significant interactions involved self-care products. Health professionals could be unaware of the risk of harm if they do not regularly ask patients about other products they are taking. Patients might not realize that they should mention vitamins (considered safe at recommended daily doses) when they tell their physicians or pharmacists what medications they are taking.

Our respondents were taking more herbal and natural health products than had been reported in previous articles based on Canadian data,¹⁹ but comparable amounts to those reported in an American study.²⁰ The very few, all clinically insignificant, potential interactions found might be owing to the paucity of reliable data in the literature. Brazier and Levine have documented thoroughly how little solid evidence there is for clinically significant drug-herb interactions.²¹

The on-line search program yielded a high proportion (87.4%) of all interactions ultimately found, including all the clinically significant interactions. Six clinically insignificant interactions were identified for every significant one (184:30). Our data suggest that the risk outlined by Magnus et al⁷ is real: "... the perception that interaction alerts were frequently irrelevant so that GPs had become used to overriding them." Interaction program developers and clinician users should collaborate to ensure that computerized support for prescribing decisions clearly flags the clinically important "must see" potential interactions so they do not get ignored during patient visits.

One serious limitation to any system for identifying potential interactions is double-doctoring and patients' use of many different pharmacies to fill their prescriptions. We did not ask patients whether their medications were prescribed by more than one physician or whether they were receiving medications from more than one pharmacy. Tamblyn et al showed that the most important risk factor for potentially inappropriate drug combinations was patients having more than one prescribing physician and that using a single dispensing pharmacy could lower this risk.²²

The survey was funded by an unconditional grant from the Nova Scotia Department of Health through the Drug Evaluation Alliance of Nova Scotia. Drug data entry and interaction analysis was funded through an unconditional grant from Merck Frosst Canada Inc.