Article Text
Abstract
Background The official prescribing information document distributed with a prescription drug is a key source of safety information, but it may include excessive or insufficient details.
Objectives To compare prescribing information approved by the US Food and Drug Administration with the UK, Canada and Australia to identify content differences in safety warnings.
Methods For 20 top-selling prescription drugs, we used an automated natural language processing tool to calculate the number and severity of reported adverse drug reactions (ADRs). We fit hierarchical Poisson models and included fixed effects for other prescribing information characteristics. Separately, we analysed the appearance and content of ‘black box’ warnings.
Results There was substantial variation in safety content of approved prescribing information. Canada had the highest median ADRs per drug (138 (IQR 86–234)) and the UK had the lowest (84 (IQR 51–111)). The number of ADRs reported was on average 50% higher in Canada compared with the USA (ratio of ADRs/document: 1.5, 95% CI 1.4 to 1.6, p<0.001). By contrast, there were on average 15% fewer ADRs listed in the UK compared with the USA (ratio of ADRs/document 0.85 (95% CI 0.78 to 0.93, p<0.001), and 21% fewer ADRs listed in Australia compared with the USS (ratio of ADRs/document 0.79, 95% CI 0.74 to 0.85, p<0.001). There were no variations in ADR severity. The presence and qualitative content of boxed warnings also showed substantial diversity.
Conclusions International variations exist in the presentation of safety data in drug prescribing information, which may have important implications for patient safety. Better international coordination is necessary to enhance use of this information for patient decision-making.
- Medication safety
- Health policy
- Communication
- Adverse events, epidemiology and detection
- Information technology
Statistics from Altmetric.com
- Medication safety
- Health policy
- Communication
- Adverse events, epidemiology and detection
- Information technology
Introduction
When a country's pharmaceutical regulatory authority approves a new drug, it also approves its accompanying official prescribing information (also known as the ‘package insert’, ‘product information’ or ‘summary of product characteristics’).1 This document describes the key studies leading to approval, the drug's approved indications and known side effects, and technical details about administration and storage. The prescribing information is written by the manufacturer and approved by the regulatory authority, and may be updated while the drug is on the market as new efficacy and safety data emerge.
The drug prescribing information is a key source of drug safety facts for prescribers or pharmacists,2 ,3 who are expected to discuss these details with patients to educate them about expected benefits and risks of therapy.4–8 However, several observers have noted over many years that the prescribing information as currently written and disseminated is unlikely to serve this goal effectively.9–13 For example, an analysis of package inserts approved by the US Food and Drug Administration in 2011 found that many listed hundreds of adverse effects, with the total number appearing to grow over time.14 The authors concluded that these drug prescribing information documents may be ‘overwarning’, listing far too many adverse effects and leading prescribers to ignore them. This practice may serve the needs of manufacturers, as providing such warnings can be seen as exculpatory in case a patient experiences one of the named adverse effects, protecting the company against charges that it failed to warn of risks.15
By contrast, others have argued that certain drug information documents contain too few warnings. In recent lawsuits in the USA, patients harmed by prescription drug side effects have charged manufacturers with negligently failing to update their prescribing information to account for emerging safety information,16 or excluding basic safety information altogether.17 Indeed, research has shown that FDA-approved drug prescribing information is frequently out of date and inadequately represents adverse drug effects.18
To provide an international comparative perspective on the issue of reporting safety details in drug prescribing information, we compared officially sanctioned prescribing information documents for the same drugs in four English-speaking settings: the USA, Canada, the UK and Australia. We sought to identify differences in the content and safety warnings and to ascertain whether aspects of the drug's approved uses or other characteristics were associated with the depiction of safety warnings.
Methods
Data sources
Our primary data sources were the prescribing information documents for prescription drugs in the USA, Canada, the UK and Australia. In each country, the administrative process for creating the document is similar, as the document is written by the manufacturer and then formally approved by the governmental drug regulatory authority. These documents are publicly available in the following repositories: DailyMed19 (USA), Health Canada20 (Canada), Electronic Medicines Compendium21 (UK), and the Therapeutic Goods Administration22 (Australia).
In October 2011, from these repositories, we obtained prescribing information documents from the four comparator settings for the 20 top-selling prescription drugs (by aggregate sales) in the USA in 2009.23 In each case, the product information relating to the brand name originator product was selected. Since these 20 drugs were available in all four settings, our final dataset consisted of prescribing information documents for 80 total products. We did not consider accompanying ‘patient information leaflets’, which are shortened versions of prescribing information documents available in some settings, such as the UK.
Characteristics of study drugs and their prescribing information documents
We first collected characteristics of the drugs and prescribing information that might serve as predictors of differences in depicting adverse drug reactions (ADRs). We identified the word counts of ADR section, the full safety section and the entire document. We also counted the number of unique indications for which each drug was approved in each of the four settings, and the duration each drug had been on the market there. We retrieved the approval information for each medication from the aforementioned repositories, except for the US drugs, for which we used the Drugs@FDA approvals database.24
Finally, we identified the number of patients in clinical trials reported as contributing to the safety data. When available, this value is reported at the beginning of the safety section or adverse reaction subsection.
We separately evaluated the use of ‘black box’ warnings, safety warnings mandated by a regulatory authority to highlight major risks at the beginning of the prescribing information. This visual mechanism of highlighting safety alerts does not appear in the UK. For US, Canadian and Australian prescribing documents, we recorded the presence of a black box warning and the specific adverse reactions cited.
ADR counts and severity
Our central analysis calculated the number and severity of ADRs reported in each prescribing information document. We used an automated tool for extracting adverse reaction descriptors that employs natural language processing. This tool, known as the Structured Product Label Information Coder and Extractor (SPLICER), was developed by one of the authors (JDD) and has been previously described.25 Briefly, SPLICER identifies the adverse reactions within each document using the Medical Dictionary of Regulatory Activities (MedDRA) terminology. The MedDRA hierarchy automates the aggregation of synonymous concepts (eg, wooziness and lightheadedness), while keeping separate different subtypes of a condition (eg, maculopapular rash, vesicular rash). In an evaluation study, SPLICER showed an accuracy of 94% (sensitivity 93% and positive predictive value 95%) in extracting coded adverse reactions.2,5 As input, we used the following comparable sections: Adverse Reactions (USA and Canada), Undesirable Effects (UK), Adverse Effects (Australia). We used SPLICER to calculate the total number of unique ADRs, disregarding redundant mentions of the same ADR.
We then determined the proportion of severe ADRs. To make this determination, we used the First DataBank National Drug Data File Side-Effect Module.26 As a curated dataset, the National Drug Data File does not include every possible reaction, but for those that it does list, a severity level is provided (severe vs non-severe). ‘Severe’ is defined in the First DataBank module documentation as ‘may be life threatening’, but reactions causing substantial morbidity may also be marked as severe. The same reaction can also have differing severity levels across different drugs, (eg, the term ‘diarrhoea’ may be considered severe for one medication and non-severe for another). To account for this variation, we considered any reaction coded as severe in at least 50% of instances to be a severe reaction.
Statistical analysis
To compare reported ADRs, we first fit hierarchical Poisson models using the number of distinct reactions on each drug prescribing information document as the outcome. The unadjusted model included a random intercept for each drug and fixed effects for country. To facilitate broad international comparisons, we chose the USA as the reference group. The adjusted model additionally included fixed effects for other characteristics: the proportion of adverse reactions that were severe, the number of approved indications, the word count of the safety section outside the adverse events section, that is, encompassing the Contraindications, Warnings, and Precautions sections (in units of 1000 words), and the total number of reported patients contributing to the safety information (dichotomised at >10 000 vs ≤10 000). The results are reported as rate ratios comparing the rate of reactions listed per prescribing information. We used R software V.2.14.1.27
Results
The 20 study drugs covered a range of indications, including psychiatric disease (6), cardiovascular disease (5) and oncology (3) (see table 1). They included 15 small molecule and 5 biological agents. By the end of 2011, the median duration of time on the market for these drugs was 12.8 years in Australia (IQR 9.3–14.5 years), 11.8 years in the USA (IQR 8.6–14.3), 11.7 years in Canada (IQR 8.5–13.6) and 10.8 years in the UK (IQR 8.8–12.7). The drugs were approved for a median of four indications in the USA (IQR 2–5), three in Australia (IQR 2–4.25), and 2.5 in Canada and the UK (IQR 2–5 for both).
The overall word length of the prescribing information in the four countries differed considerably. Canadian documents were the longest, with a median of 16 779 words (IQR 14 159–22 281). US documents had a median of 11 666 words (IQR 9709–17 990), and Australian documents had a median of 8401 words (IQR 6801–10 670). UK documents for the same drugs were the shortest, with a median of 6077 words (IQR 5268–7737). With a standard typeset page containing 400 words, Canadian prescribing information documents therefore had a median length of 42 pages compared with just 15 in the UK. This length includes safety data and pharmacokinetics, dosage guidelines and other prescribing information.
Cross-national differences in safety information
The safety information was based on widely varying numbers of patients. The UK prescribing information reported a median safety sample size of 7819 patients (IQR 4000–16 066) whereas the US prescribing information reported a median of 6278 (IQR 3005–13 907), Canadian prescribing information reported a median of 3967 (IQR 3257–11 022), and Australian prescribing information reported 3563 (IQR 2124–11 924) (data missing from 17 (21%) prescribing documents).
We also found substantial variation in the numbers of ADRs reported for the same drugs across the four countries. As shown in table 1, the overall median ADRs per drug was 105 (IQR 64–152), with Canadian prescribing information having the highest median number (138; IQR 86–234) and prescribing information from the UK having the lowest (84; IQR 51–111). Individual drugs differed substantially, with seven drugs showing variability of greater than 100 ADRs across the four countries. Canadian prescribing information contained the highest number of ADRs for 14 of the 20 drugs.
As an example of important variation, consider the antidepressant venlafaxine. The Canadian Adverse Reactions section for venlafaxine spans 25 pages (7093 words), including six tables of clinical trial results, extensive premarketing and postmarketing events, and data for multiple indications and doses. By contrast, the UK version spans just two pages (502 words) and contains a single table listing all reactions. Focusing on ADRs specifically, the venlafaxine prescribing information in Canada lists 507 ADRs compared with just 112 ADRs in the UK. Figure 1 shows gastrointestinal ADRs from UK and Canadian documents. The UK prescribing information is succinct, containing only seven gastrointestinal effects. The numerous additional ADRs found in the Canadian version vary in frequency and severity. The UK prescribing information does not include any of the postmarketing events found on the Canadian prescribing information with the exception of pancreatitis. Among those absent include serious events such as gastrointestinal bleeding and fulminant hepatitis.
Overall, differences in adverse reaction severity among the four regions were small. Across all settings, 60.7% of ADRs were classified as severe by the First DataBank definition. Prescribing information in the UK had a median proportion of severe reactions of 62% (IQR 51–72%), while US prescribing information had a median of 60% (IQR 54–76%). Australian prescribing information had a 59% median proportion of severe reactions (IQR 50–69%), while Canadian prescribing information had 58% (IQR 49–66%).
Hierarchical models
There were substantial differences in the counts of ADR reports across countries. As shown in figure 2, the number of ADRs reported per prescribing information was on average 60% higher in Canada compared with the USA (ratio of ADRs per prescribing document 1.6, 95% CI 1.5 to 1.7). By contrast, there were on average 36% fewer ADRs listed in the UK prescribing information compared with the USA (ratio of ADRs per document 0.64, 95% CI 0.61 to 0.68), and 14% fewer ADRs listed in Australian prescribing information compared with the USA (ratio of ADRs per document 0.86, 95% CI 0.81 to 0.91).
These differences were reduced slightly after multivariate analysis that accounted for the number of approved indications, years since approval, safety sample size, non-ADR word count and warning severity. After adjustment, estimated rate ratios were 1.5 (95% CI 1.4 to 1.6, p<0.001) for Canada, 0.85 (95% CI 0.78 to 0.93, p<0.001) for the UK, and 0.79 (95% CI 0.74 to 0.85, p<0.001) for Australia. The difference between the UK and the USA becomes most attenuated when adjusting for the word count outside of the adverse events section and when adjusting for the safety population sample size.
Variation in black box warnings
The presence and qualitative content of boxed warnings found in US, Canadian and Australian prescribing information also differed. As shown in table 2, 15 of the 20 (75%) medications had a boxed warning in the USA by the end of 2011. By comparison, only 8 (53%) had boxed warnings in Canada, while none (0%) had boxed warnings in Australia.
However, we found that absence of a boxed warning did not mean that the adverse event was absent. For example, bevacizumab had boxed warnings for the well known side effect of pulmonary haemorrhage in the USA and Canada. In Australia, we found no boxed warning, but the bevacizumab prescribing information listed haemorrhage in the Precaution section, stating, ‘Patients treated with AVASTIN have an increased risk of haemorrhage, especially tumour-associated haemorrhage.’
There was also substantial diversity regarding the details of these highlighted warnings. For example, pioglitazone had a boxed warning for heart failure in the USA, but not in either the Canadian or Australian prescribing information. In Canada, the Contraindication section listed ‘New York Heart Association (NYHA) Class I to IV cardiac status’, while in the Australian Contraindication section, the prescribing information warned that the drug was ‘not recommended in patients with symptomatic heart failure. Initiation of ACTOS (like other thiazolidinediones) is contraindicated in patients with NYHA Class II, III or IV heart failure.’ In addition to the boxed warning, the US prescribing information lists as a contraindication ‘Do not initiate ACTOS in patients with established NYHA Class III or IV heart failure.’
Discussion
This cross-national comparative study of drug prescribing information revealed substantial variation in the listing of safety information in top-selling drugs in the USA, Canada, the UK and Australia. Given the demographic similarities of the four countries we examined, there are few medical reasons why any of the drugs in our study should produce different ADR profiles in these settings. Nonetheless, while Canadian drug prescribing information listed significantly more ADRs than the USA, we found that prescribing information in Australia and the UK specified significantly fewer. These results have important safety implications for patients and physicians who rely on information in prescription drug prescribing information to help guide their medical decision-making. The prescribing information document is supposed to represent current knowledge about the safety of drugs, but our study reinforces that its content may be driven by social or political forces, not the best available clinical evidence.
Our results are consistent with other studies showing international differences in consumer-focused medical information.28 Yet while others have reported variation in drug prescribing information around the world for the same chemical entity,29 our study is the first to quantify these variations and demonstrate that these differences were not accounted for by clinically relevant characteristics, such as length of time on the market or number of approved indications. An alternative explanation may be that manufacturers design different prescribing documents in each setting in response to other facets of the pharmaceutical market, such as the risk of litigation for failure to warn about side effects. In the USA, drug prescribing information serves an important legal function for patients who bring failure-to-warn lawsuits against manufacturers, because it delineates the complete knowledge about a drug's risks that the manufacturer reports to the FDA at the time of approval. While the diminished risk of failure-to-warn lawsuits against drug manufacturers in Australia and the UK might explain why the documents are shorter in those settings than in the USA, it is not consistent with the finding of longer prescribing information in Canada.
Another explanation for these results is that they reflect differences in the regulatory process in these four settings. For example, Canadian regulators may be more inclusive in negotiating with the drug sponsor, while UK and Australian regulators are less inclusive. The length of Canadian prescribing information may also be partly explained by the fact that until 2006, these documents were considered the property of the manufacturer and not disseminated publicly; rather, only a substantially truncated version was made widely available to physicians and patients.30 The different role that the prescribing information document has historically played in the Canadian system compared with Australia, the USA or UK, may have affected regulators’ decision-making about including or excluding certain ADRs. Notably, while the adverse event section was less inclusive in the USA compared with Canada, we found that regulators in the USA were more apt to include black box warnings in prescribing information than regulators in Canada.
Whatever their cause, wide cross-national differences in pharmaceutical product descriptions have important implications for prescribing. For example, we found that while prescribing information may vary greatly in the number of adverse reactions listed, the average severity of ADRs did not change significantly across the settings in our study. As a result, Canadian prescribing information, which consistently listed the most ADRs, warned physicians and patients about more severe ADRs related to these drugs than prescribing information from the UK and Australia, which had the fewest ADRs. Such differences may affect the decision-making by clinicians and patients in each of these settings who choose to use these drug prescribing information documents to guide their prescription choices. While physicians and patients have gaps in their knowledge about drug safety,31 warning content in drug prescribing information affects use of the product32 ,33 and may contribute to differences in the patterns of use of drugs, such as the prevalence of non-evidence-based unapproved (‘off-label’) use.34 Though international variations in the quality of healthcare delivery and patient safety may be related to drug prescribing information, more research is needed to examine the clinical implications of the differences that we observed.
One type of drug product information that is well known to have an impact on prescribing practices is a black box warning. Previous studies suggest that physicians pay attention to the content of such warnings,35 and change their practices when they are added by a regulatory authority.36 Given the potential for black box warnings to drive physician prescribing practices in ways that avoid unsafe use of drugs, variations in this characteristic are particularly concerning, Australian authorities hold a high bar for such warnings, placing them on none of the 20 drugs in our sample. Such a choice might contribute to rates of inappropriate prescribing of drugs in Australia, such as atypical antipsychotics for older patients with dementia. The US FDA, by contrast, used black box warnings the most, and in particular appeared to favour imposing class-based warnings such as the ‘suicidality’ warning that was included with all three serotonin reuptake inhibitors on the list.
Our study is limited in that it considers only the number of unique ADR warnings, and does not assess their context, such as whether they are highlighted in some way, or listed in a table compared with in text. In addition, our analysis compared the largest safety-related section, but did not include other sections (eg, ‘Contraindications’). The way warnings are described in these other sections would enhance the rate of false positive results because of the context in which they are written (eg, the phrase ‘liver function tests’ in the ADR section is very likely to refer to abnormal liver function tests, whereas the same phrase in a different section of the prescribing information may refer to a suggestion to ‘check liver function tests every 12 months’). However, while regulatory authorities in different regions may conceivably compensate for fewer warnings in the adverse events section with more warnings in other sections, our results were robust to statistical adjustment for word length that accounted for this possibility. Though our study was limited to four settings, legislation related to medicines in Europe occurs at the European Union (EU) level, so findings from the UK are likely to generalise to the entire EU. Finally, our analysis was limited to 20 top-selling brand-name drugs, so the results may not generalise to specialty drugs with lower sales or to generic drugs.
This analysis provides quantitative evidence that regulation in different countries leads to substantial variation in drug warning information. While there is no evidence that one regulatory approach leads to better outcomes, there have been a number of efforts aimed at promoting international drug safety monitoring by coordinating data collection and review by different authorities.37 Our findings suggest that similar coordination may be necessary to determine best practices in the presentation of drug safety data in prescribing information documents. Such cooperation should take into account universal ethical principles related to the proper communication of risk information38 to reach a reasonable balance between overwarning and underwarning and thereby promote consistent and evidence-based drug prescribing.
Acknowledgments
The authors would like to thank John Connolly for his help with the background research and William Shrank for his insights in the early stages of this project.
References
Footnotes
-
Contributors ASK and JDD conceived of the study, collected data and wrote the first draft. JMF designed and conducted statistical analysis, contributed to writing the draft and reviewed the final draft. JA oversaw all writing and study design and reviewed and approved the final draft.
-
Funding ASK is supported by a career development award from the Agency for Healthcare Research & Quality (K08HS18465–01), and a Robert Wood Johnson Foundation Investigator Award in Health Policy Research. JDD is supported by a Young Investigator's award from the Indiana Clinical Translational Sciences Institute (KL2RR025760).
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.