FormalPara Key Points

Information on drugs and diseases interacting with cigarette smoking is present in some US Food and Drug Administration (FDA)-approved drug labelling.

Dosage adjustments on the basis of cigarette smoking are recommended by the FDA for ten drugs.

1 Introduction

Cigarette smoking poses significant health risks [1]. Its interactions with drugs and diseases have been extensively studied, and numerous literature reviews have been published [2–7]. However, the relevant information in US Food and Drug Administration (FDA)-approved drug labelling has not been analysed.

In an FDA document for new drug development [8], it is suggested that “a clinical study can be conducted in smokers as compared to non-smokers (in lieu of an interaction study with an inducer), when appropriate”. In the FDA draft guidance for industry on drug interaction studies, published in 2006 [9] and 2012 [10], it is recommended that “for a drug that is a substrate of CYP1A2, the evaluation of the effect of induction of CYP1A2 can be carried out by comparative pharmacokinetics (PK) studies in smokers vs. non-smokers”. Furthermore, these draft guidance documents recommend that important drug interaction information, including information on smoking–drug interactions, should be included in different sections of a drug label [9, 10].

We therefore carried out a comprehensive analysis of all FDA-approved drug labelling, aiming to assess if and how much information related to cigarette smoking is present in US drug labelling.

2 Search Strategy

The FDA Online Label Repository (at http://labels.fda.gov), the content of which is also deposited in DailyMed (at http://dailymed.nlm.nih.gov/dailymed/about-dailymed.cfm) as an alternative interface, was chosen as the source of US drug labelling. These websites provide the most comprehensive coverage (>85 %) of FDA-approved product labelling in the public domain [11].

The FDALabel database, which allows a user to search the full text of all labels deposited in DailyMed (at http://www.fda.gov/ScienceResearch/BioinformaticsTools/ucm289739.htm), was used to collect relevant information. Though the FDALabel database is publically available (at https://rm2.scinet.fda.gov/druglabel/#simsearch-0), we used the FDA intranet version (at http://ncsvmweb01.nctr.fda.gov/druglabel/#simsearch-0) because the latter is updated more frequently. Nevertheless, our preliminary results show that the final results were the same regardless of which version was used.

The keywords ‘smoke’, ‘smoker(s)’, ‘smoking’, ‘tobacco’ and ‘cigarette(s)’ were used to search the FDALabel database in the full text of all labels, using the ‘simple search’ function. Any words in the label that match the submitted query are recorded and the corresponding label is extracted. This strategy provides the most comprehensive coverage of relevant information. The resulting records were downloaded and combined, and then duplicates were removed. Labels for over-the-counter (OTC) medications, unapproved drugs and animal drugs were excluded.

The remaining labels were then downloaded directly from DailyMed and manually examined to collect all information related to cigarette smoking. Though this study was initiated to analyse smoking–drug interactions, it was noted that a lot of information regarding smoking–disease interactions was also present in the labelling. We therefore decided to scrutinize all of the information related to cigarette smoking. The permanent link to each examined label in DailyMed and the original statement relevant to cigarette smoking were recorded in an Excel table.

3 Number and Percentage of Labels Providing Smoking-Related Information

As of 19 June 2014, a total of 62,857 labels were deposited in the FDALabel database. A complete list of all of these labels is provided online in Supplementary Table 1 in the Electronic Supplementary Material. Using the keywords ‘tobacco’, ‘cigarette(s)’ and ‘smoke(r)(s)/smoking’ to query the database, 745, 1,245 and 6,695 labels were obtained, respectively. Among these 8,685 labels, there were 1,726 duplicates. The remaining 6,959 unique labels represent 11 % of all 62,677 labels, and they are also presented in Supplementary Table 1.

Among the 6,959 unique labels, 3,207 are for OTC drugs, 215 are for unapproved products, 28 are for animal drugs, 14 are for dietary supplements/vaccines/medical devices and the remaining 3,495 are for human prescription drugs.

Non-human prescription drug labels were excluded for either of two reasons: (1) they were not reviewed or approved by the FDA; or (2) the information was not relevant, as most of them are for safe handling or storage of the products.

The 3,495 human prescription drug labels are for 371 unique drugs, as a specific drug may have multiple labels from different manufacturers. Further excluded were those providing information that was not relevant, such as those dealing with safe handling or storage of flammable drugs (such as oxybutynin chloride). A list of 188 single-active-ingredient drugs and 36 multiple-active-ingredient drugs was finally chosen for detailed analysis, and the original statement regarding cigarette smoking and the link to each label in DailyMed are presented in Supplementary Table 1.

4 Drugs with a Single Active Ingredient

The summarized results for the 188 single-active-ingredient drugs are presented in Tables 1 and 2.

Table 1 Effects of cigarette smoking on drug pharmacokinetics, efficacy and dosage modifications
Full size table
Table 2 Effects of cigarette smoking on diseases or adverse drug reactions
Full size table

As shown in Table 1, smoking may affect the pharmacokinetics and/or efficacy of 34 drugs but shows no effect for 21 other drugs. For 18 drugs, the quantitative effects of smoking on drug pharmacokinetics are specified. Depending on the nature of the specific drugs, smoking could increase or decrease exposure levels, and the perturbations in pharmacokinetic parameters range from 13 to 500 %. Dosage modifications in smokers are necessary for four drugs, including erlotinib hydrochloride, riociguat, theophylline and insulin. For six drugs, including cimetidine and olanzapine, dosage modifications in smokers are necessary only when multiple factors that may affect safety or efficacy are present. For seven drugs, dosage modifications in smokers are not necessary, although the pharmacokinetic parameters of four of them are affected by smoking. This information is presented mainly in the ‘Clinical Pharmacology’ and ‘Drug Interactions’ sections of the label.

Table 2 shows that smoking is a risk factor for 16 types of diseases or adverse drug reactions. One notable example concerns nonsteroidal anti-inflammatory drugs (NSAIDs). All 22 NSAIDs may cause gastrointestinal bleeding, and smoking is a risk factor for this type of adverse reaction. The labelling clearly states that gastrointestinal bleeding is more likely to occur in smokers receiving this class of drugs. Another example concerns antihypertensive drugs. As smoking is a well-established risk factor for cardiovascular disease, it is suggested that usage of any of the 22 antihypertensives “should be part of comprehensive cardiovascular risk management, including … smoking cessation”.

Several drugs in Table 2 warrant discussion in detail. For the drug norethindrone, a black box warning (BBW), the strongest warning about adverse drug reactions issued by the FDA [12], is included in the labelling to highlight the significantly increased risks of heart attacks and strokes in female smokers using oral contraceptives, and “women who use oral contraceptives are strongly advised not to smoke”. This is the only BBW for a single-active-ingredient drug regarding smoking–drug interactions in all labelling. In contrast, for two drugs, doxycycline hyclate and calcitonin salmon, smoking appeared to have no effects on adverse reactions.

5 Drugs with Multiple Active Ingredients

As for the 36 multiple-active-ingredient drugs, none of their labels include information regarding smoking affecting pharmacokinetics. For 27 drugs, the smoking-relevant information is the same as that for the individual active ingredients (see Supplementary Table 1). Of note, the combination drug norethindrone and mestranol has the same BBW for smoking–drug interactions as does the single-active-ingredient drug norethindrone. The remaining nine multiple-active-ingredient drugs are for contraceptive use, and all of them have a BBW stating that “smoking increases the risk of serious cardiovascular side effects from oral contraceptive use” and that “this risk increases with age and with the extent of smoking (in epidemiologic studies, 15 or more cigarettes per day was associated with a significantly increased risk) and is quite marked in women over 35 years of age”. The BBW concludes that “women who use oral contraceptives should be strongly advised not to smoke”. These nine drugs are a combination of ethinylestradiol with either desogestrel, drospirenone, levonorgestrel, ethynodioldiacetate, etonogestrel, levonorgestrel, norelgestromin, norethindrone or norgestimate (see Supplementary Table 1).

6 Discussion and Perspectives

This is the first study to explore information regarding cigarette smoking in FDA-approved drug labelling. By using the FDALabel database, we were able to address this issue in a comprehensive and systematic manner. Specifically, the enormous number of readily available drug labels (62,857 labels as of 19 June 2014; and the number is growing on a daily basis in DailyMed) and the rich and diverse content in the labelling (averaging 20 pages per label; each label includes about ten sections for different types of information) poses a significant challenge for manual review of all of the documents. The FDALabel database provides the unique capability to search the full text of all labels deposited in the FDA Online Label Repository and allows the user to download the search results without limitations. To our knowledge, the FDALabel database is the only publically available tool that provides these functionalities. This database is particularly suitable for the present study, as the three keywords ‘cigarette(s)’, ‘smoke(r)(s)/smoking’ and ‘tobacco’ effectively cover all of the relevant information. In other words, as there are no alternative ways to describe the relevant information, these three keywords are sufficient to extract all of the target information. This is in contrast to more complicated applications, such as a previous study of drug-induced liver injury, for which a significant amount of unwanted information was retrieved because of the lack of specific keywords, and extensive manual annotation was therefore required [13]. For the current study, the use of three keywords enabled us to quickly extract the relevant information with no false negatives and only a minimal number of false positives. For example, we manually double-checked the labelling of 100 drugs that were not picked up by keyword searching, and we found no false negatives. We believe our data present the most comprehensive and accurate picture of cigarette smoking interacting with drugs and/or diseases in US prescription drug labelling.

Only a small number of drug labels were found to carry information related to cigarette smoking. This is not necessarily unexpected, for four possible reasons. First, inclusion is advised by the FDA but is not mandatory. The FDA draft guidance clearly states that it is “for comment purposes only” [9, 10], and no specific inclusion criteria are provided that define what types of interactions involving cigarette smoking are to be included in drug labelling. It is therefore possible that relevant studies were performed in clinical trials during the drug development process, but the results were somehow not analysed or included in the labelling. Second, the first FDA draft guidance that recommended inclusion of smokers as subjects in clinical trials and drug labelling was issued in 2006 [9]. It is therefore likely that drug labelling approved before 2006 does not contain sufficient information regarding cigarette smoking. Third, the FDA draft guidance recommended inclusion of smokers as subjects only in cytochrome P450 (CYP) 1A2-related studies [9, 10]. The effects of smoking on other drug-metabolizing enzymes, such as CYP1B1, CYP2A6, CYP2E1 and uridine 5′-diphosphate-glucuronosyltransferases (UGT) enzymes [14, 15], are currently not included in FDA labelling documents. Lastly, drug labelling usually does not include animal-based data regarding smoking–drug interactions. All of these are important reasons why, in comparison with literature reports [2, 3], only a small number of drugs were identified in this study.

A significant difference between the literature reports and the FDA drug labelling is that the latter does not tend to be exhaustive. Specifically, “only information that would be useful to” clinical decisions should be included in drug labelling [9, 10]. While a recent analysis concluded that FDA drug labelling contains excessive information that may “reduce physician comprehension of important safety warning” [11], another report argued that the FDA is not updating drug labelling consistently when new research findings become available [16]. Currently, the FDA is using its Adverse Event Reporting System (FAERS) to monitor potential signals of serious adverse drug reactions, and the results may lead to a change in drug labelling [17] (see http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformation/Surveillance/AdverseDrugEffects/ucm082196.htm). Analysis of the FAERS database may help identify potential smoking–drug reactions for regulatory considerations. This may be particularly important for drugs that were approved before the first FDA draft guidance recommending inclusion of smokers as subjects in drug development trials was issued in 2006. As the FAERS database contains mainly postmarketing information, the data would also be useful to validate some findings already presented in the labelling, which are largely from clinical trials. In addition, the newly established FDA Center for Tobacco Products (CTP) aims to reduce the adverse health impact for those who use tobacco products, and is collecting safety reports involving tobacco products (at https://www.safetyreporting.hhs.gov/fpsr/WorkflowLoginIO.aspx?metinstance=0E1F229876AB85975D41F4FF60E512B19FF9C2BC). It can be expected that data accumulated by the FDA CTP will help improve drug labelling regarding cigarette smoking in the future. However, the limitations of the FAERS database, such as under-reporting [18], should be taken into consideration for these future endeavours.

The inclusion of smoking–drug interactions in drug labelling appears not to be well recognized by many researchers. Several elegant review articles [2–4, 14] and research papers [19, 20] on smoking–drug interactions have been published, but none of them mentioned the possible inclusion of relevant information in drug labelling. There is an urgent need to increase the awareness of this issue in the research community.

7 Conclusion

Though the current study may represent an underestimate of all smoking–drug interaction investigations in the drug development process, it presents an overview of this issue with regard to drug labelling. This information may be useful to both regulatory agencies and the pharmaceutical industry for future action, such as revising or finalizing the FDA draft guidance [10]. The drug lists presented here may also help basic researchers to design future investigations, and may serve as a good reference source for physicians to better assess the risks of cigarette smoking in patient management.

It should be pointed out that the content of FDA-approved drug labelling, despite being “the official description of a drug product” (see http://www.fda.gov/Drugs/InformationOnDrugs/ucm079436.htm) and “the primary source for drug safety information for physicians” [11], is not perfect or static [13, 16]. Some authors even argue that the content in the BBW section needs to be improved [21]. Part of the statement in the drug labelling, particularly that related to pharmacokinetics, is based on limited sets of clinical trials and may not reflect the most recent findings in the literature [16]. The information presented here may serve as a useful reference but not definitive guidance. It is likely that future studies will reveal additional important smoking–drug interactions that may warrant a change in the labelling.