Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Comparison of rates of nausea side effects for prescription medications from an online patient community versus medication labels: an exploratory analysis

Comparison of rates of nausea side effects for prescription medications from an online patient... Background: While medication labels are considered the authoritative resource for medication information, emerging research suggests that patient-generated health data (PGHD) are a valuable tool to understand the ways in which patients experience medications in real world settings. However, the relationship between these two data sources has not been closely examined. Methods: To understand how rates of medication side effects compare between a source of PGHD and medication labels, the current study compares adverse drug reaction rates from FDA medication labels with those self-reported by patients from an online patient community, PatientsLikeMe (PLM). The linear association between medication label and PLM nausea rates was evaluated using Spearman correlation, with an associated 95% confidence interval calculated based on 10,000 bootstrap iterations. The reporting ratio of PLM nausea rates to medication label nausea rates was defined for all treatments with non-zero medication label nausea rates. Lognormality of the distribution of this reporting ratio was assessed based on a Kolmogorov-Smirnov test (α = 0.05). Results: Nausea rates for 163 medications were compared between the two data sources. Overall rates ranged from 0 to 60% for medication labels and 0 to 36% for PLM data with median rates of 6.4 and 3.7%, respectively. In general, nausea rates reported by patients in the online community were lower than those found in medication labels. This inconsistency was attributed to a variety of factors, including differences in data collection mechanisms and product use factors. Conclusions: Quantifiable and consistent differences exist between side effect rates reported on medication labels and those self-reported by patients based on real-world use. In general, self-reported rates of nausea associated with medication use were lower than those reported in medication labels. Although considered a definitive resource for medication information, this discrepancy demonstrates that medication labels may not comprehensively describe the patient experience. Results suggest that a combination of information from different sources may provide a more rounded and holistic view on medication safety and tolerability. Keywords: Adverse events, Drug safety, Pharmacovigilance, Adverse drug reaction reporting systems, Postmarketing product surveillance, Product labeling, Data analysis * Correspondence: DBlaser@patientslikeme.com PatientsLikeMe, 160 2nd Street, Cambridge, MA 02142, USA Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Blaser et al. AAPS Open (2017) 3:10 Page 2 of 10 Background (FDA) and the European Medicines Agency (Research The term “side effect” typically refers to an unintended Center for Devices and Radiological Health. Guidance response to a medication that is related to the pharma- for Industry Patient-Reported Outcome Measures, 2009; cological properties of that particular medicine (World 21st Century Cures Act, HR 34, 114th Congress, 2015; Health Organization, 2002). Prior research indicates that European Medicines Agency, 2011). significant differences exist between reported rates of To explore potential differences in side effect report- medication side effects during real-world use and those ing, nausea, a relatively common side effect reported observed in clinical trials (Blenkinsopp et al., 2007; across several medication classes, was selected for fur- Hakobyan et al., 2011). Clinical trials follow rigorous ther investigation. Nausea can be very debilitating and protocols for extensive safety monitoring both during has the potential to seriously impact patient health out- and after a trial; however, side effects, as observed in comes. As nausea is a subjective experience, it is particu- real-world settings, are not monitored as rigorously. larly at risk of being underestimated or underreported Outside the controlled environment of clinical trials, by providers. Additionally, patients may be able to more several variables may affect the pharmacodynamics of a accurately attribute symptomatic reactions, such as nau- particular drug, including dosage, duration of use, indi- sea, to a specific drug as opposed to reactions where cation, comorbidities, and use of concomitant medica- causality is less obvious, such as acute (e.g. heart attack, tions. In both clinical trials and during real-world use, pulmonary embolism) or systemic (e.g. fatigue, general- most side effects are reported through clinicians; this ized muscle pain) events (Research Center for Devices methodology may lead to underreporting, as clinicians and Radiological Health. Guidance for Industry Patient- may underestimate the severity of patient-reported Reported Outcome Measures, 2009). symptoms or patients may withhold side effect informa- This exploratory analysis seeks to quantify potential tion due to stigma (Pakhomov et al., 2008; Basch et al., differences in how nausea is self-reported by patients 2009; Weingart et al., 2005). (PLM database) and by investigators in the context of Advancements in technology have led to the gener- clinical trials (medication labels). The side effect profiles ation and availability of new types of medically-relevant of medications are critical components of regulator, data sources that have an array of potential health appli- payer, and physician decisions, as well as patients’ deci- cations, including patient-generated health data (PGHD) sions about whether or not to take a particular drug. (Wood et al., 2015). PGHD is defined as “health-related Quantifying the difference between side effect reporting data—including health history, symptoms, biometric in clinical trials versus PGHD offers an opportunity to data, treatment history, lifestyle choices, and other infor- better characterize the safety profile of medications and mation—created, recorded, gathered, or inferred by or to improve the patient experience. from patients or their designees (i.e. care partners or those who assist them) to help address a health concern” Methods (Shapiro et al., 2012). It is distinct from data generated The aim of this exploratory analysis was to examine the in clinical settings because it allows for patients to dir- relationship between side effect rates reported on medi- ectly report and record their own information. PGHD cation labels and those self-reported by members of the can complement and enrich other sources of medical PLM online patient community. evidence, and its use is advised when measuring con- cepts best known by the patient or best measured PLM data from the patient perspective, such as pain or nausea The initial set of medications eligible for inclusion was se- (Research Center for Devices and Radiological Health. lected based on available PLM data volume, and included Guidance for Industry Patient-Reported Outcome 220 prescription medications with side effects reports Measures, 2009). from 50 or more patients on PLM. As FDA medication la- PatientsLikeMe (PLM), an online patient community bels would be the source of comparison, only FDA- where patients track, share, and discuss health informa- approved medications were included in the analysis. For tion, is one source of PGHD. Patients use the PLM each eligible medication, the PLM nausea rate was defined platform to track their own health data, including medi- as the proportion of patients completing a treatment cation side effects. This study seeks to explore the evaluation or side effect report who had reported nausea relationship between side effect rates reported on medi- as a side effect. Data were aggregated by generic medica- cation labels (i.e., package inserts) and those reported by tion. All PLM data are as of January 1, 2016. patients in the PLM community. The value of gaining a better understanding of the ways in which side effects Medication label data are experienced during real-world medication use is rec- For all eligible treatments identified based on available ognized by both the Food and Drug Administration PLM data, corresponding medication labels were retrieved Blaser et al. AAPS Open (2017) 3:10 Page 3 of 10 using the text mining system i2E (Linguamatics, Cam- generic ingredient, 220 prescription treatments met the bridge, UK). Rule-based queries were used to identify the inclusion criteria of 50 or more patients reporting side word “nausea” within the Adverse Reactions section of effect information (Fig. 1). Of these 220 medications, a medication labels. Queries were designed in I2E OnDe- total of 12 were excluded from subsequent analyses due mand, which provides access to medication labels sourced to non-specific active ingredients (n = 5), lack of FDA from the DailyMed website sourced from the FDA. Data approval (n = 4), ongoing research investigations which extraction was limited to the most recent version of the influence side effect reporting (n = 2), and not being a medication label as of May 2016. Unique Ingredient Iden- drug product (n = 1, oxygen therapy). All data generated tifier (UNII) codes were utilized to ensure that synonyms or analyzed during this study are included in this pub- and/or lexical variants were identified for each medication. lished article [Additional file 1 and Additional file 2]. Nausea rates were recorded along with, when available, additional data from associated clinical trials (e.g. study Individual medication labels sample size, nausea rate of placebo group). Information related the medication’s nausea rates was Nausea rates reported in combination with vomiting identified and extracted from each medication label’s (e.g. methotrexate reported a combined nausea/vomiting “Adverse Reactions” section for the 208 eligible medica- rate of 10% (Methotrexate sodium [package insert], tions. Of the 208 eligible medications, 72% (n = 150) of 2003), but provided no rate associated uniquely with the medications reported at least one numerical frequency symptom “nausea”), and instances where nausea was in- for nausea. For medications in which the label did not cluded as part of a list of side effects without a numer- list nausea (n = 13), the nausea rate was considered to ical approximation were categorized as having “no be 0% for subsequent comparisons. Nausea was listed in medication label nausea frequency information”. This the label 45 of these medications, but did not have a nu- occurred in a total of 49 nausea rates associated with 45 merical frequency, and could not be included in the sub- otherwise eligible generic treatments, which were ex- sequent analysis. A total of 397 nausea rates were cluded from subsequent analyses. If nausea was not extracted for the remaining 150 medications that re- mentioned on the medication label, the nausea rated ported a numerical frequency. The number of nausea was treated as 0% in subsequent calculations. When rates identified exceeded the number of medications as nausea rates were reported as an approximation or multiple rates may be reported for a single medication range, the nearest reported value (e.g. “>5%” as 5%) or label and/or medications may have multiple brand prod- midpoint of range (e.g. “2%–6%” as 4%) was used. Nau- ucts with product-specific labels. Of the 397 extracted sea rates from pediatric studies were excluded. The me- nausea rates, approximately 5% were reported as a nu- dian of all reported nausea rates was calculated when merical approximation (n = 11) or range (n = 9). Exam- multiple nausea rates were available for a medication. ples of numerical approximations included: “At least 1%”, “fewer than 1 in 100” and “≥10%”. Data processing and analysis Data were analyzed using R version 3.2.3 (R Core Team, Aggregate information for eligible medications 2016). The linear association between medication label Medication label data were aggregated by active ingredi- and PLM nausea rates was evaluated using Spearman cor- ent, and the median nausea rate was calculated for each relation, with an associated 95% confidence interval calcu- of the 163 medications. Aggregated rates from medica- lated based on 10,000 bootstrap iterations. The reporting tion labels ranged from 0% to 60%, with a median of 6% ratio of PLM nausea rates to medication label nausea rates and an interquartile range of 3% to 14% (Fig. 2a). For was defined for all treatments with non-zero medication the same 163 medications, nausea rates reported by the label nausea rates. In cases where the PLM nausea rate PLM community ranged from 0% to 36%, with a median was 0, half the value of the lowest-detected PLM nausea of 4% and an interquartile range of 1% to 6% (Fig. 2b). rate (0.001) was added to enable calculation of log ratio. Lognormality of the distribution of the reporting ratio was assessed based on a Kolmogorov-Smirnov test (α =0.05).. Comparison of PLM vs. medication label nausea rates Confidence intervals for PLM nausea rates were calculated The Spearman correlation between PLM and medication using the Clopper-Pearson exact method, and adjusted for label nausea rates was 0.59 [95% confidence interval multiple comparisons using Bonferroni adjustment. 0.48, 0.68], demonstrating a moderate positive correl- ation between the two data sources. In both aggregate Results PLM and medication label data, cyclophosphamide had PLM data characteristics the highest observed nausea rate (Fig. 3). Medications As of January 1, 2016, the PLM community included with the largest absolute difference in reported nausea 393,371 patients. After aggregating available data by rates are shown in Table 1. Blaser et al. AAPS Open (2017) 3:10 Page 4 of 10 Fig. 1 Aggregation of available data by generic ingredient. A total of 163 medication met all of the inclusion criteria Fig. 2 Distribution of median reported nausea rates for 163 medications (aggregated by active ingredient). IQR = interquartile range Blaser et al. AAPS Open (2017) 3:10 Page 5 of 10 Cyclophosphamide 30% Doxycycline Milnacipran Mitoxantrone 20% Amoxicillin−clavulanate Mycophenolate mofetil 10% Cyclosporine Venlafaxine Naltrexone Ritonavir Tacrolimus Fluvoxamine 0% 0% 20% 40% 60% 80% Label Nausea Rate [%] Fig. 3 PatientsLikeMe nausea rates versus medication label nausea rates. Selected medications are highlighted with text labels For 57% (93/163) of drugs, the label rate was within 0.38, indicating that, for half of all observed treatments, the Bonferroni-adjusted 95% confidence interval of the the PLM nausea rate was at or below 38% of the medica- proportion of patients from PLM who reported nausea tion label nausea rate. The reporting ratio distribution (Fig. 4). For the 43% (70/163) of drugs that did not fall was well approximated by a lognormal distribution within these confidence intervals, 86% (60/70) of label (Fig. 5). nausea rates fell above the upper bound of the PLM To explore factors that may contribute to different confidence interval. In total, 60 drugs were identified as reporting rates, doxycycline, ritonavir, amoxicillin/clavu- having label nausea rates significantly higher than PLM lanate, mitoxatrone, and three organ transplant medica- nausea rates, and 10 drugs were identified as having tions as a group (cyclosporine, mycophenolate mofetil, label nausea rates significantly lower than PLM nausea and tacrolimus) were selected for further investigation rates. using both PLM data and detailed information from To compare the degree of over or underreporting of medication labels. nausea rates among patients from PLM relative to the rates reported in medication labels, the reporting ratio Ritonavir was calculated as the ratio of PLM nausea rates to medi- In the current study, ritonavir had the highest observed cation label nausea rates. Across 150 treatments with magnitude of absolute difference between medication non-zero medication label nausea rates, reporting ratios label (57.4%) and PLM (2.7%) nausea rates. This discrep- ranged from 0 to 13. The median reporting ratio was ancy cannot be attributed to a difference in indication Table 1 Medications with largest absolute differences in nausea rates between PLM data and medication label data Medication name PLM Medication Label Overall Difference Nausea % N Nausea % (median) N Ritonavir 2.7% 75 57.4% 1755 -55% Mycophenolate mofetil 9.7% 257 54% 896 -44% Fluvoxamine 0% 57 39% 1295 -39% Tacrolimus 1.9% 215 38.5% 931 -37% Mitoxantrone 20.3% 79 55% 148 -35% Trihexyphenidyl 5.6% 54 40% Not reported -34% Fentanyl transdermal patch 8.6% 220 41% 216 -32% Venlafaxine 6.5% 1035 37% 4857 -31% Naltrexone 3.9% 407 29% 440 -25% Desvenlafaxine 7.1% 197 31% 1365 -24% Values for all medications are available [see Additional file 1]. PLM Nausea Rate [%] Blaser et al. AAPS Open (2017) 3:10 Page 6 of 10 60% Within Bonferroni Adjusted Confidence Interval label rate outside of adjusted CI label rate within adjusted CI 40% 20% 0% Medication Fig. 4 Bonferroni-adjusted confidence internals for PLM nausea rates versus medication label nausea rates. For each generic treatment, PLM nausea rates and their associated 95% Bonferroni-adjusted confidence intervals are positioned along the horizontal axis, and medication label nausea rates are shown as a black line. In cases where the medication label nausea rate falls outside of the associated 95% confidence interval based on PLM data, treatments are colored red, identifying instances in which drug label nausea rates and PLM nausea rates were identified as having notable differences, after accounting for multiple comparisons. Since not all drug labels reported the number of patients included in the associated study, it was not possible to assess the statistical significance of the differences between PLM and label nausea rates for each drug for medication use, as the drug is listed as a treatment doxycycline use among patients from PLM included for HIV infection in both data sources. Instead, the ob- Lyme disease (37%) and acne (14%), with reported daily served difference is likely due to widely varying dosage dosages ranging from 20 to 400 mg/day (median daily information between the medication labels and PLM. In dose 100 mg/day). In comparison, information reported PLM, all but one patient reported a daily dosage of in the medication label was based on a clinical trial of 100 mg/day (a single patient reported a daily dosage of patients with periodontal disease who received a daily 80 mg/day). In contrast, the medication label corre- dose of 40 mg/day (Periostat® [package insert], 2001). sponds to patient experiences while taking a daily dose of 1200 mg/day. Amoxicillin/Clavulanate Like that of doxycycline, the nausea rate reported by pa- Doxycycline tients from PLM for amoxicillin/clavulanate (11.5%) The doxycycline nausea rate in PLM (27.1%) exceeded exceeded that reported in the medication label (2.6%). the medication label nausea rate (8%) by the largest Amoxicillin/clavulanate is indicated for a wide range of magnitude, which may be explained by differences in conditions; indications included sinusitis, urinary tract both dosing and indication for use. Indications for infections, and other systemic infections in both PLM Fig. 5 Empirical and theoretical lognormal distributions of reporting ratio. Empirical distribution of the reporting ratio of PLM nausea rates to medication label nausea rates and a theoretical lognormal distribution with (natural log scale) mean − 1.01 and standard deviation 1.31. Distributions based on data from 150 treatments with non-zero medication label nausea rates Nausea Rate Blaser et al. AAPS Open (2017) 3:10 Page 7 of 10 data and medication labels. Corresponding daily doses reported in medication labels reflect the overall rate for recommended by clinical guidelines also vary across in- nausea that patient experienced, but not the rate that is dications, ranging from 750 to 4000 mg/day (Amoxicillin likely to be attributed to the medication alone. Trau- and clavulanate, 2016). Unlike doxycycline and ritonavir, matic, invasive transplant surgeries and/or pain medica- there are no clear patient or product use characteristics tions are likely a significant contributor to nausea that appeared responsible for differences in nausea rates following a transplant. Patients evaluating treatments on between these data sources. Influenced by the relatively PLM may differentiate between nausea associated with low number of patients who have evaluated amoxicillin/ the overall experience of receiving an organ transplant clavulanate on the PLM platform (n = 61), the aggregate and nausea caused by transplant-related medications, medication label nausea rate for amoxicillin/clavulanate leading to lower reported rates for nausea as compared falls within the Bonferroni adjusted 95% confidence to those reported on medication labels. interval for the proportion of patients experiencing nau- sea in the PLM community. Discussion The goal of this study was to closely examine rates of Mitoxantrone patient-reported medication side effects from a real- The aggregate nausea rate reported in the medication world data source and from medication labels. This label for mitoxantrone (55%) was higher than that re- study showed that PLM nausea rates for 163 treatments ported in PLM (20.3%). A more detailed examination of ranged from 0% to 36%, with a median of 4%. Across a the medication label data suggests that the expected dif- broad range of indications and dosages, PLM nausea ference may be even more striking than the observed rates were positively correlated to those reported on difference. The median medication label nausea rate of medication labels. Moreover, this study demonstrated a 55% is based on data collected over the course of three quantifiable relationship between the reporting ratio of clinical trials, one with a nausea rate of 76% (n = 62, PLM nausea rates to medical label nausea rates, which dose = 12 mg/m ), the second with a rate of 55% provides a useful basis for comparisons against alterna- (n = 65, dose = 5 mg/m ), and a third trial where mitox- tive data sources (e.g. patient report to a physician). antrone was administered concurrently with methyl- Medication labels are often considered the definitive prednisolone and had a nausea rate of 29% (n = 21) data source for information on side effects due the (Novantrone® [package insert], 2012). The current rec- strenuous requirements of regulatory approval. However, ommended daily dose of mitoxantrone is 12 mg/m , any data source will have strengths and limitations. which is more than double the dose administered in the Many medications were approved decades ago, and up- cohort of patients whose reported nausea rate was 55% dates to their labels are infrequent. The labels also do (Marriott et al., 2010). Furthermore, administration of not contain information related to off-label indications. mitoxantrone in combination with an anti-emetic medi- An estimated 20% of outpatient prescriptions in the cation, such as methylprednisolone, would likely result United States are written for off-label use (Radley et al., in lower overall reported rates of nausea. It seems that 2006). For example, the label for amitriptyline contains the median medication label nausea rate (55%) may ac- information from clinical trials conducted over 50 years tually be an underestimate of the nausea rate observed ago in patients with depression. However, prescribing during a clinical trial that more closely mirrors real- practices have evolved significantly for amitriptyline so world prescribing practices (76%), leading to an even that the proportion of off-label use is very high (>80%) greater discrepancy between the nausea rates reported in with the most frequent use in fibromyalgia, neuralgias PLM and those reported in the medication label. and migraine prevention (Radley et al., 2006; Moore et al., 2015; Silberstein, 2015). Organ transplant medications Differences in reporting rates of nausea between medi- Our review also found that nausea rates reported on cation labels and in the PLM database may be influenced PLM were much lower than those reported in medica- by various factors, including methodological differences tion labels for the medications tacrolimus (PLM – 1.9%; in data collection and differences in indications for medication label – 38.5%), mycophenolate mofetil (PLM medication use. Data for all drugs from within the PLM rate – 9.7%; medication label – 54%), and cyclosporine database are collected through one consistent mechan- (PLM rate – 6.9%; medication label – 14%) which are ism, whereas data collected through clinical trials for use typically taken after organ transplantation to prevent re- in medication labels may be obtained using varying jection. Product use characteristics for these medications methods. Prior research has shown that the methods were examined in greater detail to determine what may used to collect side effect data significantly influence the be contributing to these wide differences. One potential results obtained (Bent et al., 2006). For example, sexual explanation for this observed trend is that nausea rates side effects due to use of selective serotonin reuptake Blaser et al. AAPS Open (2017) 3:10 Page 8 of 10 inhibitors ranged from 2% to 73% depending primarily Limitations on whether side effects were collected via a checklist, The side effect nausea can be reported in a variety of open-ended questioning, or by a detailed inquiry (Safer, ways (i.e., qualitatively or quantitatively) on medication 2002). labels. In order to collect data that were consistent with In the case of this study, we speculate that the general information reported in medication labels, PLM nausea trend of lower nausea rates in the PLM database com- cases were identified only as instances in which patients pared to those reported in corresponding medication la- specifically reported nausea. Patient reports of side ef- bels is driven largely by differences in data collection fects known to have a high co-occurrence with nausea, mechanisms. On the PLM platform, patients report side such as vomiting, were not considered as nausea cases. effects using a standard method across all drugs. Once a In cases where nausea was not mentioned on a drug patient indicates that they have experienced any particu- product label, the label nausea rate was approximated as lar side effect, they are presented with an entry field with zero for analysis purposes. FDA guidelines state that the an auto-completer that provides suggestions as the user Adverse Reactions section of a medication label does not types. include all adverse events that occurred during a clinical Although the FDA requires that studies investigating trial, but only those that meet an appropriate cutoff (e.g. new drugs report adverse events, they do not specify a all adverse reactions occurring at a rate of 10% or standard method for data collection. One common ap- greater in the treatment group and at least twice the rate proach accepted by the FDA is to ask patients to of placebo) (US Food and Drug Administration, 2006). complete a checklist of symptoms, which is likely to cap- Due to variable reporting methods, aggregating nausea ture additional symptoms unrelated to the particular rates from medication labels proved particularly challen- medication of interest (U.S. Department of Health and ging. In many instances, more than one value was given Human Services, Food and Drug Administration, 2005; for nausea as a side effect in a single medication label, Smith et al., 2013). Conversely, the process of collecting and sample sizes were not always available for each indi- data in the PLM database is more open-ended than that vidual rate, leading to challenges in calculating an overall used in clinical trials, leading patients to be more likely nausea rate. to report only their most bothersome side effects. Each In addition, aggregating nausea rates across multiple approach has both advantages and disadvantages; how- labels for the same active ingredient was challenging due ever, the use of different mechanisms for data collection to inconsistencies or redundancies in the information. across clinical trials limits the ability of researchers to For example, the medication label for levetiracetam accurately compare side effect profiles across drugs. In (Keppra®) does not list nausea as a side effect, but the contrast, the consistent method of collecting side effect label for extended-release levetiracetam (Keppra XR®) data from the PLM platform is advantageous to making lists a nausea rate of 5% (Keppra® [package insert], 2013; more valid comparisons among drugs’ safety profiles. Keppra XR® [package insert], 2016). Upon further inves- The examples highlighted in the Results section illus- tigation, it was found that nausea was previously listed trate that differences in dosage and indication contribute in the levetiracetam (Keppra®) label, described as an to differences in reported nausea rates. There are likely “event reported by at least 1% of adult levetiracetam other factors that additionally contribute to these differ- (Keppra)-treated patients but as or more frequent in the ences, including concurrent treatments, comorbidities, placebo groups.” This description, however, has been re- disease severity, and other patient population character- moved in updates to the medication label over time. Fur- istics (e.g. age, gender, location). However, information thermore, it was often difficult to determine if nausea related to these subgroups was seldom reported in medi- rates for different labels of the same active ingredient cation labels. were from the same clinical trial, as was often the case In addition to the methodological challenges associ- for immediate-release and extended-release products. ated with comparing side effects among many medica- Some medication labels included nausea rates for in- tions, differences in medication label data reporting gredients similar to the active ingredient, as was the case presented additional challenges. Data can be presented for the valproate sodium label, which contained informa- quantitatively (e.g. 5%) or qualitatively (e.g. “commonly” tion from trials of divalproex sodium. Similarly, some or “frequently”). Nausea, in particular, is reported in medication labels included nausea rates for a class of various ways, often combined with rates of vomiting and medications rather than for only the active ingredient. reported as one value. The variety of methods used for For example, the label for minocycline contains informa- 1) collecting information about side effects in clinical tri- tion on side effects of tetracyclines, a broad class of anti- als and 2) reporting information in medication labels biotics that includes minocycline, but it did not provide limits the ability to make meaningful comparisons be- information about side effects specific to minocycline tween medications. (Minocycline hydrochloride [package insert], 2012). Blaser et al. AAPS Open (2017) 3:10 Page 9 of 10 The underreporting of side effect rates by patients in Acknowledgements We are grateful to all the members of PatientsLikeMe that have contributed the PLM community may also be influenced by the ex- their data and continue to make this research possible. We also thank Jill M. istence of survival bias. Although patients are able to Serrano, PhD for her assistance in preparing and writing this manuscript. enter their entire medication history on PLM, they are Availability of data and materials prompted to enter more detailed information about ex- All data generated or analysed during this study are included in this article periences with current treatments. Patients, who stop [Additional file 1 and Additional file 2]. taking a particular medication due to intolerable side ef- fects, may not report a detailed history for that medica- Authors’ contributions DAB, SE, JL, and PP wrote the manuscript. DAB, SE, JL, SR, PW, and JW tion compared to medications they are currently taking. designed the research. DAB, SE, JL, SR and PP performed the research and Prompting patients to report on their current treatments analyzed the data. All authors read and approved the final manuscript. may lead to lower reported rates of nausea in the PLM Ethics approval and consent to participate database. Independent ethics review was not sought as members (patients) of the A common critique of Internet communities is that PatientsLikeMe Community are informed of their involvement in research patients may not be able to accurately report their diag- activities via the User Agreement and Privacy Policy before joining the site. The User Agreement states that “Member Notices: If you register as a nosis and medication history, or that anonymity provides member, you agree that PatientsLikeMe may send notices to you by email at an opportunity for individuals to report misinformation. the email address you provide when registering to become a member (or A previous study demonstrated that a large percentage which you later update using the functionality of the Site).” and the Privacy Policy states “PatientsLikeMe may also periodically ask Members to complete of the patients on PLM users can be matched to real pa- short surveys about their experiences (including questions about products tients with confirmed identities, and there is a high de- and services). Member participation in these surveys is not required, and gree of agreement between self-reported and insurance refusal to do so will not impact a Member's experience on the Site.” All data used for analysis of medication labels was from published sources. claims for diagnoses and medication history (Eichler et al., 2016). Other studies have also shown high degree Competing interests of agreement between self-reported comorbidities and DAB, SE, and PW are employees of PatientsLikeMe and hold stock options in medical records (Ye et al., 2017). the company. The PatientsLikeMe Research Team has received research funding (including conference support and consulting fees) from Abbvie, Accorda, Actelion, Alexion, Amgen, AstraZeneca, Avanir, Biogen, Boehringer Conclusions Ingelheim, Celgene, EMD, Genentech, Genzyme, Janssen, Johnson & Johnson, Merck, Neuraltus, Novartis, Otsuka, Permobil, Pfizer, Sanofi, Shire, In the current study, we observed a positive correlation Takeda, Teva, and UCB. The PatientsLikeMe R&D team has received research between nausea rates reported in medication labels and grant funding from Kaiser Permanente, the Robert Wood Johnson those self-reported by patients in an online community. Foundation, Sage Bionetworks, The AKU Society, and the University of Maryland. PW is an associate editor at the Journal of Medical Internet However, in general, nausea rates reported among pa- Research and is on the Editorial Boards of The BMJ and BMC Medicine. JW, tients in the PLM community were lower than those re- JL and SR are employees of AstraZeneca. JW and SR hold stock options in ported in medication labels. While there are likely many the company. factors driving the discrepancy between these two sources, including differences in data collection mecha- Publisher’sNote nisms and product use factors, this study demonstrates a Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. quantifiable relationship in side effect reporting frequen- cies between sources of real-world evidence and the Author details 1 2 controlled environment of clinical trials. Reporting fre- PatientsLikeMe, 160 2nd Street, Cambridge, MA 02142, USA. AstraZeneca Research & Development, Advanced Analytics Centre, Mereside, Alderley quencies of side effects have high impact on regulator, Park, Macclesfield, Cheshire SK10 4TF, UK. Northeastern University, 360 payer, physician, and patient decisions; therefore, collec- 4 Huntington Ave, Boston, MA 02115, USA. AstraZeneca Research & tion of data that accurately reflect the patient experience Development, Granta Park, Riverside 2, Cambridge CB21 6GH, UK. is critical. Using self-reported patient data may be one Received: 28 April 2017 Accepted: 13 October 2017 way to enhance the capture of subjective elements of medication tolerability and safety. References 21st Century Cures Act, HR 34, 114th Congress. (2015–2016). http://docs.house. Additional files gov/billsthisweek/20161128/CPRT-114-HPRT-RU00-SAHR34.pdf. Accessed 19 Oct 2017 Additional file 1: Datafile of nausea rates extracted from medication Amoxicillin and clavulanate: Drug information. In: UpToDate, Post TW (Ed), labels and PatientsLikeMe. (CSV 14 kb) UpToDate, Waltham, MA. Accessed on December 15, 2016 Basch E, Jia X, Heller G, Barz A, Sit L et al (2009) Adverse symptom event Additional file 2: Codebook for datafile of nausea rates extracted from reporting by patients vs clinicians: relationships with clinical outcomes. J Natl medication labels and PatientsLikeMe. (CSV 1 kb) Cancer Inst 101(23):1624–1632 Bent S, Padula A, Avins AL (2006) Brief communication: Better ways to question Abbreviations patients about adverse medical events: a andomized, controlled trial. Ann FDA: Food and drug administration; PGHD: patient-generated health data; Intern Med 144(4):257–261 Erratum in: Ann Intern Med. 2006 Jul 18;145(2): PLM: PatientsLikeMe; UNII: Unique ingredient identifier 156 Blaser et al. AAPS Open (2017) 3:10 Page 10 of 10 Blenkinsopp A, Wilkie P, Wang M, Routledge PA (2007) Patient reporting of US Food and Drug Administration. Adverse Reactions Section of labeling for suspected adverse drug reactions: a review of published literature and human prescription drug and biological products—content and format. international experience. Br J Clin Pharmacol 63(2):148–156 http://www.fda.gov/downloads/Drugs/ Eichler GS, Cochin E, Han J, Hu S, Vaughan TE et al (2016) Exploring concordance GuidanceComplianceRegulatoryInformation/Guidances/ucm075057.pdf. of patient-reported information on PatientsLikeMe and medical claims data Published 2006. Accessed 19 Oct 2017 at the patient level. J Med Internet Res 18(5):e110 Weingart SN, Gandhi TK, Seger AC, Seger DL, Borus J et al (2005) Patient-reported medication symptoms in primary care. Arch Intern Med 165(2):234–240 European Medicines Agency. Fourth report on the progress of the interaction Wood WA, Bennett AV, Basch E (2015) Emerging uses of patient generated with patients’ and consumers’ organisations (2010) and results/analysis of the health data in clinical research. Mol Oncol 9(5):1018–1024 degree of satisfaction of patients and consumers involved in EMA activities World Health Organization. The importance of pharmacovigilance - safety during 2010. 2011. http://www.ema.europa.eu/docs/en_GB/document_ monitoring of medicinal products. 2002. http://apps.who.int/medicinedocs/ library/Report/2011/10/WC500116866.pdf. Accessed 19 Oct 2017 en/d/Js4893e/. Accessed 19 Oct 2017 Hakobyan L, Haaijer-Ruskamp FM, de Zeeuw D, Dobre D, Denig PA (2011) Review Ye F, Moon DH, Carpenter WR, Reeve BB, Usinger DS et al (2017) Comparison of of methods used in assessing non-serious adverse drug events in patient report and medical Records of Comorbidities: results from a observational studies among type 2 diabetes mellitus patients. Health Qual population-based cohort of patients with prostate cancer. JAMA Oncol Life Outcomes 9:83 Keppra XR® [package insert]. UCB, inc.; Smyrna, GA; 2016. http://www.accessdata. fda.gov/drugsatfda_docs/label/2016/022285s022lbl.pdf. Accessed 19 Oct Keppra® [package insert]. UCB, Inc.; Smyrna, GA; 2013. http://www.accessdata.fda. gov/drugsatfda_docs/label/2015/021035s093,021505s033lbl.pdf. Accessed 19 Oct 2017 Marriott JJ, Miyasaki JM, Gronseth G, O'Connor PW (2010) Therapeutics and technology assessment Subcommittee of the American Academy of neurology. Evidence report:the efficacy and safety of mitoxantrone (Novantrone) in the treatment of multiple sclerosis: report of the therapeutics and technology assessment Subcommittee of the American Academy of neurology. Neurology 74(18):1463–1470 Methotrexate sodium [package insert]. Xanodyne Pharmacal, Inc., Florence, KY; 2003. http://www.accessdata.fda.gov/drugsatfda_docs/label/2004/ 11719slr106_methotrexate_lbl.pdf. Accessed 19 Oct 2017 Minocycline hydrochloride [package insert]. Teva Pharmaceuticals USA inc., Sellersville, PA.; 2012. http://www.accessdata.fda.gov/drugsatfda_docs/label/ 2013/063009s034lbl.pdf. Accessed 19 Oct 2017 Moore RA, Derry S, Aldington D, Cole P, Wiffen PJ (2015) Amitriptyline for neuropathic pain in adults. Cochrane Database Syst Rev 7:CD008242 Novantrone® [package insert]. EMD Serono, Inc., Rockland, MA; 2012. http://www. accessdata.fda.gov/drugsatfda_docs/label/2012/019297s035lbl.pdf. Acccessed 19 Oct 2017 Pakhomov SV, Jacobsen SJ, Chute CG, Roger VL (2008) Agreement between patient-reported symptoms and their documentation in the medical record. Am J Manag Care 14(8):530–539 Periostat® [package insert]. CollaGenex pharmaceuticals, Inc., Newton, PA; 2001. http://www.accessdata.fda.gov/drugsatfda_docs/nda/2001/50-783_Periostat_ prntlbl.pdf. Accessed 19 Oct 2017 R Core Team. R: a language and environment for statistical computing. R foundation for statistical computing. 2016. https://www.r-project.org/. Accessed 19 Oct 2017 Radley DC, Finkelstein SN, Stafford RS (2006) Off-label prescribing among office- based physicians. Arch Intern Med 166(9):1021–1026 Research Center for Devices and Radiological Health. Guidance for Industry Patient-Reported Outcome Measures: Use in medical product development to support labeling claims. 2009. http://www.fda.gov/downloads/Drugs/ GuidanceComplianceRegulatoryInformation/Guidances/UCM193282.pdf. Accessed 19 Oct 2017 Safer DJ (2002) Design and reporting modifications in industry-sponsored comparative psychopharmacology trials. J Nerv Ment Dis 190(9):583–592 Shapiro M, Johnston D, Wald J, Mon D. Patient-Generated Health Data: White Paper Prepared for the Office of the National Coordinator for Health Information Technology by RTI International. 2012. http://www.rti.org/sites/ default/files/resources/patientgeneratedhealthdata.pdf. Accessed 19 Dec 2016 Silberstein SD (2015) Preventive migraine treatment. Continuum (Minneap Minn) 21(4 Headache):973–989 Smith SM, Wang AT, Katz NP, McDermott MP, Burke LB et al (2013) Adverse event assessment, analysis, and reporting in recent published analgesic clinical trials: ACTTION systematic review and recommendations. Pain 154(7):997–1008 U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER). Reviewer Guidance: Conducting a Clinical Safety Review of a New Product Application and Preparing a Report on the Review. 2005. http://www.fda.gov/downloads/ drugs/guidancecomplianceregulatoryinformation/guidances/ucm072974.pdf. Accessed 19 Oct 2017 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png AAPS Open Springer Journals

Comparison of rates of nausea side effects for prescription medications from an online patient community versus medication labels: an exploratory analysis

Loading next page...
 
/lp/springer-journals/comparison-of-rates-of-nausea-side-effects-for-prescription-gf90nlZ8gX

References (15)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s)
Subject
Biomedicine; Pharmaceutical Sciences/Technology; Pharmacology/Toxicology
eISSN
2364-9534
DOI
10.1186/s41120-017-0020-y
Publisher site
See Article on Publisher Site

Abstract

Background: While medication labels are considered the authoritative resource for medication information, emerging research suggests that patient-generated health data (PGHD) are a valuable tool to understand the ways in which patients experience medications in real world settings. However, the relationship between these two data sources has not been closely examined. Methods: To understand how rates of medication side effects compare between a source of PGHD and medication labels, the current study compares adverse drug reaction rates from FDA medication labels with those self-reported by patients from an online patient community, PatientsLikeMe (PLM). The linear association between medication label and PLM nausea rates was evaluated using Spearman correlation, with an associated 95% confidence interval calculated based on 10,000 bootstrap iterations. The reporting ratio of PLM nausea rates to medication label nausea rates was defined for all treatments with non-zero medication label nausea rates. Lognormality of the distribution of this reporting ratio was assessed based on a Kolmogorov-Smirnov test (α = 0.05). Results: Nausea rates for 163 medications were compared between the two data sources. Overall rates ranged from 0 to 60% for medication labels and 0 to 36% for PLM data with median rates of 6.4 and 3.7%, respectively. In general, nausea rates reported by patients in the online community were lower than those found in medication labels. This inconsistency was attributed to a variety of factors, including differences in data collection mechanisms and product use factors. Conclusions: Quantifiable and consistent differences exist between side effect rates reported on medication labels and those self-reported by patients based on real-world use. In general, self-reported rates of nausea associated with medication use were lower than those reported in medication labels. Although considered a definitive resource for medication information, this discrepancy demonstrates that medication labels may not comprehensively describe the patient experience. Results suggest that a combination of information from different sources may provide a more rounded and holistic view on medication safety and tolerability. Keywords: Adverse events, Drug safety, Pharmacovigilance, Adverse drug reaction reporting systems, Postmarketing product surveillance, Product labeling, Data analysis * Correspondence: DBlaser@patientslikeme.com PatientsLikeMe, 160 2nd Street, Cambridge, MA 02142, USA Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Blaser et al. AAPS Open (2017) 3:10 Page 2 of 10 Background (FDA) and the European Medicines Agency (Research The term “side effect” typically refers to an unintended Center for Devices and Radiological Health. Guidance response to a medication that is related to the pharma- for Industry Patient-Reported Outcome Measures, 2009; cological properties of that particular medicine (World 21st Century Cures Act, HR 34, 114th Congress, 2015; Health Organization, 2002). Prior research indicates that European Medicines Agency, 2011). significant differences exist between reported rates of To explore potential differences in side effect report- medication side effects during real-world use and those ing, nausea, a relatively common side effect reported observed in clinical trials (Blenkinsopp et al., 2007; across several medication classes, was selected for fur- Hakobyan et al., 2011). Clinical trials follow rigorous ther investigation. Nausea can be very debilitating and protocols for extensive safety monitoring both during has the potential to seriously impact patient health out- and after a trial; however, side effects, as observed in comes. As nausea is a subjective experience, it is particu- real-world settings, are not monitored as rigorously. larly at risk of being underestimated or underreported Outside the controlled environment of clinical trials, by providers. Additionally, patients may be able to more several variables may affect the pharmacodynamics of a accurately attribute symptomatic reactions, such as nau- particular drug, including dosage, duration of use, indi- sea, to a specific drug as opposed to reactions where cation, comorbidities, and use of concomitant medica- causality is less obvious, such as acute (e.g. heart attack, tions. In both clinical trials and during real-world use, pulmonary embolism) or systemic (e.g. fatigue, general- most side effects are reported through clinicians; this ized muscle pain) events (Research Center for Devices methodology may lead to underreporting, as clinicians and Radiological Health. Guidance for Industry Patient- may underestimate the severity of patient-reported Reported Outcome Measures, 2009). symptoms or patients may withhold side effect informa- This exploratory analysis seeks to quantify potential tion due to stigma (Pakhomov et al., 2008; Basch et al., differences in how nausea is self-reported by patients 2009; Weingart et al., 2005). (PLM database) and by investigators in the context of Advancements in technology have led to the gener- clinical trials (medication labels). The side effect profiles ation and availability of new types of medically-relevant of medications are critical components of regulator, data sources that have an array of potential health appli- payer, and physician decisions, as well as patients’ deci- cations, including patient-generated health data (PGHD) sions about whether or not to take a particular drug. (Wood et al., 2015). PGHD is defined as “health-related Quantifying the difference between side effect reporting data—including health history, symptoms, biometric in clinical trials versus PGHD offers an opportunity to data, treatment history, lifestyle choices, and other infor- better characterize the safety profile of medications and mation—created, recorded, gathered, or inferred by or to improve the patient experience. from patients or their designees (i.e. care partners or those who assist them) to help address a health concern” Methods (Shapiro et al., 2012). It is distinct from data generated The aim of this exploratory analysis was to examine the in clinical settings because it allows for patients to dir- relationship between side effect rates reported on medi- ectly report and record their own information. PGHD cation labels and those self-reported by members of the can complement and enrich other sources of medical PLM online patient community. evidence, and its use is advised when measuring con- cepts best known by the patient or best measured PLM data from the patient perspective, such as pain or nausea The initial set of medications eligible for inclusion was se- (Research Center for Devices and Radiological Health. lected based on available PLM data volume, and included Guidance for Industry Patient-Reported Outcome 220 prescription medications with side effects reports Measures, 2009). from 50 or more patients on PLM. As FDA medication la- PatientsLikeMe (PLM), an online patient community bels would be the source of comparison, only FDA- where patients track, share, and discuss health informa- approved medications were included in the analysis. For tion, is one source of PGHD. Patients use the PLM each eligible medication, the PLM nausea rate was defined platform to track their own health data, including medi- as the proportion of patients completing a treatment cation side effects. This study seeks to explore the evaluation or side effect report who had reported nausea relationship between side effect rates reported on medi- as a side effect. Data were aggregated by generic medica- cation labels (i.e., package inserts) and those reported by tion. All PLM data are as of January 1, 2016. patients in the PLM community. The value of gaining a better understanding of the ways in which side effects Medication label data are experienced during real-world medication use is rec- For all eligible treatments identified based on available ognized by both the Food and Drug Administration PLM data, corresponding medication labels were retrieved Blaser et al. AAPS Open (2017) 3:10 Page 3 of 10 using the text mining system i2E (Linguamatics, Cam- generic ingredient, 220 prescription treatments met the bridge, UK). Rule-based queries were used to identify the inclusion criteria of 50 or more patients reporting side word “nausea” within the Adverse Reactions section of effect information (Fig. 1). Of these 220 medications, a medication labels. Queries were designed in I2E OnDe- total of 12 were excluded from subsequent analyses due mand, which provides access to medication labels sourced to non-specific active ingredients (n = 5), lack of FDA from the DailyMed website sourced from the FDA. Data approval (n = 4), ongoing research investigations which extraction was limited to the most recent version of the influence side effect reporting (n = 2), and not being a medication label as of May 2016. Unique Ingredient Iden- drug product (n = 1, oxygen therapy). All data generated tifier (UNII) codes were utilized to ensure that synonyms or analyzed during this study are included in this pub- and/or lexical variants were identified for each medication. lished article [Additional file 1 and Additional file 2]. Nausea rates were recorded along with, when available, additional data from associated clinical trials (e.g. study Individual medication labels sample size, nausea rate of placebo group). Information related the medication’s nausea rates was Nausea rates reported in combination with vomiting identified and extracted from each medication label’s (e.g. methotrexate reported a combined nausea/vomiting “Adverse Reactions” section for the 208 eligible medica- rate of 10% (Methotrexate sodium [package insert], tions. Of the 208 eligible medications, 72% (n = 150) of 2003), but provided no rate associated uniquely with the medications reported at least one numerical frequency symptom “nausea”), and instances where nausea was in- for nausea. For medications in which the label did not cluded as part of a list of side effects without a numer- list nausea (n = 13), the nausea rate was considered to ical approximation were categorized as having “no be 0% for subsequent comparisons. Nausea was listed in medication label nausea frequency information”. This the label 45 of these medications, but did not have a nu- occurred in a total of 49 nausea rates associated with 45 merical frequency, and could not be included in the sub- otherwise eligible generic treatments, which were ex- sequent analysis. A total of 397 nausea rates were cluded from subsequent analyses. If nausea was not extracted for the remaining 150 medications that re- mentioned on the medication label, the nausea rated ported a numerical frequency. The number of nausea was treated as 0% in subsequent calculations. When rates identified exceeded the number of medications as nausea rates were reported as an approximation or multiple rates may be reported for a single medication range, the nearest reported value (e.g. “>5%” as 5%) or label and/or medications may have multiple brand prod- midpoint of range (e.g. “2%–6%” as 4%) was used. Nau- ucts with product-specific labels. Of the 397 extracted sea rates from pediatric studies were excluded. The me- nausea rates, approximately 5% were reported as a nu- dian of all reported nausea rates was calculated when merical approximation (n = 11) or range (n = 9). Exam- multiple nausea rates were available for a medication. ples of numerical approximations included: “At least 1%”, “fewer than 1 in 100” and “≥10%”. Data processing and analysis Data were analyzed using R version 3.2.3 (R Core Team, Aggregate information for eligible medications 2016). The linear association between medication label Medication label data were aggregated by active ingredi- and PLM nausea rates was evaluated using Spearman cor- ent, and the median nausea rate was calculated for each relation, with an associated 95% confidence interval calcu- of the 163 medications. Aggregated rates from medica- lated based on 10,000 bootstrap iterations. The reporting tion labels ranged from 0% to 60%, with a median of 6% ratio of PLM nausea rates to medication label nausea rates and an interquartile range of 3% to 14% (Fig. 2a). For was defined for all treatments with non-zero medication the same 163 medications, nausea rates reported by the label nausea rates. In cases where the PLM nausea rate PLM community ranged from 0% to 36%, with a median was 0, half the value of the lowest-detected PLM nausea of 4% and an interquartile range of 1% to 6% (Fig. 2b). rate (0.001) was added to enable calculation of log ratio. Lognormality of the distribution of the reporting ratio was assessed based on a Kolmogorov-Smirnov test (α =0.05).. Comparison of PLM vs. medication label nausea rates Confidence intervals for PLM nausea rates were calculated The Spearman correlation between PLM and medication using the Clopper-Pearson exact method, and adjusted for label nausea rates was 0.59 [95% confidence interval multiple comparisons using Bonferroni adjustment. 0.48, 0.68], demonstrating a moderate positive correl- ation between the two data sources. In both aggregate Results PLM and medication label data, cyclophosphamide had PLM data characteristics the highest observed nausea rate (Fig. 3). Medications As of January 1, 2016, the PLM community included with the largest absolute difference in reported nausea 393,371 patients. After aggregating available data by rates are shown in Table 1. Blaser et al. AAPS Open (2017) 3:10 Page 4 of 10 Fig. 1 Aggregation of available data by generic ingredient. A total of 163 medication met all of the inclusion criteria Fig. 2 Distribution of median reported nausea rates for 163 medications (aggregated by active ingredient). IQR = interquartile range Blaser et al. AAPS Open (2017) 3:10 Page 5 of 10 Cyclophosphamide 30% Doxycycline Milnacipran Mitoxantrone 20% Amoxicillin−clavulanate Mycophenolate mofetil 10% Cyclosporine Venlafaxine Naltrexone Ritonavir Tacrolimus Fluvoxamine 0% 0% 20% 40% 60% 80% Label Nausea Rate [%] Fig. 3 PatientsLikeMe nausea rates versus medication label nausea rates. Selected medications are highlighted with text labels For 57% (93/163) of drugs, the label rate was within 0.38, indicating that, for half of all observed treatments, the Bonferroni-adjusted 95% confidence interval of the the PLM nausea rate was at or below 38% of the medica- proportion of patients from PLM who reported nausea tion label nausea rate. The reporting ratio distribution (Fig. 4). For the 43% (70/163) of drugs that did not fall was well approximated by a lognormal distribution within these confidence intervals, 86% (60/70) of label (Fig. 5). nausea rates fell above the upper bound of the PLM To explore factors that may contribute to different confidence interval. In total, 60 drugs were identified as reporting rates, doxycycline, ritonavir, amoxicillin/clavu- having label nausea rates significantly higher than PLM lanate, mitoxatrone, and three organ transplant medica- nausea rates, and 10 drugs were identified as having tions as a group (cyclosporine, mycophenolate mofetil, label nausea rates significantly lower than PLM nausea and tacrolimus) were selected for further investigation rates. using both PLM data and detailed information from To compare the degree of over or underreporting of medication labels. nausea rates among patients from PLM relative to the rates reported in medication labels, the reporting ratio Ritonavir was calculated as the ratio of PLM nausea rates to medi- In the current study, ritonavir had the highest observed cation label nausea rates. Across 150 treatments with magnitude of absolute difference between medication non-zero medication label nausea rates, reporting ratios label (57.4%) and PLM (2.7%) nausea rates. This discrep- ranged from 0 to 13. The median reporting ratio was ancy cannot be attributed to a difference in indication Table 1 Medications with largest absolute differences in nausea rates between PLM data and medication label data Medication name PLM Medication Label Overall Difference Nausea % N Nausea % (median) N Ritonavir 2.7% 75 57.4% 1755 -55% Mycophenolate mofetil 9.7% 257 54% 896 -44% Fluvoxamine 0% 57 39% 1295 -39% Tacrolimus 1.9% 215 38.5% 931 -37% Mitoxantrone 20.3% 79 55% 148 -35% Trihexyphenidyl 5.6% 54 40% Not reported -34% Fentanyl transdermal patch 8.6% 220 41% 216 -32% Venlafaxine 6.5% 1035 37% 4857 -31% Naltrexone 3.9% 407 29% 440 -25% Desvenlafaxine 7.1% 197 31% 1365 -24% Values for all medications are available [see Additional file 1]. PLM Nausea Rate [%] Blaser et al. AAPS Open (2017) 3:10 Page 6 of 10 60% Within Bonferroni Adjusted Confidence Interval label rate outside of adjusted CI label rate within adjusted CI 40% 20% 0% Medication Fig. 4 Bonferroni-adjusted confidence internals for PLM nausea rates versus medication label nausea rates. For each generic treatment, PLM nausea rates and their associated 95% Bonferroni-adjusted confidence intervals are positioned along the horizontal axis, and medication label nausea rates are shown as a black line. In cases where the medication label nausea rate falls outside of the associated 95% confidence interval based on PLM data, treatments are colored red, identifying instances in which drug label nausea rates and PLM nausea rates were identified as having notable differences, after accounting for multiple comparisons. Since not all drug labels reported the number of patients included in the associated study, it was not possible to assess the statistical significance of the differences between PLM and label nausea rates for each drug for medication use, as the drug is listed as a treatment doxycycline use among patients from PLM included for HIV infection in both data sources. Instead, the ob- Lyme disease (37%) and acne (14%), with reported daily served difference is likely due to widely varying dosage dosages ranging from 20 to 400 mg/day (median daily information between the medication labels and PLM. In dose 100 mg/day). In comparison, information reported PLM, all but one patient reported a daily dosage of in the medication label was based on a clinical trial of 100 mg/day (a single patient reported a daily dosage of patients with periodontal disease who received a daily 80 mg/day). In contrast, the medication label corre- dose of 40 mg/day (Periostat® [package insert], 2001). sponds to patient experiences while taking a daily dose of 1200 mg/day. Amoxicillin/Clavulanate Like that of doxycycline, the nausea rate reported by pa- Doxycycline tients from PLM for amoxicillin/clavulanate (11.5%) The doxycycline nausea rate in PLM (27.1%) exceeded exceeded that reported in the medication label (2.6%). the medication label nausea rate (8%) by the largest Amoxicillin/clavulanate is indicated for a wide range of magnitude, which may be explained by differences in conditions; indications included sinusitis, urinary tract both dosing and indication for use. Indications for infections, and other systemic infections in both PLM Fig. 5 Empirical and theoretical lognormal distributions of reporting ratio. Empirical distribution of the reporting ratio of PLM nausea rates to medication label nausea rates and a theoretical lognormal distribution with (natural log scale) mean − 1.01 and standard deviation 1.31. Distributions based on data from 150 treatments with non-zero medication label nausea rates Nausea Rate Blaser et al. AAPS Open (2017) 3:10 Page 7 of 10 data and medication labels. Corresponding daily doses reported in medication labels reflect the overall rate for recommended by clinical guidelines also vary across in- nausea that patient experienced, but not the rate that is dications, ranging from 750 to 4000 mg/day (Amoxicillin likely to be attributed to the medication alone. Trau- and clavulanate, 2016). Unlike doxycycline and ritonavir, matic, invasive transplant surgeries and/or pain medica- there are no clear patient or product use characteristics tions are likely a significant contributor to nausea that appeared responsible for differences in nausea rates following a transplant. Patients evaluating treatments on between these data sources. Influenced by the relatively PLM may differentiate between nausea associated with low number of patients who have evaluated amoxicillin/ the overall experience of receiving an organ transplant clavulanate on the PLM platform (n = 61), the aggregate and nausea caused by transplant-related medications, medication label nausea rate for amoxicillin/clavulanate leading to lower reported rates for nausea as compared falls within the Bonferroni adjusted 95% confidence to those reported on medication labels. interval for the proportion of patients experiencing nau- sea in the PLM community. Discussion The goal of this study was to closely examine rates of Mitoxantrone patient-reported medication side effects from a real- The aggregate nausea rate reported in the medication world data source and from medication labels. This label for mitoxantrone (55%) was higher than that re- study showed that PLM nausea rates for 163 treatments ported in PLM (20.3%). A more detailed examination of ranged from 0% to 36%, with a median of 4%. Across a the medication label data suggests that the expected dif- broad range of indications and dosages, PLM nausea ference may be even more striking than the observed rates were positively correlated to those reported on difference. The median medication label nausea rate of medication labels. Moreover, this study demonstrated a 55% is based on data collected over the course of three quantifiable relationship between the reporting ratio of clinical trials, one with a nausea rate of 76% (n = 62, PLM nausea rates to medical label nausea rates, which dose = 12 mg/m ), the second with a rate of 55% provides a useful basis for comparisons against alterna- (n = 65, dose = 5 mg/m ), and a third trial where mitox- tive data sources (e.g. patient report to a physician). antrone was administered concurrently with methyl- Medication labels are often considered the definitive prednisolone and had a nausea rate of 29% (n = 21) data source for information on side effects due the (Novantrone® [package insert], 2012). The current rec- strenuous requirements of regulatory approval. However, ommended daily dose of mitoxantrone is 12 mg/m , any data source will have strengths and limitations. which is more than double the dose administered in the Many medications were approved decades ago, and up- cohort of patients whose reported nausea rate was 55% dates to their labels are infrequent. The labels also do (Marriott et al., 2010). Furthermore, administration of not contain information related to off-label indications. mitoxantrone in combination with an anti-emetic medi- An estimated 20% of outpatient prescriptions in the cation, such as methylprednisolone, would likely result United States are written for off-label use (Radley et al., in lower overall reported rates of nausea. It seems that 2006). For example, the label for amitriptyline contains the median medication label nausea rate (55%) may ac- information from clinical trials conducted over 50 years tually be an underestimate of the nausea rate observed ago in patients with depression. However, prescribing during a clinical trial that more closely mirrors real- practices have evolved significantly for amitriptyline so world prescribing practices (76%), leading to an even that the proportion of off-label use is very high (>80%) greater discrepancy between the nausea rates reported in with the most frequent use in fibromyalgia, neuralgias PLM and those reported in the medication label. and migraine prevention (Radley et al., 2006; Moore et al., 2015; Silberstein, 2015). Organ transplant medications Differences in reporting rates of nausea between medi- Our review also found that nausea rates reported on cation labels and in the PLM database may be influenced PLM were much lower than those reported in medica- by various factors, including methodological differences tion labels for the medications tacrolimus (PLM – 1.9%; in data collection and differences in indications for medication label – 38.5%), mycophenolate mofetil (PLM medication use. Data for all drugs from within the PLM rate – 9.7%; medication label – 54%), and cyclosporine database are collected through one consistent mechan- (PLM rate – 6.9%; medication label – 14%) which are ism, whereas data collected through clinical trials for use typically taken after organ transplantation to prevent re- in medication labels may be obtained using varying jection. Product use characteristics for these medications methods. Prior research has shown that the methods were examined in greater detail to determine what may used to collect side effect data significantly influence the be contributing to these wide differences. One potential results obtained (Bent et al., 2006). For example, sexual explanation for this observed trend is that nausea rates side effects due to use of selective serotonin reuptake Blaser et al. AAPS Open (2017) 3:10 Page 8 of 10 inhibitors ranged from 2% to 73% depending primarily Limitations on whether side effects were collected via a checklist, The side effect nausea can be reported in a variety of open-ended questioning, or by a detailed inquiry (Safer, ways (i.e., qualitatively or quantitatively) on medication 2002). labels. In order to collect data that were consistent with In the case of this study, we speculate that the general information reported in medication labels, PLM nausea trend of lower nausea rates in the PLM database com- cases were identified only as instances in which patients pared to those reported in corresponding medication la- specifically reported nausea. Patient reports of side ef- bels is driven largely by differences in data collection fects known to have a high co-occurrence with nausea, mechanisms. On the PLM platform, patients report side such as vomiting, were not considered as nausea cases. effects using a standard method across all drugs. Once a In cases where nausea was not mentioned on a drug patient indicates that they have experienced any particu- product label, the label nausea rate was approximated as lar side effect, they are presented with an entry field with zero for analysis purposes. FDA guidelines state that the an auto-completer that provides suggestions as the user Adverse Reactions section of a medication label does not types. include all adverse events that occurred during a clinical Although the FDA requires that studies investigating trial, but only those that meet an appropriate cutoff (e.g. new drugs report adverse events, they do not specify a all adverse reactions occurring at a rate of 10% or standard method for data collection. One common ap- greater in the treatment group and at least twice the rate proach accepted by the FDA is to ask patients to of placebo) (US Food and Drug Administration, 2006). complete a checklist of symptoms, which is likely to cap- Due to variable reporting methods, aggregating nausea ture additional symptoms unrelated to the particular rates from medication labels proved particularly challen- medication of interest (U.S. Department of Health and ging. In many instances, more than one value was given Human Services, Food and Drug Administration, 2005; for nausea as a side effect in a single medication label, Smith et al., 2013). Conversely, the process of collecting and sample sizes were not always available for each indi- data in the PLM database is more open-ended than that vidual rate, leading to challenges in calculating an overall used in clinical trials, leading patients to be more likely nausea rate. to report only their most bothersome side effects. Each In addition, aggregating nausea rates across multiple approach has both advantages and disadvantages; how- labels for the same active ingredient was challenging due ever, the use of different mechanisms for data collection to inconsistencies or redundancies in the information. across clinical trials limits the ability of researchers to For example, the medication label for levetiracetam accurately compare side effect profiles across drugs. In (Keppra®) does not list nausea as a side effect, but the contrast, the consistent method of collecting side effect label for extended-release levetiracetam (Keppra XR®) data from the PLM platform is advantageous to making lists a nausea rate of 5% (Keppra® [package insert], 2013; more valid comparisons among drugs’ safety profiles. Keppra XR® [package insert], 2016). Upon further inves- The examples highlighted in the Results section illus- tigation, it was found that nausea was previously listed trate that differences in dosage and indication contribute in the levetiracetam (Keppra®) label, described as an to differences in reported nausea rates. There are likely “event reported by at least 1% of adult levetiracetam other factors that additionally contribute to these differ- (Keppra)-treated patients but as or more frequent in the ences, including concurrent treatments, comorbidities, placebo groups.” This description, however, has been re- disease severity, and other patient population character- moved in updates to the medication label over time. Fur- istics (e.g. age, gender, location). However, information thermore, it was often difficult to determine if nausea related to these subgroups was seldom reported in medi- rates for different labels of the same active ingredient cation labels. were from the same clinical trial, as was often the case In addition to the methodological challenges associ- for immediate-release and extended-release products. ated with comparing side effects among many medica- Some medication labels included nausea rates for in- tions, differences in medication label data reporting gredients similar to the active ingredient, as was the case presented additional challenges. Data can be presented for the valproate sodium label, which contained informa- quantitatively (e.g. 5%) or qualitatively (e.g. “commonly” tion from trials of divalproex sodium. Similarly, some or “frequently”). Nausea, in particular, is reported in medication labels included nausea rates for a class of various ways, often combined with rates of vomiting and medications rather than for only the active ingredient. reported as one value. The variety of methods used for For example, the label for minocycline contains informa- 1) collecting information about side effects in clinical tri- tion on side effects of tetracyclines, a broad class of anti- als and 2) reporting information in medication labels biotics that includes minocycline, but it did not provide limits the ability to make meaningful comparisons be- information about side effects specific to minocycline tween medications. (Minocycline hydrochloride [package insert], 2012). Blaser et al. AAPS Open (2017) 3:10 Page 9 of 10 The underreporting of side effect rates by patients in Acknowledgements We are grateful to all the members of PatientsLikeMe that have contributed the PLM community may also be influenced by the ex- their data and continue to make this research possible. We also thank Jill M. istence of survival bias. Although patients are able to Serrano, PhD for her assistance in preparing and writing this manuscript. enter their entire medication history on PLM, they are Availability of data and materials prompted to enter more detailed information about ex- All data generated or analysed during this study are included in this article periences with current treatments. Patients, who stop [Additional file 1 and Additional file 2]. taking a particular medication due to intolerable side ef- fects, may not report a detailed history for that medica- Authors’ contributions DAB, SE, JL, and PP wrote the manuscript. DAB, SE, JL, SR, PW, and JW tion compared to medications they are currently taking. designed the research. DAB, SE, JL, SR and PP performed the research and Prompting patients to report on their current treatments analyzed the data. All authors read and approved the final manuscript. may lead to lower reported rates of nausea in the PLM Ethics approval and consent to participate database. Independent ethics review was not sought as members (patients) of the A common critique of Internet communities is that PatientsLikeMe Community are informed of their involvement in research patients may not be able to accurately report their diag- activities via the User Agreement and Privacy Policy before joining the site. The User Agreement states that “Member Notices: If you register as a nosis and medication history, or that anonymity provides member, you agree that PatientsLikeMe may send notices to you by email at an opportunity for individuals to report misinformation. the email address you provide when registering to become a member (or A previous study demonstrated that a large percentage which you later update using the functionality of the Site).” and the Privacy Policy states “PatientsLikeMe may also periodically ask Members to complete of the patients on PLM users can be matched to real pa- short surveys about their experiences (including questions about products tients with confirmed identities, and there is a high de- and services). Member participation in these surveys is not required, and gree of agreement between self-reported and insurance refusal to do so will not impact a Member's experience on the Site.” All data used for analysis of medication labels was from published sources. claims for diagnoses and medication history (Eichler et al., 2016). Other studies have also shown high degree Competing interests of agreement between self-reported comorbidities and DAB, SE, and PW are employees of PatientsLikeMe and hold stock options in medical records (Ye et al., 2017). the company. The PatientsLikeMe Research Team has received research funding (including conference support and consulting fees) from Abbvie, Accorda, Actelion, Alexion, Amgen, AstraZeneca, Avanir, Biogen, Boehringer Conclusions Ingelheim, Celgene, EMD, Genentech, Genzyme, Janssen, Johnson & Johnson, Merck, Neuraltus, Novartis, Otsuka, Permobil, Pfizer, Sanofi, Shire, In the current study, we observed a positive correlation Takeda, Teva, and UCB. The PatientsLikeMe R&D team has received research between nausea rates reported in medication labels and grant funding from Kaiser Permanente, the Robert Wood Johnson those self-reported by patients in an online community. Foundation, Sage Bionetworks, The AKU Society, and the University of Maryland. PW is an associate editor at the Journal of Medical Internet However, in general, nausea rates reported among pa- Research and is on the Editorial Boards of The BMJ and BMC Medicine. JW, tients in the PLM community were lower than those re- JL and SR are employees of AstraZeneca. JW and SR hold stock options in ported in medication labels. While there are likely many the company. factors driving the discrepancy between these two sources, including differences in data collection mecha- Publisher’sNote nisms and product use factors, this study demonstrates a Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. quantifiable relationship in side effect reporting frequen- cies between sources of real-world evidence and the Author details 1 2 controlled environment of clinical trials. Reporting fre- PatientsLikeMe, 160 2nd Street, Cambridge, MA 02142, USA. AstraZeneca Research & Development, Advanced Analytics Centre, Mereside, Alderley quencies of side effects have high impact on regulator, Park, Macclesfield, Cheshire SK10 4TF, UK. Northeastern University, 360 payer, physician, and patient decisions; therefore, collec- 4 Huntington Ave, Boston, MA 02115, USA. AstraZeneca Research & tion of data that accurately reflect the patient experience Development, Granta Park, Riverside 2, Cambridge CB21 6GH, UK. is critical. Using self-reported patient data may be one Received: 28 April 2017 Accepted: 13 October 2017 way to enhance the capture of subjective elements of medication tolerability and safety. References 21st Century Cures Act, HR 34, 114th Congress. (2015–2016). http://docs.house. Additional files gov/billsthisweek/20161128/CPRT-114-HPRT-RU00-SAHR34.pdf. Accessed 19 Oct 2017 Additional file 1: Datafile of nausea rates extracted from medication Amoxicillin and clavulanate: Drug information. In: UpToDate, Post TW (Ed), labels and PatientsLikeMe. (CSV 14 kb) UpToDate, Waltham, MA. Accessed on December 15, 2016 Basch E, Jia X, Heller G, Barz A, Sit L et al (2009) Adverse symptom event Additional file 2: Codebook for datafile of nausea rates extracted from reporting by patients vs clinicians: relationships with clinical outcomes. J Natl medication labels and PatientsLikeMe. (CSV 1 kb) Cancer Inst 101(23):1624–1632 Bent S, Padula A, Avins AL (2006) Brief communication: Better ways to question Abbreviations patients about adverse medical events: a andomized, controlled trial. Ann FDA: Food and drug administration; PGHD: patient-generated health data; Intern Med 144(4):257–261 Erratum in: Ann Intern Med. 2006 Jul 18;145(2): PLM: PatientsLikeMe; UNII: Unique ingredient identifier 156 Blaser et al. AAPS Open (2017) 3:10 Page 10 of 10 Blenkinsopp A, Wilkie P, Wang M, Routledge PA (2007) Patient reporting of US Food and Drug Administration. Adverse Reactions Section of labeling for suspected adverse drug reactions: a review of published literature and human prescription drug and biological products—content and format. international experience. Br J Clin Pharmacol 63(2):148–156 http://www.fda.gov/downloads/Drugs/ Eichler GS, Cochin E, Han J, Hu S, Vaughan TE et al (2016) Exploring concordance GuidanceComplianceRegulatoryInformation/Guidances/ucm075057.pdf. of patient-reported information on PatientsLikeMe and medical claims data Published 2006. Accessed 19 Oct 2017 at the patient level. J Med Internet Res 18(5):e110 Weingart SN, Gandhi TK, Seger AC, Seger DL, Borus J et al (2005) Patient-reported medication symptoms in primary care. Arch Intern Med 165(2):234–240 European Medicines Agency. Fourth report on the progress of the interaction Wood WA, Bennett AV, Basch E (2015) Emerging uses of patient generated with patients’ and consumers’ organisations (2010) and results/analysis of the health data in clinical research. Mol Oncol 9(5):1018–1024 degree of satisfaction of patients and consumers involved in EMA activities World Health Organization. The importance of pharmacovigilance - safety during 2010. 2011. http://www.ema.europa.eu/docs/en_GB/document_ monitoring of medicinal products. 2002. http://apps.who.int/medicinedocs/ library/Report/2011/10/WC500116866.pdf. Accessed 19 Oct 2017 en/d/Js4893e/. Accessed 19 Oct 2017 Hakobyan L, Haaijer-Ruskamp FM, de Zeeuw D, Dobre D, Denig PA (2011) Review Ye F, Moon DH, Carpenter WR, Reeve BB, Usinger DS et al (2017) Comparison of of methods used in assessing non-serious adverse drug events in patient report and medical Records of Comorbidities: results from a observational studies among type 2 diabetes mellitus patients. Health Qual population-based cohort of patients with prostate cancer. JAMA Oncol Life Outcomes 9:83 Keppra XR® [package insert]. UCB, inc.; Smyrna, GA; 2016. http://www.accessdata. fda.gov/drugsatfda_docs/label/2016/022285s022lbl.pdf. Accessed 19 Oct Keppra® [package insert]. UCB, Inc.; Smyrna, GA; 2013. http://www.accessdata.fda. gov/drugsatfda_docs/label/2015/021035s093,021505s033lbl.pdf. Accessed 19 Oct 2017 Marriott JJ, Miyasaki JM, Gronseth G, O'Connor PW (2010) Therapeutics and technology assessment Subcommittee of the American Academy of neurology. Evidence report:the efficacy and safety of mitoxantrone (Novantrone) in the treatment of multiple sclerosis: report of the therapeutics and technology assessment Subcommittee of the American Academy of neurology. Neurology 74(18):1463–1470 Methotrexate sodium [package insert]. Xanodyne Pharmacal, Inc., Florence, KY; 2003. http://www.accessdata.fda.gov/drugsatfda_docs/label/2004/ 11719slr106_methotrexate_lbl.pdf. Accessed 19 Oct 2017 Minocycline hydrochloride [package insert]. Teva Pharmaceuticals USA inc., Sellersville, PA.; 2012. http://www.accessdata.fda.gov/drugsatfda_docs/label/ 2013/063009s034lbl.pdf. Accessed 19 Oct 2017 Moore RA, Derry S, Aldington D, Cole P, Wiffen PJ (2015) Amitriptyline for neuropathic pain in adults. Cochrane Database Syst Rev 7:CD008242 Novantrone® [package insert]. EMD Serono, Inc., Rockland, MA; 2012. http://www. accessdata.fda.gov/drugsatfda_docs/label/2012/019297s035lbl.pdf. Acccessed 19 Oct 2017 Pakhomov SV, Jacobsen SJ, Chute CG, Roger VL (2008) Agreement between patient-reported symptoms and their documentation in the medical record. Am J Manag Care 14(8):530–539 Periostat® [package insert]. CollaGenex pharmaceuticals, Inc., Newton, PA; 2001. http://www.accessdata.fda.gov/drugsatfda_docs/nda/2001/50-783_Periostat_ prntlbl.pdf. Accessed 19 Oct 2017 R Core Team. R: a language and environment for statistical computing. R foundation for statistical computing. 2016. https://www.r-project.org/. Accessed 19 Oct 2017 Radley DC, Finkelstein SN, Stafford RS (2006) Off-label prescribing among office- based physicians. Arch Intern Med 166(9):1021–1026 Research Center for Devices and Radiological Health. Guidance for Industry Patient-Reported Outcome Measures: Use in medical product development to support labeling claims. 2009. http://www.fda.gov/downloads/Drugs/ GuidanceComplianceRegulatoryInformation/Guidances/UCM193282.pdf. Accessed 19 Oct 2017 Safer DJ (2002) Design and reporting modifications in industry-sponsored comparative psychopharmacology trials. J Nerv Ment Dis 190(9):583–592 Shapiro M, Johnston D, Wald J, Mon D. Patient-Generated Health Data: White Paper Prepared for the Office of the National Coordinator for Health Information Technology by RTI International. 2012. http://www.rti.org/sites/ default/files/resources/patientgeneratedhealthdata.pdf. Accessed 19 Dec 2016 Silberstein SD (2015) Preventive migraine treatment. Continuum (Minneap Minn) 21(4 Headache):973–989 Smith SM, Wang AT, Katz NP, McDermott MP, Burke LB et al (2013) Adverse event assessment, analysis, and reporting in recent published analgesic clinical trials: ACTTION systematic review and recommendations. Pain 154(7):997–1008 U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER). Reviewer Guidance: Conducting a Clinical Safety Review of a New Product Application and Preparing a Report on the Review. 2005. http://www.fda.gov/downloads/ drugs/guidancecomplianceregulatoryinformation/guidances/ucm072974.pdf. Accessed 19 Oct 2017

Journal

AAPS OpenSpringer Journals

Published: Nov 20, 2017

There are no references for this article.