Spontaneous Reporting on Adverse Events by Consumers in the United States: An Analysis of the Food and Drug Administration Adverse Event Reporting System Database

Recent changes in pharmaceutical markets and regulations, including introduction of new accelerated approval pathways [1‐3], new lines of therapies with innovative pharmacological mechanisms from molecular-targeted drugs to immune checkpoint inhibitors and to cell therapies, and growing expectations from patients for new therapies under development, have prompted regulators and industries to introduce new drugs faster and more efficiently [4]. These trends have made postmarketing pharmacovigilance more important than ever [5]. The United States (US) Food and Drug Administration (FDA) Adverse Event (AE) Reporting System (FAERS) database, one of the largest pharmacovigilance databases, plays a key role in both collecting and providing data on drug-related AEs. In the US, the FAERS database gives the FDA critical signals and a decision basis to take regulatory actions such as ordering labeling changes in the warning and/or precaution sections to improve drug use in markets [6, 7]. Epidemiologists worldwide use the database to detect novel drug-related safety events, to identify possible mechanisms of AEs, and to explore efficient methods to detect potential drug-related AEs [8‐14]. The database has been used beyond the realm of safety. For example, several researchers have recently studied drug–drug interactions with new approaches using the FAERS database, which could lead to discovery of promising new concomitant uses of drugs in certain therapeutic areas [15, 16].

For all purposes, the integrity of databases is a critical condition for users to obtain unbiased conclusions. Under-reporting has been a serious issue that afflicts pharmacovigilance activities worldwide [6, 17]. Previous studies showed that under-reporting was caused by many factors, including inevitable dependency on reporters volunteering incentives and lack of awareness of how to use public reporting systems, or even their existence [18, 19]. A practical approach to improve the situation of under-reporting is to publicize the reporting system and to encourage not only healthcare professionals but also the consumers who actually experience AEs, and their families, to submit AE reports to the FDA. The FDA has also made efforts to reduce undesirable disproportionality (e.g., over- or under-representation of specific populations) in the FAERS. To alleviate these concerns, the FDA has continuously expanded the MedWatch program for more than 20 years. The FDA introduced the first voluntary reporting form FDA3500 in 1993 and the form FDA3500B in 2013, a voluntary consumer-friendly reporting form, to encourage reporting by patients [20, 21].

It is important for database users to acknowledge basic characteristics of spontaneous safety databases, including types and variations of data sources, because the internal and external validity of analyses using the databases totally depend on them. Regarding that perspective, however, spontaneous reports have attracted less attention than obligatory reports from the industry. There are a couple of studies on spontaneous reports by nurses and physicians in locally established databases of AE reports, but there have been few analytical studies focusing on spontaneous reporting by consumers in the US [22‐25]. In the European Union (EU), patient-oriented reporting has grown gradually since the 2000s, and current situations have been described in previous studies [25‐30]. In one study it was concluded that patient reporting successfully complemented reporting by healthcare professionals, and the conclusion was consistent with that of a study in the UK [31, 32]. Another study using the database of the Dutch Pharmacovigilance Center Lareb showed that patients report clinical information at a similar level to healthcare professionals.

In this study, we aimed to identify the characteristics of recent voluntary reports submitted by consumers and compare them with those by healthcare professionals. We discussed what backgrounds have led to the observed differences, especially focusing on how information available on AEs could influence reporters’ attitudes to AE reporting.

We analyzed the FAERS database and examined the reports submitted by consumers and healthcare professionals. We showed the transition of the numbers of voluntary reports in the US (Fig. 1), and conducted several descriptive analyses to reveal the characteristics of current voluntary reports. All the descriptive analyses were based on the reports (25,814 reports) in the first and second quarter of 2016, and the analyses related to primary suspect drugs were based on reports for which primary suspect drugs were registered in the SIDER4.1 database.

Our analysis showed that consumers’ reports, which have accounted for a significant portion of recent increases in the number of voluntary reports, seem to have information that is not necessarily provided by healthcare professionals. Differences were observed not only in AEs, suspect drugs, and health outcomes, but also in reporting quality and behaviors such as report completeness and time to report.

In the US, the FDA has facilitated voluntary reports by consumers through several activities such as the release of a new consumer-friendly reporting form FDA3500B and the refurbishment of the interface of its online reporting system through the MedWatch program in 2013 [20, 21]. These efforts apparently have achieved an excellent outcome in enhancing voluntary reporting. Figure 1 indicates that the total number of voluntary reports in the US has increased, especially since 2013, and that consumers seem to have contributed to that increase. This was probably due to encouragement and increased public recognition that patients, as well as healthcare professionals, could report AEs directly to the FDA, especially using the online reporting form.

The increase in the number of reports is beneficial to overall pharmacovigilance activities, but it inevitably draws experts’ attention to increasing heterogeneity in the safety database due to diversified reporting sources. Needless to say, statistical inconsistency between different periods is a problem. For data mining purposes, changes in the quality of information must be considered in statistically rigorous ways [17, 35]. Besides the statistical aspects, safety authorities and drug companies need to decide whether and how we should prioritize concerns for specific safety issues presented by diverse reporters in current public health needs. All these issues make it necessary to clarify how consumers are different from traditional reporters, mostly healthcare professionals, and also to test whether reports from consumers who choose the new reporting form and routes are different from reports by consumers who choose (or chose) the traditional reporting form.

Our analysis identified several interesting features in US voluntary reports. We found some disparities in patient sex reported by different occupations. Previous studies showed that female patients had a 1.5- to 1.7-fold greater risk of developing an ADR, which might be caused by sex differences in pharmacological response [36, 37]. Another recent study presented the same sex disparity and also showed that healthcare professionals tended to report ADRs of male patients more than other types of reporters did [28]. Our results indicate that consumers were more likely to report AEs of female patients than were healthcare professionals (Fig. 2), which was concordant with the findings of previous studies.

Regarding reported patient outcomes, we found that consumers were likely to report AEs that caused disability in patients. In contrast, healthcare professionals rarely reported disability cases. This indicates that reports by consumers can add information about AEs that may not be emphasized by healthcare professionals, and this is consistent with a recent systematic review of the literature on patient reporting [38].

Report completeness has been considered an indicator for well documented AE reports and used for data mining in the World Health Organization’s pharmacovigilance database VigiBase [35, 39]. A recent paper showed that voluntary reports have a higher level of report completeness than do reports from drug companies: the completeness of all four items (sex, age, event date, and medical terms) was 86.2% in serious reports submitted directly to the FDA (i.e., voluntary reports), as compared with 40.4% in manufacturer-expedited reports and 51.3% in manufacturer periodic reports, in 2014 [40]. We examined three items (i.e., event date, indication, and route of primary suspect drug) for which data are frequently missing in the FAERS database, and other important items including sex, age, and weight. The report completeness rates for these items were almost the same in reports from different reporters, but some findings suggest possible differences in reporters’ interest and environment. For example, consumers’ reports lack the event date more often than do the other reports (Online Resource 1, see ESM). This may be unsurprising, however, because AEs are detected, recorded, and reported retrospectively based on the records and/or memories available to the reporters. The fact that consumers showed the longest time to report with a median time of 22 days (Fig. 3) may support the finding.

The number of concomitant drugs may reflect reporters’ availability of information, and how meticulous the reporters are within real-world constraints. We found that consumers reported more concomitant drugs than did other reporters (Online Resource 1, see ESM). Our preliminary analysis showed that the number of concomitant drugs did not correlate with report completeness for any other items, suggesting that this would not be a quality measure for general purposes.

In general, consumers are thought to have less access to information on drug-related AEs in drug labels than do healthcare professionals. Consumers do not have expertise in pharmacovigilance activities, either. We had expected that previous knowledge about AEs would affect reporting behaviors differently for consumers and healthcare professionals, but our analysis using all the reports did not show a clear difference between the two groups (Fig. 4a). This is basically in line with a conclusion in a previous report that patients can make causality assessment based on the available information [31, 32]. We further examined serious ADR cases and found that known ADRs were reported more slowly than unknown ADRs by PHs, and vice versa by OTs (Fig. 4b). This suggests differences seem to exist among healthcare professionals.

Histories of labeling revisions were associated with observed differences between consumers and healthcare professionals in what and how they are likely to report (Online Resource 2 and 3, see ESM). A possible explanation is that healthcare professionals, and MDs in particular, are less likely to be influenced by prior information about risk and/or more likely to adhere to their own judgement than are consumers. However, it is difficult to discuss the role of prior information and environment based solely on our findings, because there are many confounding factors. The issuance of regulatory alerts, for example, is a possible confounder that has been investigated intensively in many studies [41, 42].

Time to report is an interesting measure of promptness in reporting and may help to describe reporting behaviors. A recent study in the EU showed that median time to report an ADR was approximately 30 days in spontaneous reports, regardless of whether it was reported by patients or healthcare professionals [28]. As shown in Fig. 3a, the median time to report AEs by any reporter in the US was shorter than that in the EU. These differences in time to report are likely to reflect differences in reporting pathways: the voluntary AE reports in this research were directly submitted to the FDA, while some reports in the EU arrived at the authority via companies or other regulatory agencies. Different reporting times may also be influenced by different cultures and histories surrounding the reporting systems and use of drugs, such as publicity around medicinal products, as a previous report suggested [43]. With respect to suspect drugs, Table 2 indicates that MDs report AEs of oncologic injection drugs (e.g., cisplatin and carboplatin) but consumers do not, which suggests that MDs and consumers report different AEs in different settings, and may explain our observation (at least partly) that time to report was much shorter in MDs than in consumers. It is worthwhile investigating whether this reflects differences in the publicity and maturation of the AE reporting system between the two regions.

It was also shown that time to report by consumers was longer than that by pharmacists and other healthcare professionals, and almost equal to that by physicians (Fig. 3a). Comparisons between serious and non-serious cases showed that time to report was longer in serious cases than in non-serious cases except in the reports by physicians (Fig. 3b). With regard to whether the AE(s) were already written in the labels, interesting differences were observed between healthcare professionals in reports with serious outcome(s) (Fig. 4b). Safety labeling changes of primary suspect drugs were associated with the time to report of healthcare professionals but not with those of consumers (Online Resource 3, see ESM). Time to report was associated with which section of labeling (i.e., black box warning, warning, or precaution) had been revised, but in somewhat complicated ways. Interestingly, pharmacists and other healthcare professionals tended to report earlier the primary suspect drug for which safety labeling changes occurred in the warning or precaution sections, but not in the black box warning.

Causalities behind these findings on time to report are complex and beyond our scope, but they probably reflect diversities in the environment where reporters come to experience and/or be aware of AEs and decide to report AEs to the authority. Some consumers may have difficulties in reporting their own AE immediately after their recovery, and need to take time to learn how to report AEs even when they intend to do so. Physicians, who commonly struggle with time conflicts on a daily basis, may face tradeoffs between voluntary and mandatory reporting. Causal determination for AEs did not seem to play a critical role as shown in Fig. 4 and Online Resource 3 (see ESM). This is in line with a previous report that some consumers may be able to identify suspected ADRs adequately [31, 32]. To improve efficiency and responsiveness in the current reporting system, we need to consider real-world mechanisms of reporting behaviors. For example, we need to examine the possible influence of communications such as ‘Dear Healthcare Provider’ letters, which are expected to have substantial impacts on reporting behaviors [41, 42].

Our analysis provided several clues to the heterogeneity in reports from consumers, which were conveyed in one of the two different forms: the traditional form FDA3500 (tagged as ‘CN’ in this paper) and the new form FDA3500B (tagged as ‘UN’). It is apparently the latter form that has contributed to the recent increase in the number of AE reports. Although the traditional form users and the new form users were similar in some aspects (e.g., concomitant drugs, report completeness), they were apparently different in other important aspects including reported outcomes and indications, and time to report. It is quite natural that consumers use forms differently when a new reporting form and/or route is added to traditional forms/routes, and this might be exactly what regulators intended to facilitate. Our results suggest that database users have to be careful about such heterogeneities when combining or pooling reports made using the different forms, even though they were all submitted by ‘consumers.’

Finally, we can discuss our findings with reference to previous studies in the EU. As discussed above, the average time to report in the US was much shorter than that in the EU, which may reflect the extent of differences in regulations and reporting pathway(s) [28]. Reporting behaviors in consumers, including promptness to report, may be affected by marketing environments such as direct-to-consumer advertising [43]. Irrespective of these differences between the US and European countries, we found that consumers’ reports in both regions are similar in some important aspects. For example, consumers in both regions tend to report subjective symptoms such as headache, fatigue, and nausea [28]. Avery et al. investigated patient reporting in UK’s Yellow Card Scheme and discussed how patient reporting may provide a positive complementary contribution to that of healthcare providers; however, the combination of reports from patients and healthcare providers, when used for the purposes of signal detection through disproportionality analysis, may result in the loss of some information [31, 32]. Our findings in the US lead to similar conclusions to these previous studies in European countries.

This study has several limitations. Because it relied on spontaneous reporting data, all the limitations inherent to spontaneous reporting are applicable. We focused on consumers’ reporting, but lack of publicly available information about the choice of reporting forms and reporters’ occupations made it difficult to extract data from ‘real’ consumers. There are no official rules and/or principles on how to handle consumers’ reports in publication and data mining, and our analysis had to be done under such uncertainties. Our analysis did not cover mandatory reports, which limits the generalizability of our results. A significant portion of AE reporting is done as mandatory reporting, and drug companies play a key role in it. Healthcare professionals choose reporting route(s) considering types and seriousness of AEs, applicable rules, and many other factors including opportunity costs of reporting. These considerations affect how healthcare professionals report AEs voluntarily. Further studies are needed to shed light on such broader aspects and examine the system as a whole.

Our analysis of voluntary AE reports in the US FAERS database has shown the characteristics of spontaneous reporting in the US. Voluntary reports tended to include AEs related to subjective symptoms, as in some previous studies on patient reporting in the EU. Voluntary reports by consumers seemed to be different from reports by healthcare professionals in demographics and outcomes of patients, and suspect drugs. They were also different in report completeness and time to report, which may reflect concerns and environments that are specific to each type of reporter. Consumers’ choice of voluntary reporting forms may be worth studying further. These findings suggest that the heterogeneities should be addressed appropriately when using spontaneous reports, especially in the context of international comparison.