In this study, we present one of the first prospective community-based active surveillance studies to assess NHP related AEs and NHP-drug interactions. Although not without its challenges, active surveillance in pharmacies is feasible, and markedly increases both the number and quality of AE reports in comparison to passive surveillance. 
When comparing our results to data derived from passive surveillance, we found that the Health Canada AE database (Canada Vigilance) has a total of 1544 (median
144 in 2004 to 442 as of December 31, 2009) AE reports associated with NHP use over a five year period from passive surveillance of over 30 million Canadians, which results in a median incidence rate of approximately 0.0008% (range
0.0005% in 2004 to 0.0015% in 2009). 
Given that one third of Canadians report taking NHP and prescription medications concurrently, the median incidence rate of AE reports is approximately 0.0023% (range
0.0014% in 2004 to 0.0044% in 2009). 
These values are far lower than AE rates reported by individuals taking NHPs and drugs concurrently in the SONAR study (7.4%; 95% CI: 6.0% to 9.2%), suggesting that active surveillance improved AE reporting rates substantially. It is important to note that the population screened in our study may represent a different population than those screened through passive surveillance. In particular, patients being screened in a pharmacy may be more likely to take prescription medications; since we hypothesized this may increase the risk of potential AEs, this was the population of greatest interest to our study team.
Of note, without adjudication, it is not clear how many of the AE reports from either active or passive surveillance are causally linked to NHP use. In its surveillance system, Health Canada applies a signal detection process to the incoming AE data associated with all health products, including NHPs, and causality assessment is conducted accordingly. However, often the quality of AE reports is poor and important details allowing for proper causality assessment are missing. 
Studies from several countries report that NHP use is extremely common. While several have assessed concurrent use with prescription medications, 
few have reported on the frequency of NHP-drug interactions associated with concurrent use. In the 2002 Canadian National Health Population Survey, 21% of respondents had used both prescription medications and NHPs in the previous year and 28.4% of those were using combinations with potentially harmful interactions. 
Data collected from the 2002 United States Health and Diet Survey revealed that 4% of NHP users reported at least one adverse event over the previous year. 
In addition, those NHP users reporting an AE were more likely to be taking prescription medications and NHPs concurrently than those not reporting an AE (74% vs. 58% respectively). 
The population surveyed in this study may differ in important ways from our urban community pharmacy population which may partially explain the differing AE rates.
While pharmacy-based active surveillance improved the quantity and quality of AE reports, numerous challenges were encountered. These included challenges in finding pharmacies willing to participate and pharmacy staff members not screening as many patients as expected. The most effective strategies to counter these challenges were involvement of pharmacy staff with previous research experience, involving recent graduates and students in data collection, as well as close contact and frequent visits to the sites to ensure implementation of the study protocol. This pilot program involved pharmacies/pharmacists who volunteered to participate in NHP safety research and, therefore, may not be representative of all pharmacies/pharmacists in Canada. On the other hand, this was an entirely new concept – for pharmacists to take responsibility for adverse event reporting (and with our results, perhaps more pharmacists will be interested in NHP adverse events).
Potential sampling bias also extended to the consumers screened. Only those sufficiently healthy enough to collect their or their child’s prescriptions from the pharmacy were screened. Limiting screening to outpatient community pharmacies minimizes the ability to identify current AEs requiring hospitalization or past AEs resulting in hospitalization or even death. In addition, since the pharmacy sites screened fewer patients than was originally expected, it is possible the choice of who to screen was biased in some way. Pharmacy participants stated that patient screening depended on workload. However this was difficult to quantify: limited data were provided to us on the number of prescriptions filled by each pharmacy during the study period, but this number is likely to be much lower than the number of consumers served. Recall bias with respect to patient reports of AEs was another limitation. When possible, hospital records and laboratory data were collected to ensure the highest level of accuracy. Reassuringly, our study found rates of NHP use and NHP-drug concurrent use similar to those reported by others, suggesting that our sampling frame was acceptable. 
In this study, the AEs were identified by patients in an out-patient setting rather than by clinicians or other health professionals, which is more typical of some active surveillance designs. While this limited the data available to investigate the AE, such as obtaining drug/NHP plasma concentrations, our approach helped overcome several major obstacles in detecting NHP-related harms, namely lack of inquiry about NHP use, lack of inquiry about experiences of harms, and lack of reporting, even if harms are identified.
The community setting offered a key strength to our design: SONAR screened a comparatively large sample of patients who were using NHPs under real-world conditions. The study resulted in the development of a practical tool (the NHP-drug interaction tool), 
that has been published and has since been in use by a number of clinicians who have confirmed its usefulness. To our knowledge, SONAR is the first to systematically capture NHP-drug interactions by identifying clinically relevant harms first, coupled with basic science investigation to examine the plausible mechanism. More often in the literature, NHP-drug interactions have been posited on theoretical grounds but their clinical relevance remains unknown, as these harms have not been identified in clinical practice. 
Although follow-up interviews were difficult to schedule, the quality of data obtained allowed meaningful adjudication: all cases deemed to be “probably” NHP related, and referred for lab analysis for evaluation of NHP-drug interaction, had confirmatory lab findings. Strength of our design was the opportunity to investigate the mechanism of action of NHP AEs through laboratory research, promoting the detection of novel clinically relevant NHP-drug interactions. It is important to note that with a greater degree of AE reporting comes a greater possibility of loss to follow-up, and this is an important limitation in conducting active surveillance. Further studies should investigate the sensitivity and specificity of the adjudication process, as well as methods to improve follow-up in active surveillance of NHP-related harms.
Serious harms tend to be rare, and their detection is extremely challenging. Harms assessment is a multistage process: product use must be disclosed and discussed; product-related harms must be consistently included in the differential diagnosis; suspected harms must be reported in sufficient quantity and quality for signal detection to occur. Active surveillance offers means to promote discussion and allows for meaningful adjudication, and, therefore offers an important contribution to patient safety and pharmacovigilance. It is difficult to determine causality of an AE if potential drug interactions, product authenticity adulterants/contaminants and NHP components are not investigated simultaneously. 
The paucity of literature regarding the risks involved with NHP use is not enough reason to consider these products “safe”. 
Only systematic data collection and analysis will provide the necessary evidence to consider an NHP safe, particularly in the context of concurrent use with prescription medications.
In practice, although clinicians and patients can tolerate certain degrees of uncertainty surrounding effectiveness (i.e. a therapeutic trial or “try-and-see” approach), they are understandably less tolerant of uncertainty surrounding safety. We found 7% of those concurrently using NHPs with medications report AE, with 4/15 (26.7%; 95% CI: 4.3% to 49.0%) probably due to NHP use; whether or not this is interpreted as low or high depends on one’s perspective.
Within Canada, provincial electronic health databases exist to facilitate active surveillance by capturing adverse event information relating to prescription drugs as they occur. 
Although these databases may have the ability to capture NHP and over-the-counter drug data, this is not widely known and thus they are therefore still limited in their ability to actively capture AEs within this scope of product use. 
Our study provides a novel method for investigating harms relating to NHPs to assess causality and provide clinical evidence for future patients. Active surveillance improved detection and reporting, and the rigorous investigation of detected harms generates clinical evidence to allow for practice-based change, promoting patient safety. Additional work is needed to establish if active surveillance is cost effective. According to a Health Council of Canada publication, the annual operating costs for the Marketed Health Products Directorate was $23.6 M for the year ending March 31, 2010 and is anticipated to be $30.5 M for the year ending March 31, 2012; MHPD has confirmed that approximately 10% of their budget is spent on NHP surveillance activities. 
(See File S2
) Our research team was able to start the first NHP AE active surveillance program for approximately 1% of the MHPD NHP budget. Further research is required to determine the potential economic impact of implementing active surveillance on a wider scale.
In conclusion, active surveillance for NHP-related AEs in community pharmacies is feasible. Further investigation of its potential contributions to assessment of NHP safety is warranted. Future research includes the addition of NHP AE screening within already developed provincial electronic health databases, and the use of both active surveillance and causality assessment in high risk populations, such as those seen in hospital subspecialty clinics. There is also an opportunity to test the utility of active surveillance in other countries that regulate NHPs, as it may prove informative for both NHP and drug-related adverse events based on real-world use. The limitations of passive surveillance have been well documented, but the potential advantages of active surveillance have yet to be fully explored.