Adverse Event (AE) Reporting Practices in Prior Anti-osteoarthritis (OA) Medication Trial Publications
The articles selected for the ESCEO meta-analyses on the safety of some anti-OA medications [30,31,32] hide important divergence in the reporting of AEs, which we report here. As previously stated, these meta-analyses were aimed at assessing the AEs associated with the use of topical NSAIDs, SYSADOAs and IAHA for the management of OA. The main different ways in which harms-related data were reported in the retrieved articles are summarized in the following six subsections and in Table 1. These diverse reporting practices, which have also been described in other systematic reviews on drugs for other diseases [15, 16], preclude any objective comparison of AEs and AE frequencies between various studies on the same medication, as well as between studies on different drugs for the same disease. Pooling of these kind of diversely reported data is inadequate in meta-analyses  or, at least, might lead to a systematic underestimation of harms. Thus, any objective assessment of harm/benefit balance in such context is compromised.
In addition to the main harms-reporting practices summarized hereunder, we also found that, although the total numbers of participants who withdrew from the trials due to AEs or who experienced any severe or serious AEs at least once during the trials were usually reported in the manuscripts, the specific events pertaining to each of these categories were not always reported. In almost all of the articles, we found no reported data about drug effects on biochemical and biological parameters such as full blood count, hemoglobin, and biochemical markers of liver injury and renal function.
Treatment-Emergent vs. Treatment-Related AEs
In some articles, harms were reported as “treatment-emergent AEs” (TEAEs) [36,37,38], defined by the International Conference on Harmonization (ICH) as events “that emerge during treatment having been absent pre-treatment, or that worsen relative to the pre-treatment state” . Other manuscripts reported only AEs judged by the investigators as “treatment-related AEs” (TRAEs) [40,41,42]. Finally, many other cases did not clearly specify whether the reported AEs were “treatment emergent” or “treatment related” [43, 44].
Relevance of AEs: A Threshold for AE Reporting
Reporting AEs according to frequency of occurrence was very common. Many articles reported only AEs experienced by at least a given percentage of patients in either group (treatment or control). The most common threshold percentages were 2% (i.e., AEs occurring in ≥ 2% of patients in either group) [36, 45], 3% , or 5% [46,47,48].
System Organ Class (SOC) vs. Specific AE Reporting
Many articles reported AEs by SOC, providing for each SOC the number of patients who experienced any AE related to that SOC. Sometimes only partial or no details on specific AEs pertaining to each SOC were included [42, 49, 50]. By contrast, other articles reported only a small number of selected specific AEs [47, 51] without providing the overall SOC-related frequencies when many specific AEs pertained to the same SOC.
Reporting in Summary Form and/or Reporting of the Number of AEs
The majority of manuscripts on the efficacy and safety of anti-OA medications reported AE data only in summary form. This summary was usually formulated as: “The most commonly reported AEs in the intervention and control groups were xxx” and/or “There was no statistically significant difference between the intervention and control groups” and/or “There were no serious AEs in either group” [52,53,54].
Alternatively, rather than listing the AEs and providing the number of patients who experienced each event at least once (the frequency), some manuscript authors reported only the numbers of AEs experienced during the trial [52, 55, 56]. This was usually stated in the form of “The total numbers of AEs in the intervention and placebo groups were # and ## AEs, respectively” or “The numbers of AEs in the two treatment arms were similar: ### AEs occurred in the intervention group and #### AEs in the placebo group.” This kind of information is not relevant in terms of knowledge about harms associated with a given treatment.
Combination of Many of the Previous Reporting Profiles in a Single Manuscript
Very often, many of these previously described reporting practices were combined in a single manuscript. Sometimes, “TRAEs in a defined percentage of patients” [42, 45, 57] or “TEAEs in a defined percentage of patients” [36,37,38] were reported, and these threshold percentages varied between manuscripts reporting data from various RCTs on the same drug. Other cases provided a narrative summary of treatment-related (and/or treatment-unrelated) AEs in the results section of the manuscript to describe harms associated with drugs [40, 58].
AE Coding Systems
Different coding systems were used for the assessment and reporting of AEs in the studies included in the new ESCEO meta-analyses. These included the WHO-Adverse Reaction Terminology (WHO-ART) system [55, 58], the Medical Dictionary for Regulatory Activities (MedDRA) [38, 59, 60] and the Coding Symbols for Thesaurus of Adverse Reaction Terms (COSTART) [44, 61]. Most manuscripts did not specify the AE coding dictionary used [43, 62,63,64].
The ESCEO Recommendations for Harms Reporting in Future Manuscripts on Studies Assessing Anti-OA Drugs
In total, 20 individuals were first invited to the modified Delphi process (round 1); 15 responded, representing 75% of the original experts contacted. Based on comments from these experts during this first round, the six items originally designed as a general framework for harms reporting were rephrased or improved, leading to a new draft with eight items. Only the 15 experts who completed the first round were then invited to take part in round 2 of the modified Delphi process; all of these participants (100%) rated each recommendation on a scale of 0–10, where 10 represented the highest level of agreement on the item.
For the eight items of the general framework for harms reporting, the median rating of the working group members was 9 (P25 ≥ 7; P75 = 10) for four items, 8 (P25 ≥ 6; P75 ≥ 9) for three items, and 7 (P25 = 6; P75 = 8) for one item. This suggests a high level of agreement on each item. As planned, the working group members who rated an item as < 7 were contacted and asked to comment on their rating. Based on these comments, the item was rephrased for a better consensus on the final version of the recommendation without changing its main theme.
Regarding the SOCs to be considered for AE reporting within the results section of manuscripts, the votes of the members of the working group were counted. For each drug class/group, the SOCs that received ≥ 50% of the votes (≥ 7/14 voters) were retrieved. Some additional SOCs with six votes were also ultimately included by consensus. For the SYSADOAs, “eye disorders” was not originally selected (only three votes initially) but was ultimately added to the SOCs list by consensus. Indeed, a recent RCT on glucosamine reported an increase in intraocular pressure in patients with OA, particularly in older patients .
General Framework for Harms Reporting Within the Results Sections of Manuscripts on Anti-OA Drugs
The members of the ESCEO working group on the safety of anti-OA medications agreed on a general framework of recommendations for better reporting of harms in studies assessing drugs for OA. This is summarized in Table 2 and discussed in detail here.
The ESCEO working group first focused on the types of AEs to be reported within the results section of manuscripts and how these should be reported to allow for capturing of all relevant harms. While it is understandable that AEs may differ according to individual drugs and pharmacological classes, reporting the same type of information, in the same way, from various trials of the same drug or of drugs pertaining to the same class appears to be essential. The CONSORT-harms extension statement recommends to “present the absolute risk per arm and per AE type, grade, and seriousness” . However, the “AE type” is not clarified, with the result that manuscript authors may not report the same type of information for different trials on the same drug, and even for trials on different drugs for the same disease. To allow for consistency in the future, and for comparability, the ESCEO proposes a general framework for harms reporting consisting of always reporting AEs by body system.
The other source of variation in the reporting of AEs in clinical trial manuscripts is the coding dictionary used. For the meta-analyses commissioned by the ESCEO to assess the safety profile of various anti-OA medications, almost all of the full safety reports received from pharmaceutical companies and investigators used the MedDRA coding system. Other authors have also reported that MedDRA was one of the most commonly utilized systems for harms reporting around the world [66, 67]. Thus, for consistency in this regard, the ESCEO recommends use of the MedDRA coding system (https://www.meddra.org/) when reporting harms from drugs for OA [68, 69]. However, if another coding system is used (which is not recommended), the ESCEO recommends that this coding system and the reason for not using MedDRA should always be clearly stated in the manuscript. Regardless of the coding system used, the ESCEO recommends that, for each SOC, at least the five most frequent (more if necessary) specific AEs with their frequencies (i.e., the number of participants per arm who experienced each specific AE at least once) are reported. These AEs should be reported whether or not the difference between the intervention and control groups is statistically significant. When the MedDRA dictionary is used for AE reporting, as recommended by the ESCEO, specific AEs should be reported using MedDRA “preferred terms” [68, 69].
Finally, regarding the types of AEs that should be reported, the ESCEO recommends always reporting the total number of patients who experienced at least once any severe AE (highest grade in severity—mild, moderate, severe—for the non-serious AEs) and any serious AE (an AE that results in death, is life-threatening, requires inpatient hospitalization or prolongation of existing hospitalization, results in persistent or significant disability/incapacity or is a congenital anomaly/birth defect) . As well, the total number of patients who withdrew from the trial due to AE(s) and the overall number of patients who experienced any AE (total AEs) at least once during the trial should always be reported, irrespective of the drug. These reports should be made regardless of the SOC. As for the specific AEs for reporting by SOC, at least the top five specific events that led patients to withdraw from the trial should be reported, as well as the top five total AEs. By contrast, ESCEO safety working group members emphasized that all specific severe and serious AEs should always be reported.
The ESCEO safety working group members are aware that the recommendation to report the most frequent specific events (in the intervention group) pertaining to an SOC might pose a different problem. Indeed, the most frequent events might not be clinically relevant, and some more clinically relevant events might be sparse. To avoid missing any clinically relevant information, the working group members recommend separately reporting any other specific event that would be judged as clinically relevant by the clinicians conducting a trial and that would not have been captured by the reporting of serious or severe AEs, or withdrawals due to AE(s), as recommended hereunder. However, as has been seen in studies on diacerein for patients with OA, the most well-known safety issues related to that compound were the most frequent events reported in the full clinical trial reports, particularly regarding gastrointestinal and urinary system disorders [48, 70]. On the other hand, even though the most frequent specific events pertaining to an SOC are not clinically relevant, knowing them may also be important. In fact, this could help physicians to warn patients that they might experience such events (in case of a causal relationship with the treatment) but that they are not severe or serious.
The approach that consists of reporting AEs by body system along with frequencies for specific events pertaining to each body system is supported by the following rationale: when only specific AEs are reported with multiple events pertaining to the same organ class (e.g., vomiting, nausea, diarrhea, abdominal pain), it is impossible to know whether many of them have been experienced by a single patient. In this case, a meta-analysis on the harms related to a specific SOC (e.g., gastrointestinal) will not be feasible or might lead to double counting. Furthermore, for a given drug, when authors of different clinical trials choose to report only the frequencies for specific AEs (and not the SOC level frequencies), each of these authors may not be interested in reporting the same specific events. In such a case, comparing results from various studies might not be easy, because the same events (of interest) might not be reported by different authors.
The second main issue addressed by the ESCEO working group on the safety of anti-OA medications was the thresholds usually chosen by authors to report harms-related data and the choice to report TEAEs or TRAEs. On these points, the working group members recommend that harms should always be reported considering all TEAEs. This excludes any practice consisting of choosing a threshold of occurrence for AEs reporting and/or reporting only the AEs judged as “treatment-related” by clinical investigators, i.e., only based on the judgement of clinicians. As noted by Lineberry et al.  in their paper on recommendations to improve AE reporting in clinical trial publications, such judgement by clinicians is subjective and not adequate in the context of randomized, double-blind, controlled clinical trials. Notably, the ESCEO experts recommend that these claims of “treatment-related” or “treatment-unrelated” AEs be avoided, particularly for new drugs for which a “definitive” conclusion regarding the safety profile requires results from several trials, considering Hill’s causal criteria [71, 72]. For all drugs (old and new), the working group members recommend that the interpretation of a relationship between an AE and the study drug should only be made in the discussion section of the manuscript, comparing the results obtained from the RCT with those reported from other trials and meta-analyses [71, 73]. This applies also for suspected unexpected serious adverse reactions (SUSARs) , which regulators compel clinical trial sponsors to report through expedited reporting during the course of clinical trials [27,28,29]. In fact, reporting SUSARs implies an assessment of causality (based on clinicians’ judgments), which the ESCEO discourages from the results sections of manuscripts. Since SUSARs are by default “serious” in nature, they should be reported along with all other serious AEs in the results section as recommended, without any distinction. However, if a serious event qualified as a SUSAR occurred in the course of an RCT, it should be mentioned in the manuscript conclusions (both the conclusion of the main text and that of the summary).
It should be noted that it is extremely difficult (even impossible) to draw a definitive conclusion regarding the safety of drugs in the context of a single clinical trial. Indeed, RCTs are not usually powered to detect differences in AE rates between the intervention and placebo groups but to test hypotheses on efficacy outcomes [16, 75,76,77]. In such context, a real statistically significant difference can exist but might not be detectable [75, 76]. That is why it would not be adequate to base any conclusion on the results of statistical comparison of harms data between the intervention and control groups [15, 77,78,79,80]. Furthermore, a statistically significant or non-significant difference might merely be due to sampling error, a direct result of having drawn a sample from a population [81,82,83]. On the other hand, the risk of bias in RCTs (systematic errors) may more severely affect the results [82, 84,85,86]. Hence, results from several trials and meta-analyses are required to adequately estimate the risk of AEs [15, 77, 80, 87] and to “definitively” conclude on the safety of drugs. What would be appropriate in the context of a single clinical trial, and strongly recommended by the ESCEO, particularly for serious AEs (including SUSARs), is that the characteristics of each patient reporting any serious AE (demographic characteristics, clinical characteristics, concomitant drug use, comorbidities, etc.) and the context in which the event occurred are thoroughly assessed. This information should be reported in the discussion section of the manuscript to support the interpretation on relatedness of the event with the studied medication, for each single patient involved. Such an interpretation should always be cautious, reporting in terms of probability of relatedness (strong, weak or uncertain), rather than making strong statements such as “treatment related,” “treatment-unrelated” or “treatment is safe.”
Finally, the ESCEO working group members reached consensus on the recommendation that clinical trial manuscripts should also include information on drug effect on biological and biochemical parameters for both approved and experimental drugs in OA. The relevant biological parameters may differ according to the drug, depending on the drug mechanism of action and the potential for any other adverse biological effects. However, for comparability between drugs and across studies, the ESCEO safety working group recommends that information on drug effects on liver and renal function parameters, as well as on blood cells and hemoglobin, should always be reported whatever the drug, and provide specific guidance on how this information should be reported (see item 7 in Table 2). Regarding the relevance of biological parameters according to drug type, reporting data on glycemia may be important for the SYSADOAs for example, particularly for glucosamine and chondroitin. In the case of biological drugs, it would also be important to collect and report data on antibody immune responses. Data on these particular biological parameters, which are specific to drug types, should be reported on the same model as for the biological parameters that should always be reported irrespective of the drug.
All safety information should be reported in a table that should be included in the results section of the manuscript (see Table 2 and the Electronic Supplementary Material [ESM]-1).
Recommendations on the Main Body Systems to be Considered for Harms Reporting Within the Results Section of a Manuscript
Table 3 lists the MedDRA SOCs  that the ESCEO recommends should always be considered for harms reporting within the results section of a manuscript on anti-OA medications, according to each drug class.
It is important to note that limiting the reporting of body system-related AEs to these SOCs will not result in a loss of information regarding the relevant AEs that would have occurred in the other SOCs. In fact, recommendation number 6 (other clinically relevant AEs) (Table 2) should help capture any other clinically relevant AE that would not have been captured with all the other recommendations. The approach recommended in Table 3 allows for consistency in reporting across several manuscripts, while also considering the space limitations imposed by journal requirements.