Assessing the profile and causality of adverse event related to the use of herbal medicines in patients at a public phytotherapy outpatient clinic

To assess the profile and causality of adverse events related to herbal medicines of widespread use in Brazil, we conducted an open, prospective, before-and-after clinical trial with patients at an outpatient clinic specialized in complementary and alternative medicine. Participants were submitted to laboratory tests prior to and after use of the medicines, as well as following a period of discontinuation in those experiencing adverse events. Occurrence of adverse events and their severity were verified in accordance with the WHO Toxicity Grading Scale for Determining the Severity of Adverse Events, and their causality established via the Naranjo algorithm. Forty-two subjects participated in the trial, of which 25 experienced grade 1 toxicity adverse events, mainly hyperamylasemia, and 14 continued to experience them following discontinuation, mainly hypomagnesemia. Mean/median values of laboratory tests performed in each phase were within specifications. Statistical analysis of these values in pre-treatment and treatment (n=42) and in treatment and post-treatment phases (n=22) showed statistical significance for activated partial thromboplastin time (p=0.020) in the first correlation. The Naranjo algorithm established a possible causal relationship between the use of herbal medicines and the adverse events reported. By signaling a low probability of occurrence, the algorithm indicates the safe use of these medicines under the experimental conditions employed. La ocurrencia de eventos adversos y su severidad se verificaron de acuerdo con la escala de graduación de toxicidad de la OMS para determinar la severidad de los eventos adversos, y su causalidad se estableció a través del algoritmo de Naranjo. Cuarenta y dos sujetos participaron en el ensayo, de los cuales 25 experimentaron eventos adversos de toxicidad de grado 1, principalmente hiperamilasemia, y 14 continuaron experimentándolos después de la interrupción, principalmente hipomagnesemia. Los valores medios/medianos de las pruebas de laboratorio realizadas en cada fase estuvieron dentro de las especificaciones. El análisis estadístico de estos valores en pretratamiento y tratamiento (n=42) y en las fases de tratamiento y postratamiento (n=22) mostró significación estadística para el tiempo de tromboplastina parcial activada (p=0,020) en la primera correlación. El algoritmo de Naranjo estableció una posible relación causal entre el uso de fitoterápicos y los eventos adversos informados. Al señalar una baja probabilidad de ocurrencia, el algoritmo indica el de estos medicamentos en las condiciones experimentales empleadas.


Introduction
Herbs are commonly used to cure, stabilize, prevent diseases and improve general health conditions. They are sometimes preferred over synthetic drugs, mostly for being considered a "natural" and, therefore, safer therapy. However, herbal medicines may cause harm when used irrationally (Aronson & Meyler, 2009). Adverse reactions (AR) related to herbal medicines have been reported in the literature and evidenced by pharmacovigilance systems (Ekor, 2014;Teschke & Eickhoff, 2015). Data from national drug authorities in over 110 countries, gathered by the World Health Organization's (WHO) global database for adverse drug reactions (ADR) (Vigibase®), has reported hundreds of individual cases of ADR notifications for medicinal plants (World Health Organization, 2019).
Surveillance on ADR to herbal medicines consists mainly of voluntary reports from consumers and health professionals (Jordan et al., 2010). In Brazil, between 2008 and 2012, AE notifications of products based on plant species in the System of Notifications in Health Surveillance (NOTIVISA) was less than 1% of the items notified as medicines. This reflects the fact that users do not consider these products to pose risks, and health professionals do not question their patients about the use of herbal medicines. Both attitudes contribute to reducing AE cause-effect investigations of such products (De Lima et al., 2015).
Another challenge is the large gap between what the population uses and the scientific evidence. Studies point to the need for high quality research on medicinal plants, in search of an "evidence-based phytotherapy", making health care safer (Carmona & Pereira, 2013). Randomized controlled trials are considered the best way to determine the safety and efficacy of a treatment. However, there is a scarcity of this type of study involving medicinal plants and herbal medicines, mainly due to the lack of requirement by regulatory agencies (Tachjian et al., 2010).
Most reports on ADR to herbal medicines constitute individual cases, published as single reports or case series, without proper severity and causality assessment (Teschke et al., 2013). However, the use of a standard table for interpreting and grading abnormal signs, symptoms and laboratory parameters is recommended when investigating adverse events (AEs).
The WHO Toxicity Grading Scale for Determining the Severity of Adverse Events (WHO TGS) is used to assess severity in general clinical conditions, making it possible to grade abnormal laboratory parameters as evidence of AEs and quantifying their respective severities (International Centers for Tropical Disease Research Network, 2003a;International Centers for Tropical Disease Research Network, 2003b).
The causality of an AR reflects the degree of probability of a relationship between that reaction and the drug used, and can be assessed by expert analysis, algorithmic and probabilistic methods (Jordan et al., 2010). The Naranjo Adverse Drug Reaction Probability Scale is an algorithmic scale designed for assessing AR in a variety of clinical situations, regardless of the organ affected (García-Cortés et al., 2011). It involves ten ''yes '', ''no'' or ''unknown/inapplicable'' questions, covering factors that should be considered in the assessment of any causality relationship, and the total score obtained provides the following probability categories: definite, probable, possible and doubtful (Edwards & Aronson, 2000;Naranjo et al., 1981).
To investigate the profile of AEs related to herbal medicines traditionally used by complementary and alternative medicine (CAM) in Brazil, we conducted a trial to detect the occurrence of AEs and to assess causality relationships in patients undergoing treatment with herbal medicines.

Methodology
We conducted a before/after, uncontrolled, open-label, prospective clinical trial with a single intervention group, composed of adult subjects of both genders (Estrela, 2018).

Subjects
This study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the ethics committees of the Federal University of Goiás and of the Leide das Neves Ferreira Center for Excellence in Teaching, Research and Projects, both located in Goiás state, Brazil (under protocol nos. 1.705.488 and 1.771.217, respectively). All patients gave their written informed consent and were free to withdraw from the trial at any time.
A total of 100 individuals were recruited at an outpatient clinic specialized in CAM in Goiás state from October/2016 to April/2017. Recruitment was done by a phytotherapist, from his routine of care at the Phytotherapy Outpatient Clinic. These individuals were referred by the collaborating physician for an interview with the investigator. After receiving information about the study, 89 subjects agreed to participate. Inclusion criteria were as follows: male or female, aged between 18 and 60 years, non-use of the herbal medicines under investigation in the previous three months, laboratory tests showing no abnormalities according to WHO TGS and negative results for human immunodeficiency virus (HIV), hepatitis B and C.
Patients of childbearing potential were included upon confirmation of a negative serum pregnancy test. Exclusion criteria included hypersensitivity or history of severe AR to prescribed herbal medicines, pregnancy and lactation. Fifty subjects were eligible and 42 completed the study.
Treatment interruption was recorded, and data from those participants were not considered.
During regular medical appointments, a phytotherapist prescribed these herbal medicines to patients in different combinations and dosages, according to the clinical indication of each patient. Encapsulation was performed by a pharmacy according to good manipulation practices. Formulations were supplied to participants with enough capsules for 30 days of treatment.

Study design
A before-and-after clinical trial was conducted in three phases: pre-treatment (PeT), treatment (T) and post-treatment (PoT), with T and PoT lasting 30 days each ( Figure 1). In PeT, patients were submitted to laboratory testing with a view to verifying compliance with the inclusion criteria. The researcher registered the drugs being used at that time and over the previous three months, including any self-medication and herbal drugs, as well as ADR history.    Participants were instructed to report, during T and PoT, the use of any concomitant medication, whether prescribed or over-the-counter, including all kinds of natural products, food or dietary supplements.
Serological tests for HIV, hepatitis B and C viruses were performed only in PeT, as well as those for human chorionic gonadotropin (hCG) β-subunit in female patients of childbearing potential.

AE severity assessment
Laboratory tests during T were analyzed according to WHO TGS. Results that conformed to this scale were considered AEs and were classified in four grades of toxicity, from grade 1, the mildest, to grade 4, the most severe, as shown in   found in that database, searches were conducted on MEDLINE/Pubmed and linked the name of the plant species (scientific and common) and the AE. For AEs related to liver damage, the LiverTox ® website was also consulted. Medical history and alcohol consumption reported by participants in PeT were also considered.
As regards objective confirmation of AEs (question 10), laboratory findings indicating toxicity (in line with WHO TGS) were viewed as an objective piece of evidence of AE occurrence.
Placebo administration, increased or decreased doses of the herbal medicines under investigation and analysis of their plasma concentration were not evaluated in this study. Questions 6, 7 and 8 were marked "unknown/inapplicable".

Statistical analysis
Frequency was established for patients' characteristics (gender and health complaints). Age was expressed as mean and standard deviation (SD).
Laboratory test results were expressed as mean and SD for parametric data or median and interquartile range for nonparametric data, considering all the individuals enrolled in each study phase. Two-tailed t-test (parametric data) and the Kolmogorov-Smirnov test (nonparametric data) were used to correlate the means/medians of these results between PeT and T and later between T and PoT. For laboratory test results expressed qualitatively (proteinuria and hematuria), McNemar pairedsamples test was used.
A p-value < 0.05 was considered to indicate statistical significance. Statistical analyses were performed via Système Portable d'Analyse -SPAD (version 7.4) and free software R (version 3.6.2 for Windows).

Subjects' characteristics
The socio-demographic characteristics of the 42 selected participants are presented in Table 2. The median age of the 42 participants was 46.1 years (SD + 10.2), and 92.9% were female subjects.

Safety evaluation
Mean/median values of each laboratory test performed during PeT and T, considering all participants selected (n=42), are shown in Table 3. Statistical significance was established for activated partial thromboplastin time (p=0.020). Nevertheless, the totality of results is within the range considered favorable by WHO TGS (Table 1).
Mean/median values of each laboratory parameter evaluated in T and PoT for the 22 patients that completed PoT are shown in Table 3. No unfavorable values (in accordance with WHO TGS - Table 1) were observed, and there was no statistically significant difference when comparing these values in both phases.
There was no statistical significance between PeT and T for proteinuria and hematuria (p=1 and p=0.07364, respectively). The same was reported when assessing these parameters during T and PoT (p=1 for both tests).  Table 3 presents the mean/median values of each laboratory parameter evaluated in PeT, T and PoT phases, and according to the values defined by WHO TGS, none of them was considered unfavorable. It also brings results of the statistical correlation of the mean/median values of laboratory parameter between PeT and T, and later between T and PoT, demonstrating a statistically significant difference for activated partial thromboplastin time (p=0.020) between PeT and T. thrombocytopenia; prothrombin time elevation; blood urea increase and proteinuria (n=1; 2.6%, each). The majority of participants experienced one AE (n=17), while some experienced two (n=4), three (n=3) and four (n=1) AEs.

AE causality assessment
Causality assessment was conducted in each of the 35 AEs observed during T in the patients who completed all phases of the study (n=22). From the 25 subjects who experienced AE during T, 22 repeated the laboratory tests and concluded PoT. The other 3 individuals dropped out of the study at this stage, alleging unavailability to come to the laboratory for the tests. The loss of these 3 patients did not impact the study, since the causality assessment was performed from the individual's perspective, considering the study design (before/after).
The Naranjo algorithm was selected to assess causality because it is a simple and widely applicable method that has been used to evaluate AE causality in several clinical situations, including the use of medicinal plants (Bilgi et al., 2010;Necyk et al., 2014). In addition, the application of the Naranjo scale reinforced the individual approach adopted in the study's experimental design, since each AE associated with each participant was investigated individually.   Table 4 presents the causality assessment of the evidenced AEs, as well as the aspects considered in this assessment according to the Naranjo algorithm. The laboratory alterations indicative of toxicity were all characterized as possible AEs to the use of the investigated herbal medicines.
When assessing causality, it is imperative to gauge whether an AE fits the pharmacological pattern of the substances under investigation (Edwards & Aronson, 2000). Although we have conducted an extensive and systematic search of the literature, no evidence was found relating the herbal medicines used by participants and the AEs experienced by them. In fact, some studies described opposite effects to what we observed. Participants 1B.4 and 1A.14, who respectively experienced hyperamylasemia and hematuria, used C. longa, which has been reported as effective in reversing serum amylase elevation in mice with acute pancreatitis (Yu et al., 2011) and in significantly reducing hematuria in a clinical trial with lupus nephritis patients (Khajehdehi et al., 2012).
Another aspect of major importance involves considering the temporal relationship between exposure to a substance and the emergence of an AE, as well as behavior following withdrawal, and analyzing objective parameters of organ functions both before (baseline) and after substance use (Edwards & Aronson, 2000). This study was designed as a before-and-after clinical trial, followed by a withdrawal phase, allowing the establishment of baseline laboratory parameters and the evaluation of subsequent alterations related to the use of herbal medicines for 30 days of treatment, and the same suspension interval. The decision to set the 30-day period for T was based on an observational study that investigated liver injury among users of herbal medicines, and clinical safety studies of phytotherapeutic medicines (Jeong et al., 2012;Nascimento et al., 2009;Tavares et al., 2006). All AEs submitted to the Naranjo algorithm were objectively confirmed through laboratory test alterations, which yielded a temporal relationship with the use of the herbal medicines prescribed.
Diagnosing AEs must also take into account any other drugs or natural products consumed by subjects as possible causal agents. For that, we investigated the role of each product consumed by participants during T, which are described in Table 4, as an alternative cause to the AE in question. As regards hyperamylasemia, participant 1B.3 made use of estradiol, calcium and bupropion, described as capable of causing elevation in amylase seric concentration and pancreatitis (Bupropion, 2018;Vege, 2018). Pancreatitis is also an ADR to venlafaxine, enalapril, sodium diclofenac and cyproterone, used respectively by subjects 1B.5, 1A.1, 1A.3 and 1A.4 (Cyproterone, 2018;Diclofenac (systemic), 2018;Enalapril, 2018;Venlafaxine, 2018).
In addition to these drugs, patient 1A.9 reported the use of hibiscus tea. Fakeye et al. (2009) have shown significant ALT elevation in rats that received an aqueous extract of Hibiscus sabdariffa Linn.
Besides gathering information and analyzing each product consumed as a potential alternative cause to AEs, it is important to determine how long these substances were being used. The majority are medications used continuously for more than six months prior to the study, which corroborates the causality attributed to the herbal medicines, given that they primarily represent the only difference between PeT and T.
In addition to the products consumed by participants during T, other factors may yield alternative causes to AEs. Participant 1A.6 experienced hypercalcemia during T, with a calcium concentration (corrected for albumin) equal to 10.8 mg/dL. Calcium and magnesium compete with each other for carriers in the Henle loop, and hypercalcemia may lead to a reduction in magnesium concentration (Yu, 2018); hence, it may cause the hypomagnesemia detected in that patient. Female subject 1A.13 reported being in the end of the menstrual period when T-phase laboratory tests were performed, which corresponds to an alternative cause to the transient hematuria registered. Alcohol consumption, in turn, was not considered an alternative cause since participants reported the use of acceptable quantities, in line with Friedman (2019); these are lower than 210 g and 140 g of alcohol for men and women per week, respectively, for a minimum of two years.

Conclusion
Total scores yielded by the Naranjo scale show that all AEs registered are possible ADRs caused by the investigated herbal medicines at dosages and combinations listed. A "possible" relationship corresponds to a low probability of occurrence of an AE by a product under investigation, under the present experimental conditions.
Results obtained with the individual approach proposed by the Naranjo algorithm were corroborated by the global analysis of participants. Comparisons of laboratory parameters evaluated prior to, during use, and after discontinuation of the consumption of herbal medicines failed to show statistical significance. In fact, this monitoring of toxicity and AEs profile through laboratory testing is one of the strengths of this work.
Moreover, the results showed that using WHO TGS in the detection of AEs and the Naranjo algorithm in causality assessment is a useful and effective strategy to optimize the benefits and minimize the risks of this therapy in clinical practice, thus contributing to the rational use of herbal medicines.
As perspectives for future research, it is recommended to conduct controlled, double-blind, randomized clinical trials, using herbal medicines detected as "possible" causes of AE, in order to deepen the investigation of the data obtained.