검색
검색 팝업 닫기

Ex) Article Title, Author, Keywords

Article

Split Viewer

Original Research Article

DTT 2023; 2(1): 41-48

Published online March 31, 2023 https://doi.org/10.58502/DTT.23.0003

Copyright © The Pharmaceutical Society of Korea.

Adverse Events of GLP-1 Receptor Agonists for Weight Loss: Twitter and a National Pharmacovigilance Database

Hyun-Myung Back1* , Seo-A Choi1* , Myeong Gyu Kim1,2

1College of Pharmacy, Ewha Womans University, Seoul, Korea
2Graduate School of Pharmaceutical Sciences, Ewha Womans University, Seoul, Korea

Correspondence to:Myeong Gyu Kim, kimmg@ewha.ac.kr
*These authors contributed equally to this study.

Received: January 26, 2023; Revised: March 3, 2023; Accepted: March 3, 2023

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ((http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Glucagon-like peptide-1 (GLP-1) receptor agonists became the most prescribed drugs for weight loss. This study aimed to investigate adverse events of GLP-1 receptor agonists using Twitter and KIDS-KAERS database (KIDS-KD). 116 Korean tweets on Twitter and 1,701 liraglutide-adverse event combinations from KIDS-KD were used for analysis. The most common adverse events were nausea (37.76%), vomiting (12.59%), dyspepsia (9.79%), dizziness (4.90%), and somnolence (4.90%) on Twitter. Adverse events observed in the post-marketing surveillance such as somnolence, myalgia, chest pain/ discomfort, and parosmia were found on Twitter. In KIDS-KD, injection site rash (13.46%), injection site pruritus (10.76%), nausea (10.64%), vomiting (6.41%), and injection site reaction (5.29%) were commonly reported. Signals were injection site rash (reporting odds ratio [ROR] = 616.1), injection site pruritus (ROR = 477.7), vomiting (ROR = 2.4), injection site reaction (ROR = 222.0), pruritus (ROR = 2.5), injection site bruising (ROR = 45.8), injection site pain (ROR = 36.3). Subgroup analysis revealed that injection site rash (n = 16, ROR = 46.6) and acne (n = 3, ROR = 39.2) met signal criteria in men and women, respectively. Twitter analysis identified adverse events that were not found in clinical trials, and KIDS-KD subgroup analysis suggested the possibility of acne being a safety signal. Social media and KIDS-KD were useful for pharmacovigilance.

KeywordsGLP-1 receptor agonists, KIDS-KAERS database (KIDS-KD), pharmacovigilance, adverse events, Twitter

Glucagon-like peptide-1 (GLP-1) is an incretin peptide secreted from intestinal endocrine cells and enhances glucose-dependent insulin secretion from pancreatic β-cells (Hinnen 2017; Mosavat et al. 2020; Ja'arah et al. 2021). For this reason, GLP-1 receptor agonists were approved as antidiabetic drugs. GLP-1 receptor agonists also support weight loss through delayed gastric emptying, reduced fat absorption, and reduced food intake (Baggio and Drucker 2021). Liraglutide, a GLP-1 receptor agonist, was previously approved for the treatment of type 2 diabetes under the brand name Victoza®. Saxenda®, another brand name for liraglutide, was approved by the U.S. Food and Drug Administration in December 2014 for the treatment of obesity (Nuffer and Trujillo 2015). Semaglutide was also approved for diabetes treatment (Ozempic® and Rybelsus®) in December 2017 and weight loss (Wegovy®) in June 2021 (Singh et al. 2022).

While appetite suppressants acting on the central nervous system are addictive and have side effects such as paresthesia, depression, insomnia, decreased concentration, and hallucinations (Bray 2014), GLP-1 receptor agonists are relatively safe and thus in the spotlight as a novel weight loss drug. The most common adverse events of GLP-1 receptor agonists include nausea, diarrhea, constipation, vomiting, injection site reactions, headache, hypoglycemia, and dyspepsia (Astrup et al. 2009; Pi-Sunyer et al. 2015; Rubino et al. 2021; Wilding et al. 2021).

Real-world data can have different implications from the controlled environment of clinical trials. People who are not overweight or obese also administer drugs to lose weight. In addition, long-term or rare adverse events that were not observed during clinical trials can be observed in real-world settings. The use of social media has grown exponentially over the past few years, which has become a major source of information. With the prosperity of social networks, people have become more inclined to post prescription drugs and related adverse events and share their treatment experiences (Harpaz et al. 2012; Harpaz et al. 2014). By sharing thoughts and experiences with a vast number of users, social media such as Twitter has become a useful resource for post-marketing surveillance of drugs, complementing existing pharmacovigilance systems (Sarker et al. 2015; Farooq et al. 2021). Pharmacovigilance through social media can detect adverse events before the regulatory agency’s action (Lee et al. 2021). Furthermore, real-world studies in South Korea mainly use claim data from the Health Insurance Review & Assessment Service, but there is a limitation in that non-reimbursable drugs are excluded. Since GLP-1 receptor agonists used for weight loss are non-reimbursable drugs, studies using social media are appropriate.

Spontaneous adverse event reports are another real-world data specialized for adverse events. The Korea Institute of Drug Safety & Risk Management (Ministry of Food and Drug Safety) provides Korea Adverse Event Reporting System (KAERS) database (KIDS-KD) for pharmacovigilance analysis (Shin et al. 2021). Several studies have shown that signals of adverse events were detected by data mining on KIDS-KD.

This study aimed to investigate adverse events of GLP-1 receptor agonists (Saxenda® and Wegovy®) used for weight loss in South Korea using Twitter and KIDS-KD.

Tweet collection

Tweets which were posted from August 2017 (Saxenda® was approved in South Korea on July 21, 2017) to June 2022 were collected with restrictions in the Korean language using the following search query: ‘삭센다’ (Saxenda in Korean) OR ‘삭샌다’ (Saxenda in Korean) OR ‘위고비’ (Wegovy in Korean) OR ‘리라글루타이드’ (liraglutide in Korean) OR ‘리라글루티드’ (liraglutide in Korean) OR ‘세마글루타이드’ (semaglutide in Korean) OR ‘세마글루티드’ (semaglutide in Korean). Although Wegovy® is not currently marketed in South Korea, it is included in the search because it may be mentioned on social media. Tweet search was conducted on July 6, 2022, using Python version 3.7 (Python Software Foundation, Fredericksburg, VA, USA) and Twitter application programming interface for academic research. The collected data included user identifier, tweet text, date posted, and the number of retweets, replies, likes, and quote tweets. The study was exempted from Institutional Review Board review (ewha-202203-0027-01).

Tweet annotation

All tweets were manually labeled. First, tweets that were unrelated to the GLP-1 receptor agonists were excluded. Second, tweets were classified as tweets mentioning adverse events or not. Tweets mentioning adverse events were further classified as first-hand experience or non-first-hand experience. Non-first-hand experience included adverse events from news, study results, or someone else’s experience. The adverse events mentioned in the tweet were coded as Medical Dictionary for Regulatory Activities (MedDRA) Preferred Terms version 25.0.

Tweet analysis

The annual frequency of tweets mentioning adverse events, the number of adverse events per tweet, and the number of adverse event tweets per user were analyzed. The number of adverse events was analyzed by MedDRA Preferred Terms. Drug labels in South Korea (Ministry of Food and Drug Safety 2022) and United States (U.S. Food and Drug Administration 2022), and Micromedex® (www.micromedexsolutions.com) were reviewed to identify adverse events that have been listed.

KIDS-KD data collection and processing

From August 2017 to December 2020, adverse event reports including liraglutide or other antiobesity drugs (the Anatomical Therapeutic Chemical [ATC] Classification System code A08) were provided by KIDS. The data consisted of eight tables: basic information (ADR_REPORT_BASIC), drug information (DRUG_INFO_ADR), adverse drug reaction information (ADR_INFO_REPORT), seriousness of AEs (SERIOUS_ADR), reporter information (REPORTOR_ADR), causality assessment (ASSESSMENT_ADR), medical history (DISEASE_HIST_ADR), and group information (GROUP).

Only initial reports were used for analysis. By reviewing drug information and medical history, reports with a history of diabetes or antidiabetic drugs were removed to exclude cases where liraglutide was used for diabetes treatment. If there was no record of medical history or concurrent medications, it was difficult to determine whether the data was truly missing, so they were included in the analysis. From the causality assessment table, only reports that liraglutide or other antiobesity drugs were assessed as 'certain', 'probable', and 'possible' remained for the disproportionality analysis. If information for causality assessment was missing, it was excluded from the subsequent analysis.

Disproportionality analysis of KIDS-KD

A 2-by-2 table was constructed from the drug and adverse event combination (Table 1). From the table, proportional reporting ratio (PRR), reporting odds ratio (ROR), and information component (IC) were calculated as Table 2 (Choi et al. 2021). Signals that met the three criteria of PRR, ROR, and IC were identified.

Table 1 2 × 2 table for disproportionality analysis

Specific adverse eventOther adverse events
LiraglutideAB
Other antiobesity drugsCD

Table 2 Definition and criteria of signal detection for each indicator

DefinitionCriteria of signal detection
PRR(A/(A+B))/ (C/(C+D))PRR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ROR(A/B)/ (C/D)ROR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ICLog2 (P(AE, drug)/ P(AE) × P(drug))Under limit of 95% confidence interval ≥ 0

IC, information component; PRR, proportional reporting ratio; ROR, reporting odds ratio.


Subgroup analysis was conducted according to sex. Based on sex information recorded in the basic information table (PTNT_SEX = 1 or 2), PRR, ROR, and IC values were calculated by sex. Cases with missing sex information were excluded from the subgroup analysis.

Data processing and statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary NC). Drug labels in South Korea (Ministry of Food and Drug Safety 2022) and United States (U.S. Food and Drug Administration 2022), and Micromedex® were reviewed to identify adverse events that have been listed.

Characteristics of tweets

Fig. 1 shows a tweet processing diagram. A total of 2,499 Korean tweets were collected. Of 2,499 tweets, 235 tweets mentioned adverse events. Finally, 116 first-hand experience tweets from 73 users were used for analysis.

Figure 1.Tweet processing diagram.

The annual frequency of tweets is shown in Fig. 2. An average of 20 tweets per year from 2018 to 2021 mentioned adverse events. In 2022, 36 tweets mentioned adverse events by the first half. Most tweets contained one (n = 94, 81.0%) or two (n = 18, 15.5%) adverse events in their text. Three tweets (2.6%) contained three adverse events and one tweet (0.9%) contained four adverse events. Most users posted one (n = 54, 74.0%) or two (n = 10, 13.7%) tweets mentioning adverse events.

Figure 2.Annual frequency of tweets. A solid line indicates the number of adverse event tweets which were first-hand experience. A dotted line indicates percentage of adverse event tweets to total tweets mentioning GLP-1 receptor agonists.

Adverse events from tweets

Table 3 shows MedDRA Preferred Terms sorted in the order frequently mentioned on tweets. Of 143 adverse events mentioned on tweets, two were not classified as Preferred Terms because they used a broader term. The top five adverse events were nausea (n = 54, 37.76%), vomiting (n = 18, 12.59%), dyspepsia (n = 14, 9.79%), dizziness (n = 7, 4.90%), and somnolence (n = 7, 4.90%).

Table 3 Adverse events of GLP-1 receptor agonists mentioned on Twitter

Preferred termsNumber of mentioning (n = 143)aNotoriety
Nausea54 (37.76)K/U/M
Vomiting18 (12.59)K/U/M
Dyspepsia14 (9.79)K/U/M
Dizziness7 (4.90)K/U/M
Somnolence7 (4.90)Kb/Uc
Injection site reaction5 (3.50)K/U/M
Diarrhoea3 (2.10)K/U/M
Myalgia2 (1.40)Kb
Abdominal pain2 (1.40)K/U/M
Insomnia2 (1.40)K/U
Headache2 (1.40)K/U/M
Fatigue2 (1.40)K/U/M
Hypoglycaemia2 (1.40)K/U/M
Flatulence2 (1.40)K/U/M
Injection site bruising2 (1.40)I
Injection site rash2 (1.40)I
Chest pain1 (0.70)Kb
Chest discomfort1 (0.70)Kb
Urticaria1 (0.70)K/U/M
Asthenia1 (0.70)K/U/M
Increased appetite1 (0.70)No
Urine abnormality1 (0.70)No
Vision blurred1 (0.70)Uc
Allergy1 (0.70)K/U
Rhinitis1 (0.70)No
Dermatitis acneiform1 (0.70)No
Depression1 (0.70)U/M
Parosmia1 (0.70)Kb
Gastrooesophageal reflux disease1 (0.70)K/U/M
Injection site pruritus1 (0.70)I
Disturbance in attention1 (0.70)No
Unclassifiable2 (1.40).

I, included in injection site reaction; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); U, listed on the drug labels (Saxenda® and Wegovy®) in United States.

aValues are presented as n (%).

bReported from post-marketing surveillance in South Korea.

cSuggested as signs and symptoms of hypoglycemia.


Adverse events such as somnolence, myalgia, chest pain, chest discomfort, and parosmia identified in tweets are only recorded in the Korean label. This is because they were not originally reported in clinical trials for approval, but rather reported in the post-marketing surveillance conducted in Korea over four years with 670 subjects.

Characteristics of KIDS-KD

Fig. 3 shows a KIDS-KD processing diagram. From August 2017 to December 2020, there were 3,281 reports for liraglutide or antiobesity drugs in KIDS-KD. A total of 1,701 liraglutide- adverse event combinations and 1,976 other antiobesity drugs-adverse event combinations were used for disproportionality analysis. Table 4 shows the characteristics of adverse event reports on liraglutide and other antiobesity drugs. Adverse events were primarily reported by women aged 35 to 55, and were reported by the patients through the pharmaceutical company.

Table 4 Characteristics of adverse events reports on liraglutide and other antiobesity drugs

CharacteristicsLiraglutide (n = 989)Other antiobesity drugs (n = 1,197)
Age, median (IQR), years45 (35-55)42 (33-52)
Sex, n (%)
Men113 (11.4)140 (11.7)
Women723 (73.1)973 (81.3)
Not reported153 (15.5)84 (7.0)
Reporting year, n (%)
20172 (0.2)157 (15.9)
2018199 (20.1)304 (30.7)
2019401 (40.6)428 (43.3)
2020387 (39.1)308 (31.1)
Original reporter, n (%)
Doctor48 (4.8)283 (23.6)
Pharmacist45 (4.6)584 (48.8)
Nurse89 (9.0)62 (5.2)
Customer710 (71.8)75 (6.3)
Others73 (7.4)6 (0.5)
Not reported24 (2.4)187 (15.6)
Reporter group, n (%)
Manufacturer834 (84.3)390 (32.6)
Regional PV center146 (14.8)790 (66.0)
Others9 (0.9)17 (1.4)
Seriousness, n (%)
Yes7 (0.7)21 (1.8)
No982 (99.3)1,176 (98.2)

IQR, interquartile range; PV, pharmacovigilance.


Figure 3.KIDS-KD processing diagram.

Adverse events from KIDS-KD

The adverse events that account for ≥ 1% of the adverse events reported for liraglutide are shown in Table 5. Injection site rash (n = 229, 13.46%), injection site pruritus (n = 183, 10.76%), nausea (n = 181, 10.64%), vomiting (n = 109, 6.41%), and injection site reaction (n = 90, 5.29%) were the most commonly reported adverse events of liraglutide.

Table 5 Adverse events of GLP-1 receptor agonists from KIDS-KD

Preferred termsLiraglutide (n = 1,701)aOther antiobesity drugs (n = 1,976)aRORPRRICbNotoriety
Injection site rash229 (13.46)0 (0)c533.2616.10.85I
Injection site pruritus183 (10.76)0 (0)c426.3477.70.81I
Nausea181 (10.64)133 (6.73)1.581.650.06K/U/M
Vomiting109 (6.41)54 (2.73)2.342.440.18K/U/M
Injection site reaction90 (5.29)0 (0)c210.2222.00.68K/U/M
Diarrhoea65 (3.82)44 (2.23)1.721.74−0.07K/U/M
Dyspepsia58 (3.41)59 (2.99)1.141.15−0.35K/U/M
Headache56 (3.29)204 (10.32)0.320.30−1.51K/U/M
Pruritus47 (2.76)22 (1.11)2.482.520.02Kd/U/M
Abdominal pain47 (2.76)39 (1.97)1.401.41−0.27K/U/M
Dizziness44 (2.59)155 (7.84)0.330.31−1.52K/U/M
Constipation36 (2.12)70 (3.54)0.600.59−0.98K/U/M
Rash35 (2.06)14 (0.71)2.902.94−0.001Kd/U/M
Urticaria35 (2.06)21 (1.06)1.941.96−0.17K/U/M
Fatigue22 (1.29)11 (0.56)2.322.34−0.25K/U/M
Injection site bruising19 (1.12)0 (0)c45.345.80.16I
Insomnia17 (1.00)303 (15.33)0.070.06−3.77K/U
Others (< 1% frequency)428 (25.16)847 (42.86)....

I, included in injection site reaction; IC, information component; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); PRR, proportional reporting ratio; ROR, reporting odds ratio; U, listed on the drug labels (Saxenda® and Wegovy®) in United States.

aValues are presented as n (%).

bUnder limit of 95% confidence interval.

cAdding 0.5 to all cells before disproportionality analysis.

dReported from post-marketing surveillance in South Korea.


Seven signals that met three criteria were detected from KIDS-KD: injection site rash (ROR = 616.1), injection site pruritus (ROR = 477.7), vomiting (ROR = 2.4), injection site reaction (ROR = 222.0), pruritus (ROR = 2.5), injection site bruising (ROR = 45.8), injection site pain (ROR = 36.3). All signals were listed on drug labels in South Korea, United states, and Micromedex®.

Subgroup analysis revealed that injection site rash (n = 16, ROR = 46.6) and acne (n = 3, ROR = 39.2) met signal criteria in men and women, respectively. Acne is not currently included in either drug label or Micromedex®.

This study investigated adverse events of GLP-1 receptor agonists used for weight loss. On Twitter, nausea, vomiting, and dyspepsia were reported most frequently. Gastrointestinal adverse events were also common in clinical trials. In a 20-week trial including 564 obese adults, nausea and vomiting occurred in 34.8% and 9.7% of liraglutide group, respectively (Astrup et al. 2009). In a 56-week trial involving 3,731 adults with overweight or obesity, the most common adverse events in the liraglutide group were nausea (40.2%), diarrhea (20.9%), constipation (20.0%), and vomiting (16.3%) (Pi-Sunyer et al. 2015). Similarly, in a study for adolescents with obesity, the most common adverse events of liraglutide were nausea (42.4%) and vomiting (34.4%) (Kelly et al. 2020). Social media has a lower threshold for the general public to report adverse events than spontaneous reporting systems. Therefore, adverse events that do not require professional expertise, especially gastrointestinal adverse events, are likely to be easily reported.

From Twitter, adverse events such as somnolence, myalgia, chest pain, chest discomfort, and parosmia were also detected. These adverse events were not reported in clinical trials for approval (Pi-Sunyer et al. 2015), but were observed in four year of post-marketing surveillance (Ministry of Food and Drug Safety 2022). In particular, parosmia is a rare adverse event that only 30 cases of liraglutide have been reported in the international pharmacovigilance database of the World Health Organization (Merino et al. 2022). The adverse events observed in post-marketing surveillance were identified on Twitter, indicating the usefulness of pharmacovigilance using Twitter.

Pharmacovigilance using tweets has some disadvantages. First, tweets are less reliable because they are written by users without medical staff intervention. For this reason, an adverse event that have mentioned only once in a tweet and have never been reported by other sources may have been caused by other causes or falsely written. For this reason, it is not certain that adverse events such as increased appetite, urine abnormality, disturbance in attention, and rhinitis are adverse events caused by GLP-1 receptor agonists. Second, the number of Korean tweets referring to adverse events is still small. Nevertheless, the number of tweets mentioning Saxenda® increased in 2022 and Wegovy® is expected to be introduced in South Korea, thus more adverse event tweets are expected to be available in the future.

From KIDS-KD, the most frequently reported adverse events were injection site rash and pruritus. These adverse events have been reported more frequently than nausea and vomiting. These results are different from those of clinical trials. The adverse events derived as safety signals of GLP-1 receptor agonists were already known and could not present a new insight. Nevertheless, KIDS-KD has some advantages over pharmacovigilance using Twitter. Causal relationships are classified into ‘certain’, ‘probable’, ‘possible’, ‘unlikely’, ‘unclassified’, ‘assessable’, and ‘not applicable’ by experts. Therefore, adverse events that can be causally related to drugs can be identified (Yi et al. 2019). In addition, statistical criteria for deriving safety signals such as ROR, PRR, and IC have been established.

Acne or dermatitis acneiform was found on Twitter and derived as signal in the subgroup analysis of KIDS-KD. GLP-1 receptor agonists enhance insulin secretion (Nuffer and Trujillo 2015). Increased insulin levels increase insulin-like growth factor (IGF-1) levels (Baldwin and Tan 2021). IGF-1 directly or by lowering forkhead box class O transcription factor 1 (FoxO1) level, increases sebum production and induces hyperplasia of keratinocytes (Baldwin and Tan 2021). Excess sebum production and hyperkeratinization develop acne (Baldwin and Tan 2021). However, due to the small number of cases, further research on causality and mechanisms is needed.

This study has several limitations. First, data was limited to Korean tweets and KIDS-KD. People in South Korea post tweets less than in other countries and the number of KIDS-KD data is also significantly smaller than that of the Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS). Second, not all people who have experienced adverse events report through Twitter or spontaneous reporting systems. Adverse events are likely to be underreported, and the total number of patients administered the drug is unknown. Therefore, it is impossible to identify the incidences of adverse events. Third, Twitter, and some KIDS-KD data, are missing detailed information on demographic (i.e., age, sex), drug use (i.e., dose, number of administrations), and adverse events (i.e., date of occurrence, seriousness). Last, in the disproportionality analysis, other antiobesity drugs (ATC A08 drugs) were used instead of all other drugs. Thus, some signals might not be detected. In addition, since the compared drugs were orally administered, there were no reported cases of adverse events at the injection site, which led to abnormally high ROR values for adverse events at the injection site.

In this study, adverse events of GLP-1 receptor agonists used for weight loss were investigated using real-world data. Twitter analysis identified adverse events that were not found in clinical trials, and KIDS-KD subgroup analysis suggested the possibility of acne being a safety signal. Social media or spontaneous reporting system are meaningful as real-world data for pharmacovigilance. Further research using larger data is needed to confirm our results.

The authors greatly appreciate the Korea Institute of Drug Safety & Risk Management (Ministry of Food and Drug Safety) for providing the KIDS-KAERS database (KIDS-KD) for this study.

The authors declare that they have no conflict of interest.

  1. Astrup A, Rössner S, Van Gaal L, Rissanen A, Niskanen L, Al Hakim M, Madsen J, Rasmussen MF, Lean ME; NN8022-1807 Study Group (2009) Effects of liraglutide in the treatment of obesity: a randomised, double-blind, placebo-controlled study. Lancet 374:1606-1616. doi: 10.1016/S0140-6736(09)61375-1. Erratum in: (2010) Lancet 375:984.
    Pubmed CrossRef
  2. Baggio LL, Drucker DJ (2021) Glucagon-like peptide-1 receptor co-agonists for treating metabolic disease. Mol Metab 46:101090. doi: 10.1016/j.molmet.2020.101090
    Pubmed KoreaMed CrossRef
  3. Baldwin H, Tan J (2021) Effects of diet on acne and its response to treatment. Am J Clin Dermatol 22:55-65. doi: 10.1007/s40257-020-00542-y. Erratum in: (2020) Am J Clin Dermatol 22:67. doi: 10.1007/s40257-020-00576-2
    Pubmed KoreaMed CrossRef
  4. Bray GA (2014) Medical treatment of obesity: the past, the present and the future. Best Pract Res Clin Gastroenterol 28:665-684. doi: 10.1016/j.bpg.2014.07.015
    Pubmed CrossRef
  5. Choi JY, Choi JH, Kim MG, Rhie SJ (2021) Signal detection of adverse drug reactions of cephalosporins using data from a national pharmacovigilance database. Pharmaceuticals (Basel) 14:425. doi: 10.3390/ph14050425
    Pubmed KoreaMed CrossRef
  6. Farooq H, Niaz JS, Fakhar S, Naveed H (2021) Leveraging digital media data for pharmacovigilance. AMIA Annu Symp Proc 2020:442-451.
  7. Harpaz R, Callahan A, Tamang S, Low Y, Odgers D, Finlayson S, Jung K, LePendu P, Shah NH (2014) Text mining for adverse drug events: the promise, challenges, and state of the art. Drug Saf 37:777-790. doi: 10.1007/s40264-014-0218-z
    Pubmed KoreaMed CrossRef
  8. Harpaz R, DuMouchel W, Shah NH, Madigan D, Ryan P, Friedman C (2012) Novel data-mining methodologies for adverse drug event discovery and analysis. Clin Pharmacol Ther 91:1010-1021. doi: 10.1038/clpt.2012.50
    Pubmed KoreaMed CrossRef
  9. Hinnen D (2017) Glucagon-like peptide 1 receptor agonists for type 2 diabetes. Diabetes Spectr 30:202-210. doi: 10.2337/ds16-0026
    Pubmed KoreaMed CrossRef
  10. Ja'arah D, Al Zoubi MS, Abdelhady G, Rabi F, Tambuwala MM (2021) Role of glucagon-like peptide-1 (GLP-1) receptor agonists in hypoglycemia. Clin Med Insights Endocrinol Diabetes 14:11795514211051697. doi: 10.1177/11795514211051697
    Pubmed KoreaMed CrossRef
  11. Kelly AS, Auerbach P, Barrientos-Perez M, Gies I, Hale PM, Marcus C, Mastrandrea LD, Prabhu N, Arslanian S; NN8022-4180 Trial Investigators (2020) A randomized, controlled trial of liraglutide for adolescents with obesity. N Engl J Med 382:2117-2128. doi: 10.1056/NEJMoa1916038
    Pubmed CrossRef
  12. Lee JY, Lee YS, Kim DH, Lee HS, Yang BR, Kim MG (2021) The use of social media in detecting drug safety-related new black box warnings, labeling changes, or withdrawals: scoping review. JMIR Public Health Surveill 7:e30137. doi: 10.2196/30137
    Pubmed KoreaMed CrossRef
  13. Merino D, Gérard AO, Thümmler S, Ben Othman N, Viard D, Rocher F, Destere A, Van Obberghen EK, Drici MD (2022) Drug-associated parosmia: new perspectives from the WHO safety database. J Clin Med 11:4641. doi: 10.3390/jcm11164641
    Pubmed KoreaMed CrossRef
  14. Ministry of Food and Drug Safety (2022) SAXENDA® [package insert]. Novo Nordisk Pharma Korea Limited. https://nedrug.mfds.go.kr/pbp/CCBBB01/getItemDetailCache?cacheSeq=201705815aupdateTs2023-01-20%2017:47:09.0b Accessed 30 December 2022.
  15. Mosavat M, Omar SZ, Jamalpour S, Tan PC (2020) Serum glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) in association with the risk of gestational diabetes: a prospective case-control study. J Diabetes Res 2020:9072492. doi: 10.1155/2020/9072492
    Pubmed KoreaMed CrossRef
  16. Nuffer WA, Trujillo JM (2015) Liraglutide: a new option for the treatment of obesity. Pharmacotherapy 35:926-934. doi: 10.1002/phar.1639
    Pubmed CrossRef
  17. Pi-Sunyer X, Astrup A, Fujioka K, Greenway F, Halpern A, Krempf M, Lau DC, le Roux CW, Violante Ortiz R, Jensen CB, Wilding JP; SCALE Obesity and Prediabetes NN8022-1839 Study Group (2015) A randomized, controlled trial of 3.0 mg of liraglutide in weight management. N Engl J Med 373:11-22. doi: 10.1056/NEJMoa1411892
    Pubmed CrossRef
  18. Rubino D, Abrahamsson N, Davies M, Hesse D, Greenway FL, Jensen C, Lingvay I, Mosenzon O, Rosenstock J, Rubio MA, Rudofsky G, Tadayon S, Wadden TA, Dicker D; STEP 4 Investigators (2021) Effect of continued weekly subcutaneous semaglutide vs placebo on weight loss maintenance in adults with overweight or obesity: the STEP 4 randomized clinical trial. JAMA 325:1414-1425. doi: 10.1001/jama.2021.3224
    Pubmed KoreaMed CrossRef
  19. Sarker A, Ginn R, Nikfarjam A, O'Connor K, Smith K, Jayaraman S, Upadhaya T, Gonzalez G (2015) Utilizing social media data for pharmacovigilance: a review. J Biomed Inform 54:202-212. doi: 10.1016/j.jbi.2015.02.004
    Pubmed KoreaMed CrossRef
  20. Shin H, Cha J, Lee Y, Kim JY, Lee S (2021) Real-world data-based adverse drug reactions detection from the Korea Adverse Event Reporting System databases with electronic health records-based detection algorithm. Health Informatics J 27:14604582211033014. doi: 10.1177/14604582211033014
    Pubmed CrossRef
  21. Singh G, Krauthamer M, Bjalme-Evans M (2022) Wegovy (semaglutide): a new weight loss drug for chronic weight management. J Investig Med 70:5-13. doi: 10.1136/jim-2021-001952
    Pubmed KoreaMed CrossRef
  22. U.S. Food and Drug Administration (2022) SAXENDA® [package insert]. U.S. Food and Drug Administration. https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/206321s015lbl.pdf Accessed 30 December 2022.
  23. Wilding JPH, Batterham RL, Calanna S, Davies M, Van Gaal LF, Lingvay I, McGowan BM, Rosenstock J, Tran MTD, Wadden TA, Wharton S, Yokote K, Zeuthen N, Kushner RF; STEP 1 Study Group (2021) Once-weekly semaglutide in adults with overweight or obesity. N Engl J Med 384:989-1002. doi: 10.1056/NEJMoa2032183
    Pubmed CrossRef
  24. Yi H, Lee JH, Shin JY (2019) Signal detection for cardiovascular adverse events of DPP-4 inhibitors using the Korea adverse event reporting system database, 2008-2016. Yonsei Med J 60:200-207. doi: 10.3349/ymj.2019.60.2.200
    Pubmed KoreaMed CrossRef

Article

Original Research Article

DTT 2023; 2(1): 41-48

Published online March 31, 2023 https://doi.org/10.58502/DTT.23.0003

Copyright © The Pharmaceutical Society of Korea.

Adverse Events of GLP-1 Receptor Agonists for Weight Loss: Twitter and a National Pharmacovigilance Database

Hyun-Myung Back1* , Seo-A Choi1* , Myeong Gyu Kim1,2

1College of Pharmacy, Ewha Womans University, Seoul, Korea
2Graduate School of Pharmaceutical Sciences, Ewha Womans University, Seoul, Korea

Correspondence to:Myeong Gyu Kim, kimmg@ewha.ac.kr
*These authors contributed equally to this study.

Received: January 26, 2023; Revised: March 3, 2023; Accepted: March 3, 2023

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ((http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Glucagon-like peptide-1 (GLP-1) receptor agonists became the most prescribed drugs for weight loss. This study aimed to investigate adverse events of GLP-1 receptor agonists using Twitter and KIDS-KAERS database (KIDS-KD). 116 Korean tweets on Twitter and 1,701 liraglutide-adverse event combinations from KIDS-KD were used for analysis. The most common adverse events were nausea (37.76%), vomiting (12.59%), dyspepsia (9.79%), dizziness (4.90%), and somnolence (4.90%) on Twitter. Adverse events observed in the post-marketing surveillance such as somnolence, myalgia, chest pain/ discomfort, and parosmia were found on Twitter. In KIDS-KD, injection site rash (13.46%), injection site pruritus (10.76%), nausea (10.64%), vomiting (6.41%), and injection site reaction (5.29%) were commonly reported. Signals were injection site rash (reporting odds ratio [ROR] = 616.1), injection site pruritus (ROR = 477.7), vomiting (ROR = 2.4), injection site reaction (ROR = 222.0), pruritus (ROR = 2.5), injection site bruising (ROR = 45.8), injection site pain (ROR = 36.3). Subgroup analysis revealed that injection site rash (n = 16, ROR = 46.6) and acne (n = 3, ROR = 39.2) met signal criteria in men and women, respectively. Twitter analysis identified adverse events that were not found in clinical trials, and KIDS-KD subgroup analysis suggested the possibility of acne being a safety signal. Social media and KIDS-KD were useful for pharmacovigilance.

Keywords: GLP-1 receptor agonists, KIDS-KAERS database (KIDS-KD), pharmacovigilance, adverse events, Twitter

Introduction

Glucagon-like peptide-1 (GLP-1) is an incretin peptide secreted from intestinal endocrine cells and enhances glucose-dependent insulin secretion from pancreatic β-cells (Hinnen 2017; Mosavat et al. 2020; Ja'arah et al. 2021). For this reason, GLP-1 receptor agonists were approved as antidiabetic drugs. GLP-1 receptor agonists also support weight loss through delayed gastric emptying, reduced fat absorption, and reduced food intake (Baggio and Drucker 2021). Liraglutide, a GLP-1 receptor agonist, was previously approved for the treatment of type 2 diabetes under the brand name Victoza®. Saxenda®, another brand name for liraglutide, was approved by the U.S. Food and Drug Administration in December 2014 for the treatment of obesity (Nuffer and Trujillo 2015). Semaglutide was also approved for diabetes treatment (Ozempic® and Rybelsus®) in December 2017 and weight loss (Wegovy®) in June 2021 (Singh et al. 2022).

While appetite suppressants acting on the central nervous system are addictive and have side effects such as paresthesia, depression, insomnia, decreased concentration, and hallucinations (Bray 2014), GLP-1 receptor agonists are relatively safe and thus in the spotlight as a novel weight loss drug. The most common adverse events of GLP-1 receptor agonists include nausea, diarrhea, constipation, vomiting, injection site reactions, headache, hypoglycemia, and dyspepsia (Astrup et al. 2009; Pi-Sunyer et al. 2015; Rubino et al. 2021; Wilding et al. 2021).

Real-world data can have different implications from the controlled environment of clinical trials. People who are not overweight or obese also administer drugs to lose weight. In addition, long-term or rare adverse events that were not observed during clinical trials can be observed in real-world settings. The use of social media has grown exponentially over the past few years, which has become a major source of information. With the prosperity of social networks, people have become more inclined to post prescription drugs and related adverse events and share their treatment experiences (Harpaz et al. 2012; Harpaz et al. 2014). By sharing thoughts and experiences with a vast number of users, social media such as Twitter has become a useful resource for post-marketing surveillance of drugs, complementing existing pharmacovigilance systems (Sarker et al. 2015; Farooq et al. 2021). Pharmacovigilance through social media can detect adverse events before the regulatory agency’s action (Lee et al. 2021). Furthermore, real-world studies in South Korea mainly use claim data from the Health Insurance Review & Assessment Service, but there is a limitation in that non-reimbursable drugs are excluded. Since GLP-1 receptor agonists used for weight loss are non-reimbursable drugs, studies using social media are appropriate.

Spontaneous adverse event reports are another real-world data specialized for adverse events. The Korea Institute of Drug Safety & Risk Management (Ministry of Food and Drug Safety) provides Korea Adverse Event Reporting System (KAERS) database (KIDS-KD) for pharmacovigilance analysis (Shin et al. 2021). Several studies have shown that signals of adverse events were detected by data mining on KIDS-KD.

This study aimed to investigate adverse events of GLP-1 receptor agonists (Saxenda® and Wegovy®) used for weight loss in South Korea using Twitter and KIDS-KD.

Materials|Methods

Tweet collection

Tweets which were posted from August 2017 (Saxenda® was approved in South Korea on July 21, 2017) to June 2022 were collected with restrictions in the Korean language using the following search query: ‘삭센다’ (Saxenda in Korean) OR ‘삭샌다’ (Saxenda in Korean) OR ‘위고비’ (Wegovy in Korean) OR ‘리라글루타이드’ (liraglutide in Korean) OR ‘리라글루티드’ (liraglutide in Korean) OR ‘세마글루타이드’ (semaglutide in Korean) OR ‘세마글루티드’ (semaglutide in Korean). Although Wegovy® is not currently marketed in South Korea, it is included in the search because it may be mentioned on social media. Tweet search was conducted on July 6, 2022, using Python version 3.7 (Python Software Foundation, Fredericksburg, VA, USA) and Twitter application programming interface for academic research. The collected data included user identifier, tweet text, date posted, and the number of retweets, replies, likes, and quote tweets. The study was exempted from Institutional Review Board review (ewha-202203-0027-01).

Tweet annotation

All tweets were manually labeled. First, tweets that were unrelated to the GLP-1 receptor agonists were excluded. Second, tweets were classified as tweets mentioning adverse events or not. Tweets mentioning adverse events were further classified as first-hand experience or non-first-hand experience. Non-first-hand experience included adverse events from news, study results, or someone else’s experience. The adverse events mentioned in the tweet were coded as Medical Dictionary for Regulatory Activities (MedDRA) Preferred Terms version 25.0.

Tweet analysis

The annual frequency of tweets mentioning adverse events, the number of adverse events per tweet, and the number of adverse event tweets per user were analyzed. The number of adverse events was analyzed by MedDRA Preferred Terms. Drug labels in South Korea (Ministry of Food and Drug Safety 2022) and United States (U.S. Food and Drug Administration 2022), and Micromedex® (www.micromedexsolutions.com) were reviewed to identify adverse events that have been listed.

KIDS-KD data collection and processing

From August 2017 to December 2020, adverse event reports including liraglutide or other antiobesity drugs (the Anatomical Therapeutic Chemical [ATC] Classification System code A08) were provided by KIDS. The data consisted of eight tables: basic information (ADR_REPORT_BASIC), drug information (DRUG_INFO_ADR), adverse drug reaction information (ADR_INFO_REPORT), seriousness of AEs (SERIOUS_ADR), reporter information (REPORTOR_ADR), causality assessment (ASSESSMENT_ADR), medical history (DISEASE_HIST_ADR), and group information (GROUP).

Only initial reports were used for analysis. By reviewing drug information and medical history, reports with a history of diabetes or antidiabetic drugs were removed to exclude cases where liraglutide was used for diabetes treatment. If there was no record of medical history or concurrent medications, it was difficult to determine whether the data was truly missing, so they were included in the analysis. From the causality assessment table, only reports that liraglutide or other antiobesity drugs were assessed as 'certain', 'probable', and 'possible' remained for the disproportionality analysis. If information for causality assessment was missing, it was excluded from the subsequent analysis.

Disproportionality analysis of KIDS-KD

A 2-by-2 table was constructed from the drug and adverse event combination (Table 1). From the table, proportional reporting ratio (PRR), reporting odds ratio (ROR), and information component (IC) were calculated as Table 2 (Choi et al. 2021). Signals that met the three criteria of PRR, ROR, and IC were identified.

Table 1 . 2 × 2 table for disproportionality analysis.

Specific adverse eventOther adverse events
LiraglutideAB
Other antiobesity drugsCD


Table 2 . Definition and criteria of signal detection for each indicator.

DefinitionCriteria of signal detection
PRR(A/(A+B))/ (C/(C+D))PRR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ROR(A/B)/ (C/D)ROR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ICLog2 (P(AE, drug)/ P(AE) × P(drug))Under limit of 95% confidence interval ≥ 0

IC, information component; PRR, proportional reporting ratio; ROR, reporting odds ratio..



Subgroup analysis was conducted according to sex. Based on sex information recorded in the basic information table (PTNT_SEX = 1 or 2), PRR, ROR, and IC values were calculated by sex. Cases with missing sex information were excluded from the subgroup analysis.

Data processing and statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary NC). Drug labels in South Korea (Ministry of Food and Drug Safety 2022) and United States (U.S. Food and Drug Administration 2022), and Micromedex® were reviewed to identify adverse events that have been listed.

Results

Characteristics of tweets

Fig. 1 shows a tweet processing diagram. A total of 2,499 Korean tweets were collected. Of 2,499 tweets, 235 tweets mentioned adverse events. Finally, 116 first-hand experience tweets from 73 users were used for analysis.

Figure 1. Tweet processing diagram.

The annual frequency of tweets is shown in Fig. 2. An average of 20 tweets per year from 2018 to 2021 mentioned adverse events. In 2022, 36 tweets mentioned adverse events by the first half. Most tweets contained one (n = 94, 81.0%) or two (n = 18, 15.5%) adverse events in their text. Three tweets (2.6%) contained three adverse events and one tweet (0.9%) contained four adverse events. Most users posted one (n = 54, 74.0%) or two (n = 10, 13.7%) tweets mentioning adverse events.

Figure 2. Annual frequency of tweets. A solid line indicates the number of adverse event tweets which were first-hand experience. A dotted line indicates percentage of adverse event tweets to total tweets mentioning GLP-1 receptor agonists.

Adverse events from tweets

Table 3 shows MedDRA Preferred Terms sorted in the order frequently mentioned on tweets. Of 143 adverse events mentioned on tweets, two were not classified as Preferred Terms because they used a broader term. The top five adverse events were nausea (n = 54, 37.76%), vomiting (n = 18, 12.59%), dyspepsia (n = 14, 9.79%), dizziness (n = 7, 4.90%), and somnolence (n = 7, 4.90%).

Table 3 . Adverse events of GLP-1 receptor agonists mentioned on Twitter.

Preferred termsNumber of mentioning (n = 143)aNotoriety
Nausea54 (37.76)K/U/M
Vomiting18 (12.59)K/U/M
Dyspepsia14 (9.79)K/U/M
Dizziness7 (4.90)K/U/M
Somnolence7 (4.90)Kb/Uc
Injection site reaction5 (3.50)K/U/M
Diarrhoea3 (2.10)K/U/M
Myalgia2 (1.40)Kb
Abdominal pain2 (1.40)K/U/M
Insomnia2 (1.40)K/U
Headache2 (1.40)K/U/M
Fatigue2 (1.40)K/U/M
Hypoglycaemia2 (1.40)K/U/M
Flatulence2 (1.40)K/U/M
Injection site bruising2 (1.40)I
Injection site rash2 (1.40)I
Chest pain1 (0.70)Kb
Chest discomfort1 (0.70)Kb
Urticaria1 (0.70)K/U/M
Asthenia1 (0.70)K/U/M
Increased appetite1 (0.70)No
Urine abnormality1 (0.70)No
Vision blurred1 (0.70)Uc
Allergy1 (0.70)K/U
Rhinitis1 (0.70)No
Dermatitis acneiform1 (0.70)No
Depression1 (0.70)U/M
Parosmia1 (0.70)Kb
Gastrooesophageal reflux disease1 (0.70)K/U/M
Injection site pruritus1 (0.70)I
Disturbance in attention1 (0.70)No
Unclassifiable2 (1.40).

I, included in injection site reaction; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); U, listed on the drug labels (Saxenda® and Wegovy®) in United States..

aValues are presented as n (%)..

bReported from post-marketing surveillance in South Korea..

cSuggested as signs and symptoms of hypoglycemia..



Adverse events such as somnolence, myalgia, chest pain, chest discomfort, and parosmia identified in tweets are only recorded in the Korean label. This is because they were not originally reported in clinical trials for approval, but rather reported in the post-marketing surveillance conducted in Korea over four years with 670 subjects.

Characteristics of KIDS-KD

Fig. 3 shows a KIDS-KD processing diagram. From August 2017 to December 2020, there were 3,281 reports for liraglutide or antiobesity drugs in KIDS-KD. A total of 1,701 liraglutide- adverse event combinations and 1,976 other antiobesity drugs-adverse event combinations were used for disproportionality analysis. Table 4 shows the characteristics of adverse event reports on liraglutide and other antiobesity drugs. Adverse events were primarily reported by women aged 35 to 55, and were reported by the patients through the pharmaceutical company.

Table 4 . Characteristics of adverse events reports on liraglutide and other antiobesity drugs.

CharacteristicsLiraglutide (n = 989)Other antiobesity drugs (n = 1,197)
Age, median (IQR), years45 (35-55)42 (33-52)
Sex, n (%)
Men113 (11.4)140 (11.7)
Women723 (73.1)973 (81.3)
Not reported153 (15.5)84 (7.0)
Reporting year, n (%)
20172 (0.2)157 (15.9)
2018199 (20.1)304 (30.7)
2019401 (40.6)428 (43.3)
2020387 (39.1)308 (31.1)
Original reporter, n (%)
Doctor48 (4.8)283 (23.6)
Pharmacist45 (4.6)584 (48.8)
Nurse89 (9.0)62 (5.2)
Customer710 (71.8)75 (6.3)
Others73 (7.4)6 (0.5)
Not reported24 (2.4)187 (15.6)
Reporter group, n (%)
Manufacturer834 (84.3)390 (32.6)
Regional PV center146 (14.8)790 (66.0)
Others9 (0.9)17 (1.4)
Seriousness, n (%)
Yes7 (0.7)21 (1.8)
No982 (99.3)1,176 (98.2)

IQR, interquartile range; PV, pharmacovigilance..



Figure 3. KIDS-KD processing diagram.

Adverse events from KIDS-KD

The adverse events that account for ≥ 1% of the adverse events reported for liraglutide are shown in Table 5. Injection site rash (n = 229, 13.46%), injection site pruritus (n = 183, 10.76%), nausea (n = 181, 10.64%), vomiting (n = 109, 6.41%), and injection site reaction (n = 90, 5.29%) were the most commonly reported adverse events of liraglutide.

Table 5 . Adverse events of GLP-1 receptor agonists from KIDS-KD.

Preferred termsLiraglutide (n = 1,701)aOther antiobesity drugs (n = 1,976)aRORPRRICbNotoriety
Injection site rash229 (13.46)0 (0)c533.2616.10.85I
Injection site pruritus183 (10.76)0 (0)c426.3477.70.81I
Nausea181 (10.64)133 (6.73)1.581.650.06K/U/M
Vomiting109 (6.41)54 (2.73)2.342.440.18K/U/M
Injection site reaction90 (5.29)0 (0)c210.2222.00.68K/U/M
Diarrhoea65 (3.82)44 (2.23)1.721.74−0.07K/U/M
Dyspepsia58 (3.41)59 (2.99)1.141.15−0.35K/U/M
Headache56 (3.29)204 (10.32)0.320.30−1.51K/U/M
Pruritus47 (2.76)22 (1.11)2.482.520.02Kd/U/M
Abdominal pain47 (2.76)39 (1.97)1.401.41−0.27K/U/M
Dizziness44 (2.59)155 (7.84)0.330.31−1.52K/U/M
Constipation36 (2.12)70 (3.54)0.600.59−0.98K/U/M
Rash35 (2.06)14 (0.71)2.902.94−0.001Kd/U/M
Urticaria35 (2.06)21 (1.06)1.941.96−0.17K/U/M
Fatigue22 (1.29)11 (0.56)2.322.34−0.25K/U/M
Injection site bruising19 (1.12)0 (0)c45.345.80.16I
Insomnia17 (1.00)303 (15.33)0.070.06−3.77K/U
Others (< 1% frequency)428 (25.16)847 (42.86)....

I, included in injection site reaction; IC, information component; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); PRR, proportional reporting ratio; ROR, reporting odds ratio; U, listed on the drug labels (Saxenda® and Wegovy®) in United States..

aValues are presented as n (%)..

bUnder limit of 95% confidence interval..

cAdding 0.5 to all cells before disproportionality analysis..

dReported from post-marketing surveillance in South Korea..



Seven signals that met three criteria were detected from KIDS-KD: injection site rash (ROR = 616.1), injection site pruritus (ROR = 477.7), vomiting (ROR = 2.4), injection site reaction (ROR = 222.0), pruritus (ROR = 2.5), injection site bruising (ROR = 45.8), injection site pain (ROR = 36.3). All signals were listed on drug labels in South Korea, United states, and Micromedex®.

Subgroup analysis revealed that injection site rash (n = 16, ROR = 46.6) and acne (n = 3, ROR = 39.2) met signal criteria in men and women, respectively. Acne is not currently included in either drug label or Micromedex®.

Discussion

This study investigated adverse events of GLP-1 receptor agonists used for weight loss. On Twitter, nausea, vomiting, and dyspepsia were reported most frequently. Gastrointestinal adverse events were also common in clinical trials. In a 20-week trial including 564 obese adults, nausea and vomiting occurred in 34.8% and 9.7% of liraglutide group, respectively (Astrup et al. 2009). In a 56-week trial involving 3,731 adults with overweight or obesity, the most common adverse events in the liraglutide group were nausea (40.2%), diarrhea (20.9%), constipation (20.0%), and vomiting (16.3%) (Pi-Sunyer et al. 2015). Similarly, in a study for adolescents with obesity, the most common adverse events of liraglutide were nausea (42.4%) and vomiting (34.4%) (Kelly et al. 2020). Social media has a lower threshold for the general public to report adverse events than spontaneous reporting systems. Therefore, adverse events that do not require professional expertise, especially gastrointestinal adverse events, are likely to be easily reported.

From Twitter, adverse events such as somnolence, myalgia, chest pain, chest discomfort, and parosmia were also detected. These adverse events were not reported in clinical trials for approval (Pi-Sunyer et al. 2015), but were observed in four year of post-marketing surveillance (Ministry of Food and Drug Safety 2022). In particular, parosmia is a rare adverse event that only 30 cases of liraglutide have been reported in the international pharmacovigilance database of the World Health Organization (Merino et al. 2022). The adverse events observed in post-marketing surveillance were identified on Twitter, indicating the usefulness of pharmacovigilance using Twitter.

Pharmacovigilance using tweets has some disadvantages. First, tweets are less reliable because they are written by users without medical staff intervention. For this reason, an adverse event that have mentioned only once in a tweet and have never been reported by other sources may have been caused by other causes or falsely written. For this reason, it is not certain that adverse events such as increased appetite, urine abnormality, disturbance in attention, and rhinitis are adverse events caused by GLP-1 receptor agonists. Second, the number of Korean tweets referring to adverse events is still small. Nevertheless, the number of tweets mentioning Saxenda® increased in 2022 and Wegovy® is expected to be introduced in South Korea, thus more adverse event tweets are expected to be available in the future.

From KIDS-KD, the most frequently reported adverse events were injection site rash and pruritus. These adverse events have been reported more frequently than nausea and vomiting. These results are different from those of clinical trials. The adverse events derived as safety signals of GLP-1 receptor agonists were already known and could not present a new insight. Nevertheless, KIDS-KD has some advantages over pharmacovigilance using Twitter. Causal relationships are classified into ‘certain’, ‘probable’, ‘possible’, ‘unlikely’, ‘unclassified’, ‘assessable’, and ‘not applicable’ by experts. Therefore, adverse events that can be causally related to drugs can be identified (Yi et al. 2019). In addition, statistical criteria for deriving safety signals such as ROR, PRR, and IC have been established.

Acne or dermatitis acneiform was found on Twitter and derived as signal in the subgroup analysis of KIDS-KD. GLP-1 receptor agonists enhance insulin secretion (Nuffer and Trujillo 2015). Increased insulin levels increase insulin-like growth factor (IGF-1) levels (Baldwin and Tan 2021). IGF-1 directly or by lowering forkhead box class O transcription factor 1 (FoxO1) level, increases sebum production and induces hyperplasia of keratinocytes (Baldwin and Tan 2021). Excess sebum production and hyperkeratinization develop acne (Baldwin and Tan 2021). However, due to the small number of cases, further research on causality and mechanisms is needed.

This study has several limitations. First, data was limited to Korean tweets and KIDS-KD. People in South Korea post tweets less than in other countries and the number of KIDS-KD data is also significantly smaller than that of the Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS). Second, not all people who have experienced adverse events report through Twitter or spontaneous reporting systems. Adverse events are likely to be underreported, and the total number of patients administered the drug is unknown. Therefore, it is impossible to identify the incidences of adverse events. Third, Twitter, and some KIDS-KD data, are missing detailed information on demographic (i.e., age, sex), drug use (i.e., dose, number of administrations), and adverse events (i.e., date of occurrence, seriousness). Last, in the disproportionality analysis, other antiobesity drugs (ATC A08 drugs) were used instead of all other drugs. Thus, some signals might not be detected. In addition, since the compared drugs were orally administered, there were no reported cases of adverse events at the injection site, which led to abnormally high ROR values for adverse events at the injection site.

In this study, adverse events of GLP-1 receptor agonists used for weight loss were investigated using real-world data. Twitter analysis identified adverse events that were not found in clinical trials, and KIDS-KD subgroup analysis suggested the possibility of acne being a safety signal. Social media or spontaneous reporting system are meaningful as real-world data for pharmacovigilance. Further research using larger data is needed to confirm our results.

Acknowledgements

The authors greatly appreciate the Korea Institute of Drug Safety & Risk Management (Ministry of Food and Drug Safety) for providing the KIDS-KAERS database (KIDS-KD) for this study.

Conflict of Interest

The authors declare that they have no conflict of interest.

Fig 1.

Figure 1.Tweet processing diagram.
Drug Targets and Therapeutics 2023; 2: 41-48https://doi.org/10.58502/DTT.23.0003

Fig 2.

Figure 2.Annual frequency of tweets. A solid line indicates the number of adverse event tweets which were first-hand experience. A dotted line indicates percentage of adverse event tweets to total tweets mentioning GLP-1 receptor agonists.
Drug Targets and Therapeutics 2023; 2: 41-48https://doi.org/10.58502/DTT.23.0003

Fig 3.

Figure 3.KIDS-KD processing diagram.
Drug Targets and Therapeutics 2023; 2: 41-48https://doi.org/10.58502/DTT.23.0003

Table 1 2 × 2 table for disproportionality analysis

Specific adverse eventOther adverse events
LiraglutideAB
Other antiobesity drugsCD

Table 2 Definition and criteria of signal detection for each indicator

DefinitionCriteria of signal detection
PRR(A/(A+B))/ (C/(C+D))PRR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ROR(A/B)/ (C/D)ROR ≥ 2, chi-squared ≥ 4, and A ≥ 3
ICLog2 (P(AE, drug)/ P(AE) × P(drug))Under limit of 95% confidence interval ≥ 0

IC, information component; PRR, proportional reporting ratio; ROR, reporting odds ratio.


Table 3 Adverse events of GLP-1 receptor agonists mentioned on Twitter

Preferred termsNumber of mentioning (n = 143)aNotoriety
Nausea54 (37.76)K/U/M
Vomiting18 (12.59)K/U/M
Dyspepsia14 (9.79)K/U/M
Dizziness7 (4.90)K/U/M
Somnolence7 (4.90)Kb/Uc
Injection site reaction5 (3.50)K/U/M
Diarrhoea3 (2.10)K/U/M
Myalgia2 (1.40)Kb
Abdominal pain2 (1.40)K/U/M
Insomnia2 (1.40)K/U
Headache2 (1.40)K/U/M
Fatigue2 (1.40)K/U/M
Hypoglycaemia2 (1.40)K/U/M
Flatulence2 (1.40)K/U/M
Injection site bruising2 (1.40)I
Injection site rash2 (1.40)I
Chest pain1 (0.70)Kb
Chest discomfort1 (0.70)Kb
Urticaria1 (0.70)K/U/M
Asthenia1 (0.70)K/U/M
Increased appetite1 (0.70)No
Urine abnormality1 (0.70)No
Vision blurred1 (0.70)Uc
Allergy1 (0.70)K/U
Rhinitis1 (0.70)No
Dermatitis acneiform1 (0.70)No
Depression1 (0.70)U/M
Parosmia1 (0.70)Kb
Gastrooesophageal reflux disease1 (0.70)K/U/M
Injection site pruritus1 (0.70)I
Disturbance in attention1 (0.70)No
Unclassifiable2 (1.40).

I, included in injection site reaction; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); U, listed on the drug labels (Saxenda® and Wegovy®) in United States.

aValues are presented as n (%).

bReported from post-marketing surveillance in South Korea.

cSuggested as signs and symptoms of hypoglycemia.


Table 4 Characteristics of adverse events reports on liraglutide and other antiobesity drugs

CharacteristicsLiraglutide (n = 989)Other antiobesity drugs (n = 1,197)
Age, median (IQR), years45 (35-55)42 (33-52)
Sex, n (%)
Men113 (11.4)140 (11.7)
Women723 (73.1)973 (81.3)
Not reported153 (15.5)84 (7.0)
Reporting year, n (%)
20172 (0.2)157 (15.9)
2018199 (20.1)304 (30.7)
2019401 (40.6)428 (43.3)
2020387 (39.1)308 (31.1)
Original reporter, n (%)
Doctor48 (4.8)283 (23.6)
Pharmacist45 (4.6)584 (48.8)
Nurse89 (9.0)62 (5.2)
Customer710 (71.8)75 (6.3)
Others73 (7.4)6 (0.5)
Not reported24 (2.4)187 (15.6)
Reporter group, n (%)
Manufacturer834 (84.3)390 (32.6)
Regional PV center146 (14.8)790 (66.0)
Others9 (0.9)17 (1.4)
Seriousness, n (%)
Yes7 (0.7)21 (1.8)
No982 (99.3)1,176 (98.2)

IQR, interquartile range; PV, pharmacovigilance.


Table 5 Adverse events of GLP-1 receptor agonists from KIDS-KD

Preferred termsLiraglutide (n = 1,701)aOther antiobesity drugs (n = 1,976)aRORPRRICbNotoriety
Injection site rash229 (13.46)0 (0)c533.2616.10.85I
Injection site pruritus183 (10.76)0 (0)c426.3477.70.81I
Nausea181 (10.64)133 (6.73)1.581.650.06K/U/M
Vomiting109 (6.41)54 (2.73)2.342.440.18K/U/M
Injection site reaction90 (5.29)0 (0)c210.2222.00.68K/U/M
Diarrhoea65 (3.82)44 (2.23)1.721.74−0.07K/U/M
Dyspepsia58 (3.41)59 (2.99)1.141.15−0.35K/U/M
Headache56 (3.29)204 (10.32)0.320.30−1.51K/U/M
Pruritus47 (2.76)22 (1.11)2.482.520.02Kd/U/M
Abdominal pain47 (2.76)39 (1.97)1.401.41−0.27K/U/M
Dizziness44 (2.59)155 (7.84)0.330.31−1.52K/U/M
Constipation36 (2.12)70 (3.54)0.600.59−0.98K/U/M
Rash35 (2.06)14 (0.71)2.902.94−0.001Kd/U/M
Urticaria35 (2.06)21 (1.06)1.941.96−0.17K/U/M
Fatigue22 (1.29)11 (0.56)2.322.34−0.25K/U/M
Injection site bruising19 (1.12)0 (0)c45.345.80.16I
Insomnia17 (1.00)303 (15.33)0.070.06−3.77K/U
Others (< 1% frequency)428 (25.16)847 (42.86)....

I, included in injection site reaction; IC, information component; K, listed on the drug label (Saxenda®) in South Korea; M, listed on the Micromedex® (liraglutide and semaglutide); PRR, proportional reporting ratio; ROR, reporting odds ratio; U, listed on the drug labels (Saxenda® and Wegovy®) in United States.

aValues are presented as n (%).

bUnder limit of 95% confidence interval.

cAdding 0.5 to all cells before disproportionality analysis.

dReported from post-marketing surveillance in South Korea.


References

  1. Astrup A, Rössner S, Van Gaal L, Rissanen A, Niskanen L, Al Hakim M, Madsen J, Rasmussen MF, Lean ME; NN8022-1807 Study Group (2009) Effects of liraglutide in the treatment of obesity: a randomised, double-blind, placebo-controlled study. Lancet 374:1606-1616. doi: 10.1016/S0140-6736(09)61375-1. Erratum in: (2010) Lancet 375:984.
    Pubmed CrossRef
  2. Baggio LL, Drucker DJ (2021) Glucagon-like peptide-1 receptor co-agonists for treating metabolic disease. Mol Metab 46:101090. doi: 10.1016/j.molmet.2020.101090
    Pubmed KoreaMed CrossRef
  3. Baldwin H, Tan J (2021) Effects of diet on acne and its response to treatment. Am J Clin Dermatol 22:55-65. doi: 10.1007/s40257-020-00542-y. Erratum in: (2020) Am J Clin Dermatol 22:67. doi: 10.1007/s40257-020-00576-2
    Pubmed KoreaMed CrossRef
  4. Bray GA (2014) Medical treatment of obesity: the past, the present and the future. Best Pract Res Clin Gastroenterol 28:665-684. doi: 10.1016/j.bpg.2014.07.015
    Pubmed CrossRef
  5. Choi JY, Choi JH, Kim MG, Rhie SJ (2021) Signal detection of adverse drug reactions of cephalosporins using data from a national pharmacovigilance database. Pharmaceuticals (Basel) 14:425. doi: 10.3390/ph14050425
    Pubmed KoreaMed CrossRef
  6. Farooq H, Niaz JS, Fakhar S, Naveed H (2021) Leveraging digital media data for pharmacovigilance. AMIA Annu Symp Proc 2020:442-451.
  7. Harpaz R, Callahan A, Tamang S, Low Y, Odgers D, Finlayson S, Jung K, LePendu P, Shah NH (2014) Text mining for adverse drug events: the promise, challenges, and state of the art. Drug Saf 37:777-790. doi: 10.1007/s40264-014-0218-z
    Pubmed KoreaMed CrossRef
  8. Harpaz R, DuMouchel W, Shah NH, Madigan D, Ryan P, Friedman C (2012) Novel data-mining methodologies for adverse drug event discovery and analysis. Clin Pharmacol Ther 91:1010-1021. doi: 10.1038/clpt.2012.50
    Pubmed KoreaMed CrossRef
  9. Hinnen D (2017) Glucagon-like peptide 1 receptor agonists for type 2 diabetes. Diabetes Spectr 30:202-210. doi: 10.2337/ds16-0026
    Pubmed KoreaMed CrossRef
  10. Ja'arah D, Al Zoubi MS, Abdelhady G, Rabi F, Tambuwala MM (2021) Role of glucagon-like peptide-1 (GLP-1) receptor agonists in hypoglycemia. Clin Med Insights Endocrinol Diabetes 14:11795514211051697. doi: 10.1177/11795514211051697
    Pubmed KoreaMed CrossRef
  11. Kelly AS, Auerbach P, Barrientos-Perez M, Gies I, Hale PM, Marcus C, Mastrandrea LD, Prabhu N, Arslanian S; NN8022-4180 Trial Investigators (2020) A randomized, controlled trial of liraglutide for adolescents with obesity. N Engl J Med 382:2117-2128. doi: 10.1056/NEJMoa1916038
    Pubmed CrossRef
  12. Lee JY, Lee YS, Kim DH, Lee HS, Yang BR, Kim MG (2021) The use of social media in detecting drug safety-related new black box warnings, labeling changes, or withdrawals: scoping review. JMIR Public Health Surveill 7:e30137. doi: 10.2196/30137
    Pubmed KoreaMed CrossRef
  13. Merino D, Gérard AO, Thümmler S, Ben Othman N, Viard D, Rocher F, Destere A, Van Obberghen EK, Drici MD (2022) Drug-associated parosmia: new perspectives from the WHO safety database. J Clin Med 11:4641. doi: 10.3390/jcm11164641
    Pubmed KoreaMed CrossRef
  14. Ministry of Food and Drug Safety (2022) SAXENDA® [package insert]. Novo Nordisk Pharma Korea Limited. https://nedrug.mfds.go.kr/pbp/CCBBB01/getItemDetailCache?cacheSeq=201705815aupdateTs2023-01-20%2017:47:09.0b Accessed 30 December 2022.
  15. Mosavat M, Omar SZ, Jamalpour S, Tan PC (2020) Serum glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) in association with the risk of gestational diabetes: a prospective case-control study. J Diabetes Res 2020:9072492. doi: 10.1155/2020/9072492
    Pubmed KoreaMed CrossRef
  16. Nuffer WA, Trujillo JM (2015) Liraglutide: a new option for the treatment of obesity. Pharmacotherapy 35:926-934. doi: 10.1002/phar.1639
    Pubmed CrossRef
  17. Pi-Sunyer X, Astrup A, Fujioka K, Greenway F, Halpern A, Krempf M, Lau DC, le Roux CW, Violante Ortiz R, Jensen CB, Wilding JP; SCALE Obesity and Prediabetes NN8022-1839 Study Group (2015) A randomized, controlled trial of 3.0 mg of liraglutide in weight management. N Engl J Med 373:11-22. doi: 10.1056/NEJMoa1411892
    Pubmed CrossRef
  18. Rubino D, Abrahamsson N, Davies M, Hesse D, Greenway FL, Jensen C, Lingvay I, Mosenzon O, Rosenstock J, Rubio MA, Rudofsky G, Tadayon S, Wadden TA, Dicker D; STEP 4 Investigators (2021) Effect of continued weekly subcutaneous semaglutide vs placebo on weight loss maintenance in adults with overweight or obesity: the STEP 4 randomized clinical trial. JAMA 325:1414-1425. doi: 10.1001/jama.2021.3224
    Pubmed KoreaMed CrossRef
  19. Sarker A, Ginn R, Nikfarjam A, O'Connor K, Smith K, Jayaraman S, Upadhaya T, Gonzalez G (2015) Utilizing social media data for pharmacovigilance: a review. J Biomed Inform 54:202-212. doi: 10.1016/j.jbi.2015.02.004
    Pubmed KoreaMed CrossRef
  20. Shin H, Cha J, Lee Y, Kim JY, Lee S (2021) Real-world data-based adverse drug reactions detection from the Korea Adverse Event Reporting System databases with electronic health records-based detection algorithm. Health Informatics J 27:14604582211033014. doi: 10.1177/14604582211033014
    Pubmed CrossRef
  21. Singh G, Krauthamer M, Bjalme-Evans M (2022) Wegovy (semaglutide): a new weight loss drug for chronic weight management. J Investig Med 70:5-13. doi: 10.1136/jim-2021-001952
    Pubmed KoreaMed CrossRef
  22. U.S. Food and Drug Administration (2022) SAXENDA® [package insert]. U.S. Food and Drug Administration. https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/206321s015lbl.pdf Accessed 30 December 2022.
  23. Wilding JPH, Batterham RL, Calanna S, Davies M, Van Gaal LF, Lingvay I, McGowan BM, Rosenstock J, Tran MTD, Wadden TA, Wharton S, Yokote K, Zeuthen N, Kushner RF; STEP 1 Study Group (2021) Once-weekly semaglutide in adults with overweight or obesity. N Engl J Med 384:989-1002. doi: 10.1056/NEJMoa2032183
    Pubmed CrossRef
  24. Yi H, Lee JH, Shin JY (2019) Signal detection for cardiovascular adverse events of DPP-4 inhibitors using the Korea adverse event reporting system database, 2008-2016. Yonsei Med J 60:200-207. doi: 10.3349/ymj.2019.60.2.200
    Pubmed KoreaMed CrossRef

Stats or Metrics

Share this article on :

  • line