Gastrointestinal toxicities associated with immune checkpoint inhibitors: a disproportionality analysis leveraging VigiBase, the WHO Adverse Drug Reaction Database.

Many clinical oncologists and gastroenterologists are unfamiliar with gastrointestinal irAEs caused by ICI treatment, making it impossible to detect and deal with the resulting immunotoxicity in time. In addition, the previous scientific literature lacks comprehensive and detailed reports on gastrointestinal irAEs caused by ICIs, especially on cases of megacolon. In this study, we used the World Health Organization (WHO) global individual case safety report database, known as VigiBase (herein also referred to as Adverse Drug Reaction Database), which is managed by the Uppsala Monitoring Center in Sweden, to fully understand the gastrointestinal irAEs caused by ICI treatment, and their clinical characteristics.

The present study is a disproportionality analysis leveraging VigiBase, which includes reports from more than 130 countries, and represents over 90% of the world's population. It contains more than 19 million individual case safety reports (ICSRs) submitted by national pharmacovigilance centers since 1967. These reports are generally notified post-marketing and originate from many kinds of sources, including pharmacists, healthcare professionals, patients, and physicians. Adverse drug reactions (ADRs) in VigiBase are coded according to the Medical Dictionary for Regulatory Activities (MedDRA).

In the present observational, retrospective, and disproportionate analysis, based on the VigiBase database, we report suspicious gastrointestinal irAEs related to ICI therapy from January 1, 2011 to March 6, 2019, and describe the clinical features of cases in detail. The reference group consists of all the adverse reaction reports in VigiBase during the same study period. The main gastrointestinal irAEs are divided into 13 categories, namely enterocolitis, gastrointestinal perforation, megacolon, duodenitis, intestinal obstruction, esophagitis, diarrhea, colon operation, gastritis, inflammatory bowel disease, gastrointestinal hemorrhage, abdominal discomfort, and vomiting (Table 1).

Table 1

Reports of ICI-associated gastrointestinal irAEs versus those included in the full database of VigiBase (from January 1, 2011 to March 6, 2019)

Gastrointestinal irAE	ICSRs for ICIs (ICIs reports in total: 61 931)	ICSRs in full database (reports in total: 13 300 773)	Information component (95% CI)
Enterocolitis	3066 (4.82%)	23 850 (0.18%)	4.80 (4.75‒4.85)
Gastrointestinal perforation	380 (0.40%)	12 983 (0.10%)	2.68 (2.51‒2.80)
Megacolon	15 (0.01%)	535 (0.004%)	2.46 (1.62‒3.05)
Duodenitis	44 (0.06%)	4358 (0.03%)	1.19 (0.71‒1.54)
Intestinal obstruction	259 (0.35%)	25 760 (0.19%)	1.19 (0.99‒1.33)
Esophagitis	62 (0.10%)	7344 (0.06%)	0.85 (0.43‒1.15)
Diarrhea	3246 (5.24%)	416 055 (3.13%)	0.74 (0.69‒0.79)
Colon operation	22 (0.04%)	2858 (0.02%)	0.70 (-0.01‒1.21)
Gastritis	166 (0.27%)	39 951 (0.30%)	-0.16 (-0.42‒0.02)
Inflammatory bowel disease	122 (0.20%)	42 884 (0.32%)	-0.71 (-1.01‒0.49)
Gastrointestinal hemorrhage	365 (0.59%)	131 244 (0.99%)	-0.74 (-0.92‒0.62)
Abdominal discomfort	1136 (1.83%)	520 487 (3.91%)	-1.09 (-1.19‒1.02)
Vomiting	2616 (4.22%)	1 248 590 (9.39%)	-1.15 (-1.22‒1.11)

ICSRs are described as number (percentage). The information component (IC) reflects the strength of association between the specific treatment and the adverse drug event. Positive IC values are thresholds of signal detection within VigiBase. Different ICIs include ipilimumab, nivolumab, pembrolizumab, atezolizumab, durvalumab, and combination therapy (ipilimumab plus nivolumab, ipilimumab plus pembrolizumab, ipilimumab plus atezolizumab, pembrolizumab plus nivolumab, ipilimumab plus pembrolizumab plus nivolumab). All gastrointestinal irAEs were classified by group queries according to the Medical Dictionary for Regulatory Activities (MedDRA, Version 20.1). ICIs: immune checkpoint inhibitors; irAE: immune-related adverse event; ICSRs: individual case safety reports; CI: confidence interval.

Each adverse reaction report associated with ICI treatment includes patient characteristics (age and sex), management information (country, reporting date, and notifier), drug information (WHO drug name, indication, route of administration, dosage regimen, start and end dates), and ADR information (MedDRA classification terms, onset of ADRs, and final outcomes). The five ICI drugs reported in the study are anti-PD-L1 antibodies (atezolizumab and durvalumab), anti-CTLA-4 antibody (ipilimumab), and anti-PD-1 antibodies (nivolumab and pembrolizumab). A severe adverse event is defined as an event leading to any other medically important conditions that require hospitalization (initial or prolonged), is life-threatening, or causes death. It is worth noting that reports of ADRs were used for the purposes of this study starting from January 1, 2011.

VigiBase is known as a case‒non-case analytic database; therefore, it was used for disproportionality analysis to assess whether suspected drug-induced gastrointestinal events were differentially reported between ICIs and other drugs in the full database. Gastrointestinal irAEs among different ICI regimens were also compared by disproportionality analyses, i.e., anti-PD-1 or anti-PD-L1 versus anti-CTLA-4 monotherapy, and monotherapy versus combination ICI therapy (anti-PD-1 or anti-PD-L1 plus anti-CTLA-4 combination therapy). The analyses aimed to identify adverse drug events by relative frequency, which is different from the overall reporting of events in the full database. The proportion of selected ADRs reported for ICIs was compared to the proportion in the full database.

Disproportionality analysis can be calculated by the information component (IC) or the reported odds ratio (ROR). The earlier can provide a conservative association metric, and reduce the risk of highlighting false associations compared to the ROR, especially for events with very low expected frequencies in the VigiBase database. Values of IC are only used to compare adverse drug events with the full database, but cannot be used for comparisons of different ICI reporting events among ICI therapies. Therefore, the IC value was used to identify meaningful ICIs that induced irAEs, while the ROR was used to compare the reporting events of ADRs in different ICI subgroups.

The value of IC, whose calculation is derived from a Bayesian confidence propagation neural network, is used to reflect the strength of the adverse drug event association (Zhao et al., 2001), which is expressed by the logarithm of prior and posterior probabilities of occurrence of ADRs. A positive IC value indicates that adverse drug events are reported more often than all reported events in the full database. The 95% confidence interval (CI) indicates the stability of IC values; the stability is higher when the 95% CI is narrower. Disproportionality for gastrointestinal irAEs induced by ICI subgroups was analyzed by calculating the ROR (95% CI). When the lower value of ROR (95% CI) was larger than one, adverse reactions for the target ICI were notable ADR signals in the adverse drug event. The Bonferroni adjustment method was used to perform the comparisons among different drug treatment regimens, and the ROR CI was adjusted by using P=0.01. The clinical characteristics of ADRs were summarized by descriptive statistical methods, while the quantitative variables were compared by using one-way analysis of variance (ANOVA).

According to the VigiBase database, patients who received ICIs reported 61 931 adverse events, whereas patients who were administered any drug reported 13 300 773 adverse events between January 1, 2011 and March 6, 2019. Table 1 shows the IC values for various gastrointestinal adverse events induced by ICI therapy compared to those for the same adverse event in the full database: ICI-associated enterocolitis (3066 reports for ICIs vs. 23 850 reports for the full database; IC 4.80 (95% CI 4.75‒4.85)); ICI-associated gastrointestinal perforation (380 reports for ICIs vs. 12 983 reports for the full database; IC 2.68 (95% CI 2.51‍‒‍2.80)); ICI-associated duodenitis (44 reports for ICIs vs. 4358 reports for the full database; IC 1.19 (95% CI 0.71‍‒‍1.54)); ICI-associated intestinal obstruction (259 reports for ICIs vs. 25 760 reports for the full database; IC 1.19 (95% CI 0.99‍‒‍‍1.33)). It is worth noting that 15 cases of the rarely reported ICI-associated megacolon were found (535 reports for the full database; IC 2.46 (95% CI 1.62‒3.05)). All IC values were positive for the above five gastrointestinal irAEs induced by ICIs, and the values of 95% CI were very narrow, which shows that these adverse events are markedly associated with ICI therapies. Thus, we focus on these five gastrointestinal irAEs to analyze differences between various ICI subgroups, and to explain the clinical characteristics of each gastrointestinal irAE.

As demonstrated in Table 2, the ROR values of the five gastrointestinal irAE were markedly different between ICI therapies, and the ROR values of duodenitis, enterocolitis, and gastrointestinal perforation were the highest in ICI combination treatments (ROR 11.68 (95% CI 7.43‍‒‍18.34), ROR 93.09 (95% CI 85.85‍‒‍100.94), and ROR 5.77 (95% CI 3.98‍‒‍8.37), respectively). We also found that the three gastrointestinal irAEs of duodenitis, enterocolitis, and intestinal obstruction were overreported for patients subjected to ICI combination treatment, versus those who received ICI monotherapy: duodenitis (ROR 8.43 (95% CI 4.70‍‒‍15.10)), enterocolitis (ROR 3.98 (95% CI 3.64‍‒‍4.35)), and intestinal obstruction (ROR 1.94 (95% CI 1.37‍‒‍2.75)). Enterocolitis, gastrointestinal perforation, and megacolon were overreported for patients who received anti-CTLA-4 treatment, versus those subjected to anti-PD-1/PD-L1 treatment (ROR 3.87 (95% CI 3.57‒4.20), ROR 5.42 (95% CI 4.37‍‍‒‍6.73), and ROR 4.92 (95% CI 1.75‒13.82), respectively).

Table 2

Major ICI-associated gastrointestinal ADRs compared to the full database of VigiBase (from January 1, 2011 to March 6, 2019)

Characteristics	Duodenitis			Enterocolitis			Gastrointestinal perforation			Intestinal obstruction			Megacolon
	n/n	ROR (95% CI)		n/n	ROR(95% CI)		n/n	ROR (95% CI)		n/n	ROR (95% CI)		n/n	ROR (95% CI)
Anti-CTLA-4 vs. full database	2/4 358	0.42 (0.11‒1.68)		1239/23 850	56.67 (53.42‒60.11)		214/12 983	15.97 (13.96‒18.27)		46/25 760	1.75 (1.32‒2.32)		9/535	15.71 (8.13‒30.36)
Anti-PD-1/Anti-PD-L1 vs. full database	25/4358	1.68 (1.14‒2.47)		1143/23 850	14.52 (13.67‒15.41)		140/12 983	2.92 (2.46‒3.46)		182/25 760	2.06 (1.78‒2.38)		5/535	3.17 (1.42‒7.08)
ICI combination vs. full database	17/4358	11.68 (7.43‒18.34)		684/23 850	93.09 (85.85‍‒‍100.94)		26/12 983	5.77 (3.98‒8.37)		31/25 760	3.84 (2.78‒5.31)		1/535	4.98 (0.70‒35.40)
Anti-CTLA-4 vs. Anti-PD-1/Anti-PD-L1	2/25	0.25 (0.06‒1.06)		1 239/1 143	3.87 (3.57‒4.20)		214/140	5.42 (4.37‒6.73)		46/182	0.85 (0.62‒1.17)		9/5	4.92 (1.75‒13.82)
ICI combination vs. monotherapy	17/ 27	8.43 (4.70‒15.10)		684/2382	3.98 (3.64‒4.35)		26/354	0.98 (0.67‒1.44)		31/228	1.94 (1.37‒2.75)		1/14	0.83 (0.11‒6.25)

Anti-CTLA-4 refers to any ICSR treated with ipilimumab. Anti-PD-1/PD-L1 refers to any ICSR treated with any of the following four drugs: nivolumab, pembrolizumab, atezolizumab, and durvalumab. ICI combination refers to any ICSR treated with anti-CTLA-4 (ipilimumab) in combination with any one or two of the anti-PD-1/PD-L1, or to any ICSR treated with anti-PD-1 in combination with anti-PD-L1. Comparisons between different drug regimens were performed using the Bonferroni adjustment method. ROR confidence intervals were calculated using P=0.01 to adjust. ROR: reported odds ratio； CI: confidence interval; ICI: immune checkpoint inhibitor； ADR: adverse drug reaction； Anti-CTLA-4: anti-cytotoxic T lymphocyte-associated antigen-4; ICSR: individual case safety report; Anti-PD-1: anti-programmed cell death protein 1; Anti-PD-L1: anti-programmed death-ligand 1.

Events of ICI-associated enterocolitis, gastrointestinal perforation, intestinal obstruction, and duodenitis were reported mainly from the Americas (1292/3066 (42.14%), 193/379 (50.92%), 138/259 (53.28%), and 19/44 (43.18%), respectively), while most of the megacolon cases were recorded in Europe (5/15 (33.33%)), as shown in Table 3. There were more male patients than female ones for the five gastrointestinal irAEs except for megacolon (enterocolitis 60.88% vs. 39.12%; gastrointestinal perforation 67.24% vs. 32.76%; intestinal obstruction 60.40% vs. 39.60%; duodenitis 62.79% vs. 37.21%; megacolon 40.00% vs. 60.00%). We also found that reported events of enterocolitis, duodenitis, gastrointestinal perforation, megacolon, and intestinal obstruction, associated with ICI treatment, were observed mainly in patients with lung cancer or malignant melanoma.

Table 3

Basic information of ICI-associated gastrointestinal ADR cases reported to VigiBase (from January 1, 2011 to March 6, 2019)

Basic information	Enterocolitis （n=3066）	Gastrointestinal perforation （n=379）	Intestinal obstruction （n=259）	Duodenitis （n=44）	Megacolon （n=15）
Year
2011	23 (0.75%)	8 (2.11%)	4 (1.54%)	0	0
2012	84 (2.74%)	28 (7.39%)	9 (3.47%)	0	3 (20.00%)
2013	103 (3.36%)	20 (5.28%)	2 (0.77%)	0	2 (13.33%)
2014	233 (7.60%)	49 (12.93%)	10 (3.86%)	2 (4.55%)	0
2015	369 (12.04%)	79 (20.84%)	28 (10.81%)	5 (11.36%)	3 (20.00%)
2016	425 (13.86%)	45 (11.87%)	30 (11.58%)	7 (15.91%)	1 (6.67%)
2017	754 (24.59%)	55 (14.51%)	62 (23.94%)	10 (22.73%)	6 (40.00%)
2018	934 (30.46%)	88 (23.22%)	98 (37.84%)	15 (34.09%)	0
2019	141 (4.60%)	7 (1.85%)	16 (6.18%)	5 (11.36%)	0
Gender
Female	1080/2761 (39.12%)	115/351 (32.76%)	99/250 (39.60%)	16/43 (37.21%)	9/15 (60.00%)
Male	1681/2761 (60.88%)	236/351 (67.24%)	151/250 (60.40%)	27/43 (62.79%)	6/15 (40.00%)
Age （year）
0‒17	3/2173 (0.14%)	0	0	0	0
18‒44	218/2173 (10.03%)	16/290 (5.52%)	14/205 (6.83%)	5/36 (13.89%)	1/13 (7.69%)
45‒65	870/2173 (40.04%)	117/290 (40.34%)	109/205 (53.17%)	16/36 (44.44%)	6/13 (46.15%)
>65	1082/2173 (49.79%)	157/290 (54.14%)	82/205 (40.00%)	15/36 (41.67%)	6/13 (46.15%)
Region
Africa	3/3066 (0.10%)	0	1/259 (0.39%)	0	0
Americas	1292/3066 (42.14%)	193/379 (50.92%)	138/259 (53.28%)	19/44 (43.18%)	4/15 (26.67%)
Asia	403/3066 (13.14%)	65/379 (17.15%)	61/259 (23.55%)	9/44 (20.45%)	4/15 (26.67%)
Europe	1198/3066 (39.07%)	109/379 (28.76%)	51/259 (19.69%)	15/44 (34.09%)	5/15 (33.33%)
Oceania	170/3066 (5.54%)	12/379 (3.17%)	8/259 (3.09%)	1/44 (2.27%)	2/15 (13.33%)
Notifiera
Notifier1	2391/2912 (82.11%)	289/366 (78.96%)	205/248 (82.66%)	40/42 (95.24%)	9/14 (64.29%)
Notifier2	521/2912 (17.89%)	77/366 (21.04%)	43/248 (17.34%)	2/42 (4.76%)	5/14 (35.71%)
Indicationb
Malignant melanoma	1533/2350 (65.23%)	168/277 (60.65%)	67/180 (37.22%)	14/34 (41.18%)	8/10 (80.00%)
Lung cancer	608/2350 (25.87%)	78/277 (28.16%)	53/180 (29.44%)	11/34 (32.35%)	2/10 (20.00%)
Kidney cancer	83/2350 (3.53%)	4/277 (1.44%)	9/180 (5.00%)	4/34 (11.76%)	0
Bladder cancer	44/2350 (1.87%)	10/277 (3.61%)	29/180 (16.11%)	1/34 (2.94%)	0
Gastric cancer	12/2350 (0.51%)	8/277 (2.89%)	8/180 (4.44%)	0	0
Prostate cancer	20/2350 (0.85%)	3/277 (1.08%)	2/180 (1.11%)	0	0
Colon cancer	13/2350 (0.55%)	1/277 (0.36%)	8/180 (4.44%)	2/34 (5.88%)	0

Data are expressed as number (percentage). a Notifier1 refers to physician, pharmacist, or other healthcare professional; Notifier2 refers to lawyer or consumer. b Indication refers to the primary tumors treated by ICIs. ICI: immune checkpoint inhibitor； ADR: adverse drug reaction.

As shown in Table 4, reports of enterocolitis (1239 cases, 40.41%) and gastrointestinal perforation (214 cases, 56.32%) were the two most frequent ipilimumab-associated ADRs, while the most of intestinal obstruction cases (86 cases, 33.20%) were nivolumab-associated, and of duodenitis cases (16 cases, 36.36%) were ipilimumab plus nivolumab combination-associated. Megacolon was reported mainly under ipilimumab treatment (9 cases, 60.00%). In addition, there were a small number of reports of anti-PD-L1-associated gastrointestinal irAEs (87 cases of atezolizumab and 43 cases of durvalumab). We conducted a further analysis of follow-up reports on ICI-associated gastrointestinal irAEs. It was found that 1713 cases had the final outcomes reported, including "fatal," "not recovered/not resolved," and "recovered/resolved." Most patients in the enterocolitis, intestinal obstruction, and duodenitis cases recovered or resolved after appropriate treatment, while most of the gastrointestinal perforation cases (46.61%) had fatal consequences. The number of recovery reports of ICI-associated enterocolitis (71.70%) was higher than that for ICI-associated gastrointestinal perforation, intestinal obstruction, duodenitis, and megacolon combined. Four megacolon cases reported final outcomes: the patient recovered in one case and did not recover in another case, while two cases had fatal consequences.

Table 4

Comparison of ICI regimen for ICI-associated digestive system ADRs reported to VigiBase (from January 1, 2011 to March 6, 2019)

ICI regimen/outcome	Enterocolitis (n=3066)	Gastrointestinalperforation (n=380)	Intestinalobstruction (n=259)	Duodenitis (n=44)	Megacolon (n=15)
Drug group
Ipilimumab	1239 (40.41%)	214 (56.32%)	46 (17.76%)	2 (4.55%)	9 (60.00%)
Nivolumab	660 (21.53%)	67 (17.63%)	86 (33.20%)	11 (25.00%)	5 (33.33%)
Pembrolizumab	406 (13.24%)	60 (15.79%)	58 (22.39%)	12 (27.27%)	0
Atezolizumab	46 (1.50%)	10 (2.63%)	30 (11.58%)	1 (2.27%)	0
Durvalumab	31 (1.01%)	3 (0.79%)	8 (3.09%)	1 (2.27%)	0
Ipilimumab plus nivolumab	656 (21.40%)	25 (6.58)	29 (11.20%)	16 (36.36%)	1 (6.67%)
Ipilimumab plus pembrolizumab	19 (0.62%)	1 (0.26%)	1 (0.39%)	1 (2.27%)	0
Ipilimumab plus atezolizumab	1 (0.03%)	0	0	0	0
Pembrolizumab plus nivolumab	1 (0.03%)	0	0	0	0
Ipilimumab plus pembrolizumab plus nivolumab	7 (0.23%)	0	1 (0.39%)	0	0
Drug doses
Monotherapy with ipilimumab
3 mg/kg	392/742 (52.83%)	61/136 (44.85%)	7/26 (26.92%)	0	2/7 (28.57%)
10 mg/kg	120/742 (16.17%)	22/136 (16.18%)	9/26 (34.62%)	1/1 (100.00%)	1/7 (14.29%)
<500 mg	156/742 (21.02%)	43/136 (31.62%)	3/26 (11.54%)	0	4/7 (57.14%)
500‒1000 mg	47/742 (6.33%)	7/136 (5.15%)	4/26 (15.38%)	0	0
>1000 mg	27/742 (3.64%)	3/136 (2.21%)	3/26 (11.54%)	0	0
Monotherapy with nivolumab
1‒2 mg/kg	13/408 (3.19%)	5/53 (9.43%)	5/69 (7.25%)	0	0
≥3 mg/kg	246/408 (60.29%)	26/53 (49.06%)	38/69 (55.07%)	4/8 (50.00%)	0
≤240 mg	112/408 (27.45%)	19/53 (35.85%)	21/69 (30.43%)	4/8 (50.00%)	3/3 (100.00%)
240‒480 mg	37/408 (9.07%)	3/53 (5.66%)	5/69 (7.25%)	0	0
Monotherapy with pembrolizumab
2 mg/kg	61/247 (24.70%)	4/45 (8.89%)	7/44 (15.91%)	1/9 (11.11%)	0
≤200 mg	186/247 (75.30%)	41/45 (91.11%)	37/44 (84.09%)	8/9 (88.89%)	0
Combination therapy
Ipilimumab (3 mg/kg) plusnivolumab (1‒2 mg/kg）	150/367 (40.87%)	6/9 (66.67%)	8/17 (47.06%)	2/7 (28.57%)	0
Ipilimumab (3 mg/kg) plusnivolumab （≥3 mg/kg）	86/367 (23.43%)	2/9 (22.22%)	5/17 (29.41%)	2/7 (28.57%)	0
Ipilimumab (≤500 mg) plusnivolumab (≤240 mg)	124/367 (33.78%)	1/9 (11.11%)	3/17 (17.65%)	2/7 (28.57%)	0
Ipilimumab (≤500 mg) plusnivolumab (>240 mg)	7/367 (1.91%)	0	1/17 (5.88%)	1/7 (14.29%)	0
Outcome
Recovered/resolved	960/1339 (71.70%)	82/221 (37.10%)	73/134 (54.48%)	8/15 (53.33%)	1/4 (25.00%)
Not recovered/not resolved	252/1339 (18.82%)	36/221 (16.29%)	35/134 (26.12%)	7/15 (46.67%)	1/4 (25.00%)
Fatal	127/1339 (9.48%)	103/221 (46.61%)	26/134 (19.40%)	0	2/4 (50.00%)

Data are expressed as number (percentage). Drug doses are expressed in VigiBase in two forms: mg/kg and total dose. Drug doses are grouped according to dosage and administration recommended in the drug prescription information. Recommended dosage for nivolumab: 3 mg/kg every 2 weeks, 240 mg every 2 weeks, or 480 mg every 4 weeks for different cancers. Recommended dosage for pembrolizumab: 200 mg every 3 weeks, 2 mg/kg every 3 weeks for different cancers. Recommended dosage for ipilimumab: 3 mg/kg every 3 weeks or 10 mg/kg every 3 weeks for different cancers. Recommended dosage for combined ICI therapy: 3 mg/kg nivolumab followed by 1 mg/kg Ipilimumab on the same day every 3 weeks for 4 doses, and then 240 mg nivolumab every 2 weeks or 480 mg every 4 weeks. ICI: immune checkpoint inhibitor； ADR: adverse drug reaction.

Additionally, we analyzed time to onset for ICI-associated gastrointestinal irAEs. We found that the time to onset of enterocolitis associated with anti-CTLA-4 and ICI combination treatment was earlier compared to that associated with anti-PD-1/PD-L1 monotherapy (anti-CTLA-4 (55.44±3.25) d, anti-PD-1/PD-L1 (119.50±5.15) d, and combination therapy (58.49±4.64) d), and the time to onset of enterocolitis was not significantly different between anti-CTLA-4 and ICI combination treatments (Fig. 1a). As shown in Fig. 1b, the time to onset of gastrointestinal perforation was not significantly different between anti-CTLA-4 and anti-PD-1/PD-L1 monotherapy (anti-CTLA-4 (58.08±6.45) d and anti-PD-1/PD-L1 (61.67±8.44) d, P=0.7998). Furthermore, the time to onset of intestinal obstruction, as shown in Fig. 1c, was not significantly different between anti-CTLA-4 and anti-PD-1/PD-L1 monotherapy (anti-CTLA-4 (45.40±9.89) d and anti-PD-1/PD-L1 (91.60±14.45) d, P=0.1760).

Fig. 1

Time to onset of ICI-associated gastrointestinal irAEs reported to VigiBase (from January 1, 2011 to March 6,2019). (a) ICI-associated enterocolitis (n=1145); (b) ICI-associated gastrointestinal perforation (n=113); (c) ICI-associated intestinal obstruction (n=86). Date are expressed as mean±standard deviation. ICI: immune checkpoint inhibitor； irAE: immune-related adverse event; CTLA-4: cytotoxic T lymphocyte-associated antigen-4; PD-1: programmed cell death protein 1； PD-L1: programmed death-ligand 1.

Early clinical outcomes with ICI therapies have indicated that ICIs can improve patient survival; therefore, such therapies have been widely used to treat many malignant tumors, such as malignant melanoma (Wolchok et al., 2017; Afzal et al., 2018), renal cell carcinoma (Derosa et al., 2018), lung cancer (Doroshow et al., 2019; Madden and Kasler, 2019), and urothelial cancer (Rodriguez-Vida et al., 2018). Meanwhile, ICIs can also disturb the maintenance of patient immune homeostasis, as they suppress coinhibitory molecules between T-cells and other immune cells, which mediate the tumor killing process (Sangro et al., 2020). In addition, ICIs have side effects and toxicities in the process of tumor therapy, commonly referred to as irAEs. Therapies involving ICIs, such as anti-PD-1, anti-PD-L1, or anti-CTLA-4 antibodies, have been proved to effectively treat metastatic malignant melanoma and lung cancer (Afzal et al., 2018; Madden and Kasler, 2019). However, they cause gastrointestinal irAEs, including enteritis, enterocolitis, microscopic colitis, and gastritis (Tandon et al., 2018; de Malet et al., 2019). Additionally, certain severe or fatal irAEs were indicated in a previous study, including gastrointestinal perforation and intestinal obstruction associated with ICI therapy (Abu-Sbeih and Wang, 2020).

Thus far, the present study includes the largest and most extensive database of the clinical characteristics of ICI-associated gastrointestinal irAEs, acquired by the analysis of individual case safety reports from the WHO pharmacovigilance database. Consistently with previous studies, our results indicate that ICI-associated enterocolitis, duodenitis, gastrointestinal perforation, and intestinal obstruction are dramatically increased, and the associated risk posed by ICI combination treatment may be higher than that by ICI monotherapy. In addition, 15 cases of the rarely reported ICI-associated megacolon are included herein, which is infrequent in the clinical setting.

Previous research has shown that gastrointestinal irAEs are closely related to anti-CTLA-4 or anti-PD-1/PD-L1 therapy (Xu et al., 2018). It was observed in the present study that, compared to anti-PD-1/PD-L1 therapy, patients receiving anti-CTLA-4 therapy may have a higher risk of developing enterocolitis (ROR 3.87 (95% CI 3.57‍‒‍4.20)), gastrointestinal perforation (ROR 5.42 (95% CI 4.37‍‒‍6.73)), and megacolon (ROR 4.92 (95% CI 1.75‍‒‍13.82)). The combined ICI therapy of anti-CTLA-4 and anti-PD-1/PD-L1 has been suggested as highly efficient in treating malignant tumors (Yuan et al., 2019). Combined ICI therapy, however, may result in more severe immune-related ADRs (Gu et al., 2019). Our study indicates that, in agreement with previous research, patients receiving combined ICI therapy may have a higher risk of developing gastrointestinal irAEs compared to ICI monotherapy, including duodenitis (ROR 8.43 (95% CI 4.70‍‒‍15.10)), enterocolitis (ROR 3.98 (95% CI 3.64‍‒‍4.35)), and intestinal obstruction (ROR 1.94 (95% CI 1.37‍‒‍2.75)).

In the current study, it is demonstrated that most of the gastrointestinal irAEs occurred in patients with melanoma and lung cancer, with more cases in males than females, except for megacolon. Most of the gastrointestinal irAEs occurred within six weeks, and immune-related enterocolitis in ICI combination therapy manifested earlier than those in anti-PD-1/PD-L1 treatment, which was consistent with previous research (Wang et al., 2018). Additionally, it is indicated that anti-CTLA-4 therapy had a dramatically earlier time to onset of enterocolitis than anti-PD-1/PD-L1 therapy, but no significant difference between the onset time of gastrointestinal perforation and intestinal obstruction was observed.

We also analyzed cases of ICI-associated megacolon, which were very rare in the clinical setting, and had only been recorded as case reports in previous research (Cho and Choi, 2020). Fifteen megacolon cases are featured herein, most of which involved patients with lung cancer or malignant melanoma, and occurred in anti-CTLA-4 therapy (ipilimumab). Interestingly, ICI-associated megacolon was mainly reported from Europe but not Americas. The possible reason is that megacolon is commonly named grade 4 colitis in Americas, and has been described similarly in previous research (Wang et al., 2017). Clinically toxic megacolon requires rapid multidisciplinary assessment, and individual treatments need to be developed based on the underlying causes (Leppkes et al., 2015). If the cause is likely an infection or inflammation, perioperative morbidity and mortality can be reduced by conservative treatment. The effect of treatment needs to be re-evaluated after 72 h and after 7 d to further ensure patient safety. Thus, the toxic megacolon adverse event needs to be closely monitored during treatment by ICIs, especially by anti-CTLA-4. The risk factors of ICI-related megacolon are not yet clear, and hence they need to be investigated in future clinical trials. It was observed in our study that most patients of the enterocolitis, intestinal obstruction, and duodenitis cases were recovered or resolved after precise treatment, while most of the gastrointestinal perforation cases had fatal consequences. The number of recovery reports for ICI-associated enterocolitis was found to be larger compared to ICI-associated gastrointestinal perforation, intestinal obstruction, duodenitis, and megacolon combined. Therefore, ICI-related enterocolitis is treated conservatively with most patients capable of recovery, and there is usually no need to discontinue the drug (Zhang ML et al., 2020). Additionally, one of the most common drug side effects, esophagitis, was absent in gastrointestinal irAEs of our comprehensive assessment, demonstrating that the esophagus is likely not involved in gastrointestinal irAEs associated with ICIs therapy. Thus, it seems fortunate that the esophagus, an integral section of the gastrointestinal tract, appears to be spared in the procession of ICI therapy.

The VigiBase database has some limitations, which are also important to mention for the scope of this study. First, detailed information on clinical data, which may help to better evaluate patient responses related to irAEs, was missing from the VigiBase database. Second, when a report involved several drugs and/or several adverse events, we treated the adverse drug event pair as the basic unit rather than the reported event; the results of this pharmacovigilance analysis may thus be biased. Third, case reports of gastrointestinal disease associated with ICI therapies may not have been submitted to the VigiBase database, leading to a further bias in the database. Finally, the IC value does not indicate a causal relationship between ICIs and irAEs, but shows the related dependence of the ICI drugs and adverse event reports. In contrast to existing studies, however, the intensity of the large number of records at the national level supports our research aimed at quantifying the potential risks, but the true risks of these adverse events should be determined in prospective studies. Hence, a complete understanding of the limitations of VigiBase can help us better analyze ADRs, and produce useful clinical research hypotheses.

The present study shows that the number of gastrointestinal irAEs, including enterocolitis, gastrointestinal perforation, duodenitis, and intestinal obstruction cases, has been rising dramatically, with several reported events of megacolon, which is rare in the clinical setting. Most gastrointestinal irAEs occur early during the course of ICI therapy. Consequently, it is important to raise an awareness of gastrointestinal irAEs associated with treatment by the ICI therapy of some advanced malignant tumors.