Bevacizumab and Gastrointestinal Perforations

A Review From the FDA Adverse Event Reporting System (FAERS) Database

Thomas A. Wichelmann; Sufyan Abdulmujeeb; Eli D. Ehrenpreis


Aliment Pharmacol Ther. 2021;54(10):1290-1297. 

In This Article


US post-market medication surveillance began after World War II with reports of chloramphenicol-associated aplastic anaemia.[8] Today, the FAERS database is accessible to the public via the FDA's online portal and contains international, voluntarily reported adverse events from medications, devices, vaccines and nutraceuticals.[9] Pharmaceutical companies are required by law to report all ADRs to the FAERS database, and as such, information obtained through the FAERS database is vital for post-market medication surveillance. While the FDA cautions against using the database to define ADR incidence rates because of reporting limitations and case duplication, the database represents a large, easily identifiable and attainable collection of ADRs. The FAERS database thus represents a valuable resource to further knowledge and understanding of the blackbox warning of bevacizumab-induced GI perforations.

Our study identified over 1300 cases of bevacizumab-induced GI perforations making this the largest collective description on the interaction to date. We included only reports from healthcare professionals in our analysis and eliminated consumer reports assuming superior reporter accuracy and more complete description of the GI perforation. Nearly half of reported cases (45.5%) occurred in the Medicare population (over age 65) and 26.0% of cases occurred in patients over age 70. For the top three primary malignancies, NSCLC represented the oldest cohort of patients, followed by colorectal cancer and ovarian cancer (avg. 64.7 years of age, 62.7 years of age and 59.8 years of age respectively).

There are two likely hypotheses that support the link between advanced age of bevacizumab-induced GI perforations and the frequency of cases reported in the FAERS database. First, the incidence of bevacizumab treated cancers typically increase with age and a higher proportion of patients in the elderly demographic would be expected to receive bevacizumab therapy for treatment. Second, advanced age could predispose patients to perforations through weakened mucosal integrity. Advanced age is linked to structural changes in gastric mucosa. Tarnawski et al previously illustrated replacement of normal rat gastric mucosa with increased perivascular connective tissue in older rats utilising transmission electron micrographs.[10] It was hypothesised that these gastric mucosal changes could significantly impact nutrient and oxygen transport leading to localised hypoxia and accumulation of toxic metabolites.[10] Liu et al's research on human subjects showed increased vascular collagen and oxidative stress in the gastric mucosa of elderly patients (aged 60–85) in comparison to younger participants.[11] In addition to the normal ageing of gastric mucosa, other insults to gastrointestinal mucosal integrity, like peptic ulcer disease (PUD) and diverticulosis, occur more frequently as patients age.[3,12–14] Draelos et al reviewed 17 802 colonoscopies to assess for elevated risk of GI perforation.[15] Nine perforations were identified and all occurred in patients over the age of 65 (average age of 78.4 years old).[15] Intestinal mucosal health arguably begins to decline around this age and, while speculative, our research supports the notion that patients over the age of 65 could be at a higher risk for GI perforation as reflected by the higher frequency of reporting contained within the FAERS database.

Present literature on the specific sites of bevacizumab-induced GI perforations is limited and disconnected. Most studies still reference early clinical trials that describe relatively few collective perforations.[16,17] To the best of our knowledge, this is the first large collective description of specific locations where bevacizumab-induced GI perforations occur. When descriptive locations were used, the large intestine was the most commonly identified perforation location (385 cases, 55.2% of descriptive locations and 28.0% of total cases). There are several plausible explanations for this finding. First, diverticulosis is relatively common in western culture. Prior research has identified diverticular presence in approximately 60% of patients aged 60 and older in the US and it is estimated that 5%-10% of the US population has diverticulosis by the age of 40.[13,14] Prior research in the UK identified an overall diverticular perforation incidence rate of 2.66 per 100 000 person-years (1.79–2.95, 95% CI).[13] Diverticular disease has also been previously identified as a risk factor for GI perforations in association with bevacizumab therapy.[18] Badgwell et al's study on bevacizumab-induced GI perforations found 29% of patients (seven cases) had diverticulosis.[3] It is possible that transmural and vascular disruption involved with colonic diverticulosis could contribute to the higher rates of large intestine perforations seen in the US population in association with bevacizumab use. Second, the majority of cases cited colorectal cancer as the indication for bevacizumab use (691 cases, 50.3% of all cases). As described previously, one of the competing hypotheses behind the pathophysiology of bevacizumab-induced GI perforations is the local impact of tumour destruction on surrounding normal mucosa leading to relative mucosal disruption contributing to increased perforation risk.[2] This notion could be supported if reported cases of colorectal cancer patients experiencing GI perforations within the large intestine were seen more frequently in the FAERS database; however, only 28.4% of colorectal cancer-associated GI perforation cases occurred in the large intestine which was only slightly higher than that of NSCLC (25.8%). Similar rates of reported large intestine perforations in colorectal cancer and NSCLC suggest a mechanism other than local tumour destruction by bevacizumab may be in effect.

Ovarian cancer is a frequently fatal entity due to the high probability of metastatic involvement at the time of diagnosis.[19,20] Abu-Hejleh et al's study identified ovarian cancer as having an overall increased incidence of bevacizumab-induced GI perforation.[4] Ovarian cancer represented the second most common primary malignancy in our study, however, reported cases were shadowed in comparison to colorectal cancer. There are two plausible explanations for this finding. First, bevacizumab was not FDA approved for ovarian cancer until 2018, nearly 14 years after colorectal cancer treatment approval (2004) leading to disproportionate frequency of bevacizumab use for colorectal cancer in comparison to ovarian cancer and subsequently higher frequency of reported GI perforations.[1] Second, colorectal cancer is more prevalent in the US than ovarian cancer (approximately 149 500 new cases vs 21 410 new cases respectively in 2020) and, subsequently, it could be assumed that bevacizumab is more frequently used for colorectal cancer thus increasing the prevalence of GI perforations.[21] Analysis of the FAERS data revealed that 138 of the 197 cases of bevacizumab-induced GI perforations associated with use for ovarian cancer (70.0% of all ovarian cancer cases) occurred before US FDA approval (June 13, 2018). Since FDA approval, there have been 59 cases of bevacizumab-induced GI perforation associated with ovarian cancer reported to the FAERS database in contrast to 80 cases associated with colorectal cancer. While unable to estimate incidence in any patient group, it appears that cases of patients with ovarian cancer experiencing GI perforation are becoming more frequently reported in the FAERs database. The comparative frequency of GI perforation reporting between colorectal cancer and ovarian cancer cases in the FAERS database in recent years highlights the possible increased risk of perforation given the much higher prevalence of colorectal cancer in the US. Metastatic ovarian cancer most frequently impacts and spreads first to the fallopian tube, omentum and peritoneum.[18] It is plausible that metastatic ovarian cancer thus imposes the increased risk of bevacizumab-induced GI perforations as a result of metastatic mesenteric burden impacting intestinal mucosal integrity at the time of treatment; we hypothesise that this will become more apparent in FAERS data trends in the future.

In 2009, Van Cutsem et al published their phase III clinical trial findings of bevacizumab in combination with gemcitabine and erlotinib for the treatment of metastatic pancreatic cancer.[22] While the primary endpoint of overall survival was not met, the progression-free survival was significantly longer in the treatment group with bevacizumab.[22] While bevacizumab has not received FDA approval for use in pancreatic cancer, bevacizumab is utilised off-label in this group and has been associated infrequently with GI perforations. Badgwell et al identified nine cases of bevacizumab-induced GI perforation in which pancreatic cancer was the indication for use (37.5% of their total cases).[3] This is a substantial finding given that only 18 cases of bevacizumab-induced GI perforation associated with pancreatic cancer were reported to the FAERS database from 2004 to the present and made up just 1.3% of the total cases. We hypothesise that bevacizumab-induced GI perforations in association with pancreatic cancer are a much rarer entity than previously suggested.

A comprehensive literature review reveals no prior large retrospective case series focused on bevacizumab-induced GI perforations within the gastro-oesophageal region. Our study represents the first large-scale description of patients suffering a perforation in the gastro-oesophageal region after bevacizumab therapy and highlights some interesting differences in comparison to bevacizumab-induced GI perforations as a collective whole. First, gastric cancer and pancreatic cancer were more commonly reported in patients suffering from gastro-oesophageal perforations. 14.5% (nine cases) of patients undergoing treatment for gastric cancer with bevacizumab in comparison to 1.0% of the collective group; and, 6.5% (four cases) of patients treated for pancreatic cancer with bevacizumab in comparison to 1.3% of the collective group (see Figure 11). Local reduction of tumour or tumour necrosis by bevacizumab leading to disruption of mucosal integrity and predisposition to perforations near the location of the primary tumour could explain this finding.[2] Second, gastro-oesophageal perforations occurred more commonly in males (66.1% of cases in the gastro-oesophageal group vs 48.9% in the overall group of bevacizumab-induced GI perforations). Khan et al recently published a registry-based analysis of 1106 oesophageal perforations which identified a male predominance of 62%.[23] More research is needed to illuminate the potential aetiology behind this association. Overall mortality was lower amongst gastro-oesophageal perforation cases compared to the total group of GI perforations (29.0% vs 40.3%). Prior research by Khan et al determined that mortality from all causes of oesophageal perforations was 14% and 90-day mortality was 22%.[23] The location of gastro-oesophageal perforations inherently makes them more amenable to endoscopic repair and stenting which we hypothesise could positively impact mortality. Overall, given the lack of previously reported data/cases this subgroup warrants additional investigation.

The meta-analysis by Hapani et al reported a mortality rate of 21.7% (range 11.5%-37.0%, 95% CI) for bevacizumab-induced GI perforations and prior research on gastrointestinal perforations as a whole indicates an overall mortality rate of approximately 30%.[5,24] Over 40% of bevacizumab-induced GI perforation cases identified in the FAERS database were fatal (see Table 2). There are three plausible explanations for increased mortality reporting within the database. First, the age demographic of patients experiencing bevacizumab-induced GI perforations was predominantly over the age of 65 (over 45% of total cases) with 25% of total cases occurring in those over age 70. It can be assumed that with advanced age comes inevitable comorbidities leading to poorer prognosis and higher mortality rates. In addition, perforation precedes peritonitis and the outcome is likely dependent on the patient's ability to mount an immune response. Cancer patients have multiple insults to their immune system, including medication effect, decreased nutritional intake, and tumour-mediated pathways.[25] Thus, we hypothesise that higher mortality is expected in bevacizumab-induced GI perforations given the cohort's underlying comorbidities, advanced age and presumed blunted immune response. Another plausible explanation is that mortality could be over represented in the FAERS database as a direct result of voluntary reporting. We hypothesise that lethal outcomes associated with medications would be more likely to be reported than non-lethal adverse events and, subsequently, the perceived rate of mortality included within the FAERS database represents an overestimation.

Our study highlights several shortcomings of the FAERS database and areas for potential improvement. First, frequent duplication of reported cases made data analysis challenging and time-consuming. Roughly 50% of all reported cases had complete demographics/information necessary for this analysis and nearly 50% of reported cases lacked a descriptive site of perforation. Under-reporting of ADRs and specific details in the FAERS database is known to occur.[26] Undoubtedly this could impact data analysis and arguably limits the utility of the calculated reporting odds ratio. Additionally, media attention and the recent publication of an adverse drug reaction in the literature might affect reporting behaviors.[27] While a more robust, well-defined data source could confer more reliable information about bevacizumab-induced GI perforations, the large number of cases reported in this series is useful to shed further light on this important and frequently encountered ADR from bevacizumab. Lastly, although bevacizumab-induced GI perforations are an established ADR, it is possible that some cases of GI perforations contained within the FAERS database are not a result of bevacizumab therapy.

In conclusion, an evaluation of the FAERS database for bevacizumab-induced GI perforations identified over 1300 reported cases. This is the largest reported case series to date. The most frequent location of GI perforation was in the large intestine. Elderly patients with colorectal cancer made up the majority of cases and 40% of the cases were fatal. We also identified and described a rare subgroup of patients experiencing bevacizumab-induced gastro-oesophageal perforations. Additional research is needed to describe this cohort of patients in order to understand the nuances of bevacizumab-induced gastro-oesophageal perforations.