Upper Gastrointestinal Perforations: A Possible Danger of Antibiotic Overuse

Abstract

Background

The role of changes in gut microflora on upper gastrointestinal (UGI) perforations is not known. We conducted a retrospective case-control study to examine the relationship between antibiotic exposure—a proxy for microbiome modulation— and UGI perforations in a national sample.

Methods

We queried a 5% random sample of Medicare (2009–2013) to identify patients ≥ 65 years old hospitalized with UGI (stomach or small intestine) perforations using International Classification of Diseases diagnosis codes. Cases with UGI perforations were matched with 4 controls, each based on age and sex. Exposure to outpatient antibiotics (0–30, 31–60, 61–90 days) prior to case patients’ index hospitalization admission data was determined with Part D claims. Univariate and multivariable regression analyses were performed to evaluate the effect of antibiotic exposure on UGI perforation.

Results

Overall, 504 cases and 2016 matched controls were identified. Compared to controls, more cases had antibiotic exposure 0–30 days (19% vs. 3%, p < 0.001) and 31–60 days (5% vs. 2%, p < 0.001) prior to admission. In adjusted analyses, antibiotic exposure 0–30 days prior to admission was associated with 6.8 increased odds of an UGI perforation (95% CI 4.8, 9.8); 31–60 days was associated with 1.9 increased odds (95% CI 1.1, 3.3); and 61–90 days was associated with 3.7 increased odds (95% CI 2.0, 6.9).

Conclusions

Recent outpatient antibiotic use, in particular in the preceding 30 days, is associated with UGI perforation among Medicare beneficiaries. Exposure to antibiotics, one of the most modifiable determinants of the microbiome, should be minimized in the outpatient setting.

Introduction

In the last decade, there has been tremendous growth in our understanding of the relationship between the trillions of bacteria hosted by humans in their gastrointestinal tracts and disease pathogenesis.^1,2 There is mounting evidence that gut microbiota affects disease incidence and outcomes, in particular in surgical disease.^3–6 Various studies have demonstrated that antibiotic exposure reduces microbiome diversity.^7,8 Although general implications of alteration of gut microbiota by antimicrobial treatment have been well described,^9–11 the relationship between antibiotic use and acute surgical disease remains to be elucidated.

Not only have alterations in gut microbiota due to antibiotic exposure been shown to be responsible for C. difficile infections,^10,12,13 but perturbations of microbial communities promote other pro-inflammatory responses, which may play a role in the pathogenesis of gastrointestinal perforations.¹⁴ Upper gastrointestinal (UGI) perforation, a devastating complication of a number of inflammatory conditions, most commonly peptic ulcer disease, is associated with high morbidity and mortality.^15–18 More than 70% of deaths associated with peptic ulcer disease are due to UGI perforations.¹⁶ Older persons are at greatest risk of dying after developing perforated peptic ulcer disease.^19–21

With the rampant overuse of antibiotics among older adults,²² there is a pressing need to better understand the adverse effects of antibiotics. Older adults use 50% more antibiotics per capita than younger adults; ^23–25 furthermore, older adults also have the highest risk of poor outcomes from the adverse effects of antibiotics.²⁶ Yet, little is known about the association between antibiotic use and pathogenesis of UGI perforations, one possible adverse effect of antibiotics, in the elderly. The aim of this study is to utilize a national sample of Medicare beneficiaries to examine the association between antibiotic exposure—a proxy for microbiome modulation— and UGI perforations.

Materials and Methods

Data Source

We queried a 5% random sample of Medicare claims, files (MEDPAR) 2009–2013 (N = 1,123,017), to identify patients ≥ 65 years old with UGI perforations (stomach or small intestine) using International Classification of Diseases diagnosis codes (Appendix 1 Table 3). All patients were greater than or equal to 65 years on January 1, 2009 and had 12 months of continuous Part D enrollment. Medicare denominator files provide beneficiary demographic data (age, sex, race). Medicare Provider and Analysis Review (MEDPAR) provides data on Part A inpatient claims (diagnosis and procedures). Inpatient claims, outpatient claims, and chronic conditions files were used to calculate each Elixhauser comorbidity index.²⁷ Part D outpatient prescription drug claims provided data on type of outpatient medication, formulation, dosage, strength, date of prescription, and date of prescription filled.

Case Identification and Control Group

For cases, inclusion criteria included emergent or urgent admission. Exclusion criteria included any skilled nursing facility or previous hospitalization in the last 90 days. For cases, patients with an UGI (stomach or small intestine) perforation were defined as any hospital admission with a primary diagnosis of UGI perforation documented in MEDPAR. For controls, we randomly selected four controls without UGI perforation for each case and matched on sex and age.

Determination of Exposure to Outpatient Oral Antibiotics

Using Part D claims, we classified previous antibiotic exposure, a proxy for microbiome modulation, as having received oral antibiotics within 30 days, 31–60 days, and 61–90 days prior to the data of the case’s admission for UGI perforation. Time frame of antibiotic exposure was based on the last date of antibiotic available to the patient based on date of dispensation, directions for medication usage, and total number of doses dispensed. Topical and ophthalmic antimicrobials were excluded due to the lack of systemic bioavailability. Antibiotic exposure was further classified into drug classes. If more than one oral antibiotic was prescribed, the date of exposure was determined by the antibiotic that was most recently consumed by the patient. For all cases, a priori adherence to prescribed antibiotic was assumed.

Statistical Analysis

Categorical data are presented as percentage frequencies, with continuous data presented as mean values ± standard deviation. Categorical data were analyzed using Rao-Scott chisquare test. Continuous data were analyzed using analysis of variance. The Elixhauser comorbidity software was used, as previously described, for univariate and multivariable analyses.²⁷ A multivariable logistic regression model was constructed for the primary outcome, UGI perforation. Previous antibiotic use (0–30 days, 31–60 days, 61–90 days) was the variable of interest. The model was adjusted for race, Elixhauser index, current proton pump inhibitor use, current nonsteroidal anti-inflammatory drug use, and current or recent history (within 3 months) of steroid use. Statistical significance was defined as p < 0.05. All statistical analyses were performed using SAS 9.4 statistical software (SAS Institute, Cary, NC). All data were de-identified by the originator prior to our acquisition. This study was deemed exempt by the University of Massachusetts Medical School Institutional Review Board and was approved by the Centers for Medicare and Medicaid Services via the Research Data Assistance Center.

Results

Overall, 504 cases were identified and matched with 2016 age- and sex- matched controls. The majority of the cases were elderly females who were non-Hispanic whites (Table 1). In addition, the majority of cases and controls did not have significant comorbidities (Elixhauser score = 0: Cases 68% vs. Controls 78%, p < 0.001). Compared to controls, cases had a higher use of proton pump inhibitors (PPI), nonsteroidal anti-inflammatory drugs (NSAID), and steroids.

Table 1.

Baseline characteristics and antibiotic exposure of cases and controls

	Upper gastrointestinal perforations (N = 504)	Controls (N = 2016)	p value
Female	340 (67.5)	1360 (67.5)	Matched
Age (years)			Matched
65–74	160 (31.7)	640 (31.7)
74–85	165 (32.7)	660 (32.7)
85–94	164 (32.5)	656 (32.5)
> 95	15 (3)	60 (3)
Race/ethnicity			0.002
Non-Hispanic white	444 (88.1)	1633 (81)
Non-Hispanic black	33 (6.5)	200 (9.9)
Hispanic	9 (1.8)	54 (2.7)
Other	18 (3.6)	129 (6.4)
Average Elixhauser score (SD)	1.86 (3.06)	1.18 (2.52)	< 0.001
Elixhauser score			< 0.001
Elixhauser index = 0	343 (68.1)	1564 (77.6)
Elixhauser index = 1	2 (0.4)	19 (0.9)
Elixhauser index = 2	12 (2.4)	48 (2.4)
Elixhauser index ≥ 3	147 (29.2)	385 (19.1)
Comorbidities
Rheumatoid arthritis/collagen vascular	281 (55.8)	775 (38.4)	< 0.001
Diabetes with and without chronic complications	157 (31.2)	550 (27.3)	0.08
Hypertension	126 (25)	365 (18.1)	< 0.001
Renal failure	120 (23.8)	264 (13.1)	< 0.001
Depression	125 (24.8)	300 (14.9)	< 0.001
Hypothyroidism	90 (17.9)	366 (18.2)	0.87
Solid tumor without metastasis	11 (2.2)	14 (0.7)	0.01
Deficiency anemia	39 (7.7)	128 (6.3)	0.26
Current proton pump inhibitor use	46 (9.1)	96 (4.8)	< 0.001
Current nonsteroidal anti-inflammatory drug use	49 (9.7)	17 (0.8)	< 0.001
Current or recent history (within 3 months) of steroid use	38 (7.5)	30 (1.5)	< 0.001
Antibiotic exposure prior to admission
0–30 days	96 (19.0)	59 (2.9)	< 0.001
31–60 days	24 (4.7)	48 (2.4)	< 0.001
61–90 days	22 (4.4)	25 (1.2)	0.45
No antibiotic use (0–90 days prior to admission)	362 (71.8)	1884 (93.4)	< 0.001

In addition, compared to controls, cases had a greater number of comorbidities using the Elixhauser index. Overall, 274 used antibiotics. Among all antibiotic users, 57% had antibiotic exposure 0–30 days prior to admission, 26% had antibiotic exposure 31–60 days prior to admission, and 17% had antibiotic exposure 61–90 days prior to admission. Compared to controls, more cases had antibiotic exposure 0–30 days (19% vs. 3%, p < 0.001) and 31–60 days (5% vs. 2%, p < 0.001) prior to admission. There was no significant difference between the number of cases and controls with antibiotic exposure 61–90 days prior to admission. Fewer cases had no antibiotic use 90 days prior to admission compared to controls (72% vs. 93%, p < 0.001).

In adjusted analyses, antibiotic exposure 0–30 days prior to admission was associated with 6.8 increased odds of developing an UGI perforation (Fig. 1); antibiotic exposure 31–60 days prior to admission was associated with 1.9 increased odds of developing an UGI perforation; and antibiotic exposure 61–90 days prior to admission was associated with 3.7 increased odds of developing an UGI perforation. Table 2 shows differences in UGI perforation based on antibiotic class. In sub-analyses, to determine the association between class of antibiotics and UGI perforation, we found no differences in any time period of antibiotic exposure. (Supplemental Figs. 1, 2, 3).

Fig. 1.

Odds of upper gastrointestinal perforations based on timing of outpatient antibiotic exposure (0–90 days prior to admission)

Table 2.

Specific antibiotics/antibiotic classes used among cases and controls between 0–90 days prior to admission

	0–30 days (N = 155 )			31–60 Days (N = 72)			61–90 Days (N = 47 )
	UGI perforations	Controls	p value	UGI perforations	Controls	p value	UGI perforations	Controls	p value
Any antibiotic use	96 (61.9)	59 (38.1)	< 0.001	24 (33.3)	48 (66.7)	< 0.001	22 (46.8)	25 (53.2)	0.45
Cephalosporins	16 (16.7)	8 (13.6)	0.60	7 (29.2)	11 (22.9)	0.56	7 (31.8)	1 (4)	0.02
Fluoroquinolones	29 (30.2)	15 (25.4)	0.52	7 (29.2)	10 (20.8)	0.44	2 (9.1)	4 (16)	0.67
Penicillins	13 (13.5)	14 (23.7)	0.10	2 (8.3)	10 (20.8)	0.31	5 (22.7)	4 (16)	0.71
Macrolides	13 (13.5)	19 (32.2)	0.005	3 (12.5)	8 (16.7)	0.74	5 (22.7)	9 (36)	0.32
Clindamycin	4 (4.2)	4 (6.8)	0.48	1 (4.2)	1 (2.1)	0.99	1 (4.5)	0 (0)	0.47
Bactrim	29 (30.2)	6 (10.2)	0.003	1 (4.2)	6 (12.5)	0.41	3 (13.6)	3 (12)	1.0
Tetracyclines	5 (5.2)	0	0.07	2 (8.3)	6 (12.5)	0.71	1 (4.5)	3 (12)	0.61
Other antibiotics	9 (9.4)	3 (5.1)	0.54	2 (8.3)	4 (8.3)	1.0	1 (4.5)	2 (8)	1.0

Discussion

Although antibiotics have revolutionized medical treatment since the twentieth century, there is abundant evidence of widespread overuse of antibiotics, in particular in the outpatient setting for non-microbial disease such as upper respiratory viral infections.^24,25 Meanwhile, despite the rise in medical management of peptic ulcer disease, the rates of hospitalization and emergency surgery for perforated peptic ulcer disease are increasing.^17,28,29 We hypothesized that perturbations in gut microbiota due to antibiotic exposure might result in a pro-perforation milieu. We found that exposure to outpatient antibiotics across all time frames was associated with increased odds of UGI perforation, with the most recent exposure to antibiotics (0–30 days prior to admission) associated with the greatest increase.

A number of studies have shed light on the pervasive alteration of taxonomic, genomic, and functional capacity of the human gut microbiota by antibiotics.^7,30 It is not fully understood if these effects of antibiotics are rapid or persistent. As few as 3 days of antibiotic treatment can result in sustained reduction in microbiota diversity.³¹ Although the short-term antibiotic effects of perturbation of a microbial community are better characterized, long-term consequences of antibiotic effects on human microbiota are less widely known.^7,32 Data show that repopulating the microbiome may occur in days to weeks; however, the perturbation of a microbial community may also last for years before return to pre-antibiotic treatment community.^33–36 In the present study, we found that the most recent antibiotic use (0–30 days) was associated with the greatest increased risk of UGI perforation, which is consistent with the possible explanation that the perturbations of the gut microbial community have not yet recovered from the recent antibiotic use. Interestingly, antibiotic use (31–60 days) was also associated with increased, although attenuated, odds of development of UGI perforation compared to use between 61 and 90 days. While it seems surprising that the relationship is non-linear as time since antibiotic exposure decreases, this may be due to unmeasured clinical variables within our administrative dataset. Nevertheless, our results show that antibiotics at any time in the previous 90 days increase the risk of perforation compared to no exposure.

Given that our study demonstrates an association, we can speculate a biologically plausible rationale for the relationship between antibiotic use and gastrointestinal perforations similar to cigarette smoking and gastrointestinal perforations through reactive oxygen species production. Excessive dosing of antibiotics leads to the loss of naturally occurring intestinal microbiota.³⁷ Imbalance of the intestinal microbiota induces intestinal inflammatory responses through the increase in reactive nitrogen species and reactive oxygen species.³⁷ It is known that cigarette smoking is a major risk factor for the development of ulcers. The mechanisms of smoking in inflammatory diseases such as ulcers includes alteration of mucosal cell proliferation, change of blood flow in the inflammatory site, the increase of viral or bacterial infection, and the dysfunction of the immune system in the gastrointestinal mucosa.³⁸ Specifically, smoking induces apoptosis in the mucosa of the gastrointestinal tract, and this apoptosis is mediated through reactive oxygen species. Therefore, we speculate that antibiotics may lead to the imbalance of intestinal microbiota, which subsequently may induce apoptosis in the mucosa via reactive oxygen species which may play a role in ulcerogenesis, similar to cigarette smoke.

To our knowledge, there is only one other study examining the association between antibiotic use and UGI perforations. This study from Hsu et al. ³⁹ conducted a case-control study using a national health insurance database in Taiwan. The researchers found that fluoroquinolone use (0–60 days prior to admission) was associated with a twofold increased risk of development of gastric perforations after controlling for age and gender, possibly via a mechanism of collagen degradation;³⁹ however, there was no association of fluoroquinolones use with the development of small or large intestine perforation after controlling for patient characteristics. Given the increased global consumption of fluoroquinolones from 4.75 billion doses to 7.81 billion doses from 2000 to 2010,⁴⁰ this finding is concerning. In our study, fluoroquinolones were not associated with UGI perforations (Supplemental Figs. 1, 2, 3). One explanation for our discrepant findings compared to Hsu et al. ³⁹ may be the small number of cases and controls using fluoroquinolones in the present study, or more endemic nature of peptic ulcer disease and H. pylori infection in Taiwan. Furthermore, given the differences in anatomy, microbial community, and indications for perforations, we would argue that combining small or large intestine perforations into one category may create inaccurate results.

The strength of the study is not only the large sample size among a national sample of Medicare beneficiaries but also the detailed classification of antibiotic use using the Part D Medicare claims data. However, similar to other pharmacoepidemiology studies, this study is subject to limitations. First, this study is limited by the assumption that each subject adhered to the prescribed antibiotic. Second, although we controlled for many risk factors for the development of gastrointestinal perforations, unmeasured confounders, such as alcohol use and tobacco use, may exist given this is an administrative claims database. Since we matched on age and sex and controlled for baseline health, using a comorbidity index as well as use of certain medications, we have attempted to equalize the groups and account for variables that may have confounded the relationship between antibiotic use and perforations. Third, we did not measure the dose-response of antibiotics nor any other medications. Fourth, although we controlled for prescription medications that could be associated with gastrointestinal perforations, over-the-counter proton pump inhibitors and nonsteroidal anti-inflammatory drugs are not accounted for. However, due to our matching algorithm, we suspect that these factors may be distributed between the groups. In addition, given known demographic characteristics of those who utilize Medicare Part D for prescription drug benefit,⁴¹ our findings might not be generalizable to all Medicare beneficiaries or older Americans who do not use Medicare.

Many of the adverse effects, such as ototoxicity and Stevens-Johnson syndrome, of antibiotics are well known;⁴² however, to our knowledge, we are the first study to examine the link between antibiotic use and UGI perforations in the USA. A possible explanation of the relationship may be related to perturbation of the gut microbial community by antibiotics, although further research is needed.

Conclusion

Recent outpatient antibiotic use is associated with an increased odds of an UGI perforation, with the most recent antibiotic use associated with the greatest odds. Exposure to antibiotics, one of the most modifiable determinants of microbiome, should be minimized in the outpatient setting and only prescribed when warranted. Minimizing exposure to antibiotics may ultimately reduce the morbidity and mortality of UGI perforations by reducing disease incidence.