Association between antibiotic use and subsequent risk of prostate cancer: A retrospective cohort study in South Korea

Several studies suggest that antibiotic use may affect overall cancer incidence, but the association between antibiotics and prostate cancer is still unclear. This retrospective cohort study aimed to assess the association between antibiotics and the risk of prostate cancer.


INTRODUCTION
In 2019, cancer ranked as the first or second leading cause of death in 112 countries and third or fourth in another 23 countries, according to the World Health Organization (WHO). 1 Among all types of cancer, prostate cancer caused 375 304 deaths in 2020, ranking as the fifth leading cause of cancer death among the global male population. 2 In South Korea, the mortality rate of prostate cancer increased from 4.2 in 2000 to 5.9 in 2007, and 6.0 in 2013 (per 1 000 000, age-adjusted). 3 In recent years, most research on prostate cancer has concentrated on the genetic aspects primarily, since prostate cancer is known as a highly heritable disease. 4However, an epidemiological understanding, such as risk factors, is also crucial for the prevention or treatment of prostate cancer.
6][7][8] The effects of antibiotics on gut microbiota diversity reduction and alteration were revealed. 9Mounting evidence also suggests that dysbiosis of the gut microbiota is linked to the pathophysiology of both intestinal and extra-intestinal illnesses. 10o prevent prostate cancer, it is vital to manage the risk factors such as age, family history, obesity, smoking, alcohol consumption, benign prostatic hyperplasia, and prostatitis. 11,12uch risk factors are well known in the prevention and early diagnosis of prostate cancer, while there is limited research on the possible association between antibiotics and prostate cancer risk.Previously, a Canadian population-based casecontrol study suggested a duration-dependent relationship between antibiotic use up to 15 years in the past and prostate cancer risk. 7Contrarily, in a case-control study based on medical records, there was no support that antibiotics decrease the prostate cancer risk. 13In addition, only a few studies investigated this issue in the Asian population, and residual confounding effects such as comorbidities and medication use were not fully considered.
Considering conflicting results, we sought to determine the association between antibiotic use and prostate cancer risk by using the National Health Insurance Service (NHIS) database.

Data source
The study population was acquired from the Korean National Health Insurance Service (NHIS) database.NHIS provides all Korean citizens with health insurance and complementary healthcare services.The health screening examinations are available to citizens who are 40 years or older, and they include a questionnaire on health-related behaviors, measurements of their anthropometric characteristics, and laboratory tests for blood and urine.NHIS collects data on hospitalizations, screenings, diagnostic and treatment-related procedures, and pharmaceutical prescriptions related to health service exploitation for claims purposes.

Study population
Figure S1 represents the study sample selection procedure.Among 1 075 313 men who lived in seven metropolitan areas (Seoul, Incheon, Busan, Daegu, Daejeon, Gwangju, and Ulsan) in South Korea and aged 50 years or older, 24 and 42 892 participants were excluded, respectively, as they passed away or were diagnosed with cancer before the index date of January 1, 2007.Consequently, 1 032 397 individuals between January 1, 2007, and December 31, 2019, were included in the main study population.Among these participants, 22 824 individuals were diagnosed with prostate cancer during the follow-up period.

Key variables
The primary outcome was the first diagnosis of prostate cancer between 2007 and 2019.The prostate cancer diagnosis was defined with the International Classification of Diseases 10th revision (ICD-10) code of C61.
From 2002 to 2006, antibiotic prescriptions and classes were used as exposure variables.Classes of antibiotics were based on the claims database and the World Health Organization (WHO) Anatomical Therapeutic Chemicals (ATC) classification of drugs; classes include macrolides, penicillins, cephalosporins, fluoroquinolones, sulfonamides, lincosamides, tetracyclines, vancomycin, carbapenems, monobactams, and linezolid (Table S1).The study classified antibiotic prescription days into 0, 1-14, 15-89, 90-179, and 180 or more days.The number of antibiotic groups was classified into zero, one, two, three, and four or more for the main study population, and zero, one, two, and three or more for participants who underwent health examinations.Separate statistical analyses were executed utilizing different models to eliminate possible interactions between cumulative antibiotic prescription days and classes.In the research, antibiotic non-users or the lowest users of antibiotics were used as reference groups.The group with the smallest number of antibiotic users was utilized as a reference when adjusting for indications of antibiotic use to reduce the possibility of multicollinearity and address the potential carcinogenic implications from inflammation, which has been investigated in various clinical and epidemiological studies. 14,15he considered confounders included age (continuous, years), household income (categorical, upper half or lower half), Charlson comorbidity index (categorical, 0, 1, or ≥2), non-steroidal anti-inflammatory drug use (categorical, yes or no), aspirin use (categorical, yes or no), anti-diabetic drugs (categorical, yes or no), 5a-reductase inhibitor (5-ARI) use (categorical, yes or no), Alpha-blockers use (categorical, yes or no), prostatitis diagnosis (categorical, yes or no), benign prostatic hyperplasia diagnosis (categorical, yes or no), hypertension diagnosis (categorical, yes or no), and infectious diseases (categorical, yes or no).The infectious diseases included respiratory diseases, urinary tract infections, skin, soft tissue, bone, and joint infections, intra-abdominal infections, intestinal infectious diseases, and other infectious diseases (Table S2).Infectious diseases were adjusted for six variables, not one variable.
For health examination participants, covariates including smoking (categorical, never, past, or current smokers), alcohol intake (categorical, less than one, one-two, three-four, more than five times per week), exercise (categorical, zero or more than one time per week), body mass index (BMI) (continuous, kg/m 2 ), systolic blood pressure (continuous, mmHg), fasting serum glucose (continuous, mg/dL), and total cholesterol (continuous, mg/dL) were additionally adjusted.

Statistical analysis
To evaluate the adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) of prostate cancer affected by antibiotic use after adjusting for the covariates, multivariate Cox proportional hazards models were utilized.Antibiotic non-users were assessed as a reference group for analyzing prostate cancer risk according to cumulative prescription days.For the analysis that included the source of infection, individuals who utilized antibiotics were assessed as a reference group.To minimize reverse causality, we applied the wash-out period by excluding the individuals who were diagnosed with prostate cancer within the defined period, 3 and 5 years.We also shifted the index date from January 1, 2007, to January 1, 2009, by prolonging antibiotic exposure in the sensitivity analysis.
Data mining and statistical analyses were executed utilizing SAS 9.4 (SAS Institute, Cary, NC, USA).On a two-sided basis, a p-value of 0.05 was considered statistically significant.

RESULTS
Among 1 032 397 participants of the main study population (Figure S1), 512 650 individuals were antibiotics non-users.Subgroups for the cumulative antibiotic prescription duration (1-14, 15-89, 90-179, and ≥180 days) included 403 054, 112 010, 3451, and 1232 participants, respectively.Tables S3  and S4 show the descriptive qualities of the main study population and participants who underwent health examinations, respectively.Participants who were prescribed 180 or more days of antibiotics tended to be older and have low-income status than those who were not.Health examination participants with 180 or more days of antibiotics subscription tended to be older, never smokers, and have low-income status than antibiotics non-users.
The relationship between the accumulative days of antibiotics prescription and prostate cancer was depicted in Table 1 (main study population) and Table S5 (participants who underwent health examinations).A duration-response relationship between prostate cancer and antibiotic use was observed.Within the main study population, compared with antibiotic non-users, antibiotic users for 180 or more days had a higher prostate cancer risk (aHR, 1.46; 95% CI, 1.11-1.91) in a fully adjusted model (Model 3).All models (Models 1, 2, and 3) revealed a higher prostate cancer risk with the constant use of antibiotics.In addition, the durationresponse relationship was maintained (p for trend = 0.004) in a fully modified model including infectious diseases (Model 4) with antibiotics 1-14 days user group as a reference.Individuals with 180 or more days of antibiotic exploitation had a higher prostate cancer risk (aHR, 1.36; 95% CI, 1.03-1.78)than those who prescribed antibiotics 1-14 days.An overall trend was maintained among participants who underwent health examinations.
The prostate cancer risk in proportion to the number of antibiotic classes prescribed was represented in Table 2 (main study population) and Table S6 (participants who went through health examinations).As the number of prescribed antibiotic classes increased, prostate cancer risk increased for both populations.Compared to the antibiotic non-users, participants with four or more kinds of antibiotics had a higher risk of prostate cancer (aHR, 1.18; 95% CI, 1.07-1.30).Similarly, antibiotic users with four or more types had a higher risk of prostate cancer (aHR, 1.11; 95% CI, 1.01-1.22)than those who utilized one kind of antibiotic.Among health examination participants, those who were prescribed three or more types of antibiotics had a higher risk of prostate cancer (aHR, 1.12; 95% CI, 1.04-1.21)than the antibiotic non-users.Additionally, compared to individuals who were prescribed one type of antibiotics, individuals who were prescribed three or more kinds of antibiotics had a higher risk of prostate cancer (aHR, 1.05; 95% CI, 0.97-1.13).
Table 3 shows a sensitivity analysis result with a wash-out period or shifting the index date for prostate cancer risk in proportion to cumulative days of antibiotics prescription.When 5 years of wash-out period was applied to a model with adjustments for various covariates, the adjusted hazard ratios for 180 or more days of antibiotics utilization were 1.36 (95% CI, 1.02-1.81).The risk-increasing trend as the accumulative days of antibiotic utilization increased was discovered.A similar trend was observed in a model adjusted with infectious diseases, along with the confounders reported above.In addition, compared to antibiotics non-users, antibiotics users for 180 or more days had a higher risk of prostate cancer (aHR, 1.31; 95% CI, 1.03-1.67)when a 7-year exposure period was given.
Table 4 represents prostate cancer risk with exposure to a single, specific class of antibiotics compared to the non-user group.Among classes of antibiotics, macrolides, penicillins, cephalosporins, fluoroquinolones, and tetracyclines users had a higher risk of prostate cancer than non-users.Compared to non-users, macrolides, penicillins, cephalosporins, fluoroquinolones, and lincosamides users had a higher risk of prostate cancer among participants who went through health examinations (Table S7).

DISCUSSION
In this retrospective, longitudinal study with more than one million participants aged 50 years or older, a durationresponse relationship between prostate cancer and antibiotic  The estimates were based on fully adjusted models (reference: none).bThe aHRs were calculated by Cox proportional hazards regression after adjustments for age, household income, Charlson comorbidity index, NSAID use, aspirin use, anti-diabetic drugs, 5-ARI use, Alpha-blockers use, prostatitis, benign prostatic hyperplasia, hypertension, and infectious diseases (respiratory diseases, urinary tract infections, skin, soft tissue, bone, and joint infections, intra-abdominal infections, intestinal infectious diseases, and other infectious diseases).The estimates were based on fully adjusted models (reference: 1-14 days).
use was discovered.Our study showed that prostate cancer risk may increase as the duration of antibiotic prescription and the number of antibiotic classes increase.Previously, several studies studied the association between antibiotics and the subsequent risk of various types of cancer. 5,6Petrelli et al. performed a systematic review proving a similar result-antibiotic usage was an independent risk factor for cancer, including prostate cancer. 5However, these studies mostly covered studies with Western populations.
Moreover, some studies specifically investigated the relationship between antibiotic use and prostate cancer risk.Tanim et al. suggested that antibiotic exposure significantly increased the risk of prostate cancer with a follow-up period of 15 years through a case-control study.Nonetheless, the authors concluded that there is a noncausal relationship between antibiotic exposure and the risk of prostate cancer due to the lack of temporal trends and the absence of classspecific effects. 7Our study conducted analyses with the inclusion of individuals' clinical status and numerous indications, complementing the primary limitation of that study.
Although a duration-dependent relationship was observed in our study, prospective studies are needed to evaluate the causation between antibiotic exposure and prostate cancer.Another study from the United States explored the relationship between antibiotics and anti-inflammatory use and the risk of prostate cancer.This study also proposed no association between antibiotics and the use of anti-inflammatory medication with prostate cancer risk. 13However, this study included a small sample size of 65 men with prostate cancer among 260 participants.This study also adjusted limited variables in logistic regression.Unlike these studies, our large population-based study suggested a duration-dependent response between antibiotics and the risk of prostate cancer, along with additional adjustments for potential confounders.We also showed that exposure to specific classes of antibiotics, such as penicillins, macrolides, cephalosporins, and fluoroquinolones, increased prostate cancer risk.
Several mechanisms may explain the positive durationresponse relationship between antibiotic use and prostate cancer risk.First, antibiotics may influence the microorganisms and metabolites in prostate microflora and promote prostate cell genetic instability. 16Induced oxidative stress and the subsequent cellular damage can promote the growth of atrophic luminal epithelial cells, resulting in proliferative inflammatory atrophy (PIA), a precursor to prostate cancer. 17Thus, prostate cell genetic instability may lead to the development of prostate tumors. 16Second, the impact of antibiotics on gut microbiota can be a potential link between antibiotic exposure and cancer development.Antibiotics may cause a change in bacterial diversity that leads to dysbiosis. 18,19In this regard, recent research established a "gut-prostate axis" in the association between gut microbiota and prostate health, including prostate cancer. 20,21Insulin-like growth factor 1 (IGF-1), affected by the gut microbiome, is linked to prostate cancer, which acts as an aggravating factor. 21Fujita et al. reported that folate and arginine pathways in the gut microbiota were significantly altered in men with prostate cancer, revealed in predicted metagenome analysis.They further noted that Bacteroides massiliensis in the gut microbiota of men with prostate cancer was elevated. 20Furthermore, prolonged antibiotic exposure caused the abundance of Proteobacteria in mice. 22This gut dysbiosis increased gut permeability and intratumoral lipopolysaccharide (LPS), which promoted prostate cancer growth via the NF-jB-IL6-STAT3 axis. 22Third, the interplay between gut microbiota and androgen metabolism in prostate cancer could be a hypothesis. 23Androgen plays an essential role in the development of prostate cancer through binding to the androgen receptor (AR) in prostate cancer cells. 20,24The gut microbiota plays a role in this androgenesis. 20Fourth, specific classes of antibiotics may have different mechanisms for cancer development.Tetracyclines, macrolides, and fluoroquinolones may impact the production of cytokines and prostaglandins. 6Cytokines can be exploited by tumor cells to promote growth, increase resistance to apoptosis, and foster dissemination, 25 and prostaglandins play a role in tumorigenesis and the progressions of cancers, including prostate cancer. 26Also, different effects among the type of antibiotics may be influenced by different levels of penetration into the prostate depending on the class of antibiotics.Charalabopoulos et al. suggested that penetration levels into the prostate varied based on the types of antibiotics. 27In our study, macrolides, penicillins, cephalosporins, fluoroquinolones, and tetracyclines, increased prostate cancer with statistical significance, which may be influenced by the different penetration levels into the prostate.However, since there is little to no evidence of this relationship, further research is needed to investigate this mechanism.Finally, antibiotics may be carcinogenic.Although there is minimal evidence of the carcinogenicity of antibiotics, Klkkinen et al. suggested that some antibiotics have a genotoxic potential. 6Genotoxin is a chemical agent that can cause damage to the DNA of a cell, which may result in somatic mutation and malignant transformation (cancer). 28ur study revealed several limitations.First, we could not include the participants' prostate specific antigen (PSA) level, one of the indicators to detect prostate cancer, 29 due to the limitation of our database.Second, we could not verify whether the antibiotics were taken as prescribed.Third, although this study considered 5-or 7-year antibiotic exposure in the analyses, further validation is needed with the utilization of longer periods of antibiotic exposure.Fourth, this cohort study was observational, so could not entirely rule out residual confounding effects through various confounders we adjusted for.Therefore, the findings should be interpreted with caution about causality.Finally, because the study was conducted among Asians, the results may not apply to other racial or ethnic individuals.
Despite its limitations, this study has multiple strengths.To our knowledge, this is the first retrospective cohort study that investigated the association between antibiotic exposure and prostate cancer risk in Asia.Second, this study included various potential confounders such as BMI, smoking, alcohol consumption, medications, and infectious diseases.Third, various sensitivity analyses were consistent with the results of the primary conclusions.This study's findings show that there may be a clear duration-dependent relationship between antibiotic prescription and prostate cancer risk.As the number of cumulative days on antibiotic subscription and antibiotic classes increased, the risk of prostate cancer also increased.Consequently, cautious prescription of antibiotics with consideration of potential detrimental effects is needed.

TABLE 1
Risk for prostate cancer according to cumulative days of antibiotics prescription.

TABLE 2
Risk for prostate cancer according to the number of antibiotic classes prescribed.Number of antibiotic classes prescribed during 5 years before the index date of follow-up investigation The aHRs were calculated by Cox proportional hazards regression after adjustments.Model 1 adjusted for age, household income, Charlson comorbidity index, non-steroidal anti-inflammatory drug use, aspirin use, anti-diabetic drugs, 5-ARI use, Alpha-blockers use, prostatitis, benign prostatic hyperplasia, and hypertension.Model 2: Model 1 + adjusted for infectious diseases (respiratory diseases, urinary tract infections, skin, soft tissue, bone, and joint infections, intra-abdominal infections, intestinal infectious diseases, and other infectious diseases).Antibiotics were divided into 11 classes consisting of penicillin, cephalosporin, macrolide, fluoroquinolone, sulfonamides, tetracyclines, lincosamides, and others.Abbreviations: aHR, adjusted hazard ratio; CI, confidence interval; ref, reference.

TABLE 3
Sensitivity analysis of risk for prostate cancer according to cumulative days antibiotics prescribed.
Note:The antibiotic exposure period in analyses about the wash-out period was set from 2002 to 2006, 5 years in total.The shifting index date indicates that the antibiotic exposure period was set from 2002 to 2008, 7 years in total.Abbreviations: aHR, adjusted hazard ratio; CI, confidence interval; ref., reference.aThe aHRs were calculated by Cox proportional hazards regression after adjustments for age, household income, Charlson comorbidity index, non-steroidal anti-inflammatory drug (NSAID) use, aspirin use, anti-diabetic drugs, 5-ARI use, Alpha-blockers use, prostatitis, benign prostatic hyperplasia, and hypertension.

TABLE 4
Risk of prostate cancer among participants with exposure to a single specific class of antibiotics compared with antibiotics non-user group.