In recent years, it has become increasingly obvious that the consumption of antibiotics may play an important role in the development of cancer. Both experimental and clinical analyses have demonstrated an increased incidence of cancer in patients with a history of antibiotic intake, but results have remained inconclusive (Bach et al. 2020; Bagley et al. 2017; Ramírez-Labrada et al. 2020; Wertman et al. 2021; Bi et al. 2021; Cheung et al. 2021; Amadei and Notario 2020; Martins Lopes et al. 2020). Using the population-based Disease Analyzer database (IQVIA) (Rathmann et al. 2018), we demonstrate that the probability of cancer is significantly higher among patients with a history of antibiotic intake. This effect was especially striking in patients receiving penicillins and cephalosporins, while tetracyclines and macrolides appeared to protect against the development of cancer. Our analyses stratified by cancer site revealed an even more complex picture. Here, the consumption of penicillins was significantly and positively associated with cancer of the respiratory organs and, in the case of low consumption, also with cancers of the lymphoid and hematopoietic tissue, while the intake of cephalosporin was significantly associated with respiratory organ cancer, breast cancer, and cancer of the lymphoid and hematopoietic tissue, highlighting a potentially distinctive pathophysiological connection between different classes of antibiotics and different cancer entities.
Cancer is a multifactorial disease with a complex pathophysiology. Genetic causes have been studied extensively in past; however, it has recently become more obvious that many cancer cases are even more closely associated with environmental factors or are associated equally with a combination of both (Wu et al. 2018). For example, human papilloma virus infection and tobacco smoking are responsible for up to 90% of cervical squamous cell carcinomas and lung cancers, respectively (Lewandowska et al. 2019). Recent epidemiological studies have established the commensal microbiota as a previously neglected modulator of carcinogenesis, immune response, and response to anti-cancer therapy (Bhatt et al. 2020; Fessler et al. 2019; Raza et al. 2019). Today, various authors theorize that individual microbial pathogens contribute to cancer development in approximately 15–20% of all cases (González-Sánchez and DeNicola 2021; Martel et al. 2012). It has been suggested that different lifestyle factors that have been found to be associated with an elevated risk of cancer development might act via microbiota-related mechanisms (González-Sánchez and DeNicola 2021). In this context, it was hypothesized that the intake of antibiotics, which are related to profound and long-term changes in the human microbiome, might be associated with an increased risk of various forms of neoplasia. Supporting this hypothesis, our findings show that the probability of cancer in general is significantly higher among patients with a history of antibiotic intake than in matched controls. In this context, recent studies have suggested that cancer initiation and progression are complex processes that are impacted in a very specific manner by global changes in the microbiome rather than by single pathogens (González-Sánchez and DeNicola 2021; Bhatt et al. 2017; Schwabe and Jobin 2013). In line with this, we demonstrate that while penicillins and cephalosporins increase the odds of cancer, tetracyclines and macrolides may instead have a preventive effect with respect to cancer. Notably, these data are in line with recent data on colorectal cancer (CRC) showing that the intake of both penicillins and cephalosporins is associated with colorectal cancer while the consumption of tetracyclines is not (Simin et al. 2020). By contrast, no such differences between different antibiotics were found for breast cancer, in which tetracycline actually had the strongest cancer-promoting effect, underscoring the complexity of the association between antibiotics and cancer.
The vast majority of human microbiota reside in the gastrointestinal tract, particularly in the colon, and can interact both locally and systemically with cancer cells. Consequently, most epidemiological analyses in the past have focused on this interaction. Aneke-Nash and colleagues recently performed a meta-analysis of six studies providing 16 estimates of the association between the level of antibiotic consumption and colorectal neoplasia and showing that individuals with the highest levels of antibiotic exposure had a 10% higher risk of colorectal neoplasia than those with the lowest exposure (Aneke-Nash et al. 2021). In addition, this effect differed between broad- and narrow-spectrum antibiotics, and possibly within the colorectal continuum. Due to the small sample sizes available, we are unable to draw conclusions about the antibiotic-related cancer risk specifically in the colorectum. Beyond colorectal cancer, we also provide a comprehensive analysis on the association between the intake of antibiotics and many different cancer entities by analyzing > 220,000 patients from a large population-based database in Germany. We also demonstrate an elevated risk of non-digestive tract cancers in patients with a history of antibiotic consumption. Notably, specific microbial populations have been described for many organs, revealing a different microbiome signature for each (summarized in González-Sánchez and DeNicola (2021)). Clearly, then, antibiotic-induced alterations in the local microbiome are likely to play a key role in the development of cancer in organs distant from the gut, which is in line with our findings. Of course, a potential bias must not be forgotten when considering the cause of the antibiotic-associated increase in cancer rates suggested by our data. We cannot exclude the possibility that the accumulation of patients suffering from certain diseases that frequently lead to antibiotic consumption and are associated with cancer development may amplify the effect. Examples of such diseases include ulcerative colitis and primary sclerosing cholangitis.
We acknowledge the fact that our study is subject to various limitations, most of which are due to the chosen study design and cannot be avoided (Labenz et al. 2020a; Loosen et al. 2021; Roderburg et al. 2021). Most importantly, diagnoses within our database are coded as ICD-10 codes, which might be associated with the misclassification of certain diagnoses. In addition, data might be incomplete for certain patients; in particular, information regarding lab values or drug intake was not available for all patients, leading to their exclusion from this analysis (Table 1). Furthermore, data on the socioeconomic status (e.g., education and income of patients) as well as lifestyle-related risk factors (e.g., smoking, alcohol consumption, and physical activity) are lacking within the Disease Analyzer database and thus cannot be taken into account in our study. However, the IQVIA Disease Analyzer database used for the present analyses has been used extensively for various academic publications (e.g., (Huber et al. 2020; Jacob et al. 2021; Labenz et al. 2020b)) and its validity has been well demonstrated (Rathmann et al. 2018). We also want to highlight the fact that a selection bias must be assumed due to the fact that patients who take antibiotics more frequently presumably have more frequent contact with their physicians than those who take no antibiotics. Such a bias might also explain the finding that the step to the minimum observation time of 5 years prior to the index date reduces the number of persons to about 1/3 in cancer cases but 1/5 in non-cancer cases. This may be due to the fact that cancer patients had presumably attended the doctor for a greater number of diseases previously and were, therefore, observed for a longer period or more often and that patients with a longer medical history have a better chance of having cancer detected. Next, subgroup analyses of individual cancer sites (e.g., left/ right sided colorectal cancer) were not feasible due to the small sample sizes available. We, therefore, grouped different tumor entities with similar pathomechanisms (e.g., digestive or respiratory organs), which might be associated with a presentation bias as described in detail in a recent study (Loosen et al. 2021). Finally, the information regarding antibiotic consumption was only available for a sufficient number of patients for a 5-year period, meaning that it was not possible to assess the influence of longer periods of antibiotic consumption, potentially leading to an underestimation of the association.
In summary, by analyzing data from a large German primary care provider database, we demonstrate that the intake of various antibiotics is associated with an increased risk of cancer in a dose- and tumor site-specific manner, irrespective of patients’ age and sex. Thus, along with previous data, our study including > 220,000 patients suggests that the clinical management of patients needing antibiotics should include a careful and structured risk assessment for the development of cancer to improve long-term outcomes in these patients. For example, patients with high/frequent consumption of penicillins or cephalosporins might be presented in a specific “board” and discussed with dedicated infectiologists and oncologists, as was recently suggested in the context of specific antibiotic stewardship programs.