Incidence of interstitial lung disease in patients with breast cancer: a nationwide database study in Japan

Breast cancer is the most commonly diagnosed cancer among women worldwide. Approximately 2.3 million new cases were diagnosed in 2020 and 685,000 breast cancer-associated deaths occurred in the same year [1]. Breast cancer is the leading cancer among women in Japan, accounting for approximately 20% of all cancer cases [2]. Advancements in molecular targeted therapies have remarkably improved the survival rate of patients with breast cancer [3]. However, some of these therapies have been associated with an increased risk of drug-induced interstitial lung disease (ILD) [4–7].

ILD is a potentially life-threatening disorder characterized by inflammation and fibrosis of the pulmonary interstitium. ILD can curtail the application of certain anticancer therapies, as caution or contraindications are often advised in the prescribing information. A population-based study reported that the incidence rate of respiratory failure attributed to ILD was 6.6 per 100,000 patient-years and that more than half of these instances were concomitant with the administration of chemotherapeutic agents [8]. The use of novel drugs with mechanisms that target novel molecular pathways, including molecular targeting agents and immune checkpoint inhibitors, have been associated with an increased incidence of ILD [9]. Additionally, a comparative analysis of drug-induced pulmonary diseases on an international scale demonstrated that the incidence rates in Japan were substantially higher than those reported in clinical trials and observational studies conducted in other countries [7,10,11]. However, the epidemiological data on ILD in patients with breast cancer in Japan, particularly in the real-world setting, remain scarce.

This study aimed to address the knowledge gap by estimating the incidence of ILD in patients with breast cancer using a large nationwide claims database in Japan.

Materials & methods

Study design & data source

This study utilized Japanese insurance claims and Diagnosis Procedure Code data from the Medical Data Vision database to estimate the incidence of ILD in patients with breast cancer who were treated with anticancer chemotherapeutic regimens. The Medical Data Vision database is a large nationwide claims database that contains anonymized patient-level data of approximately 40 million patients in Japan from 469 hospitals that use a diagnosis procedure combination per diem payment system [12,13]. These institutions represent 27% of all the acute care hospitals in Japan. According to the Japanese Ethical Guidelines for Medical and Health Research Involving Human Subjects, informed consent was not required for this observational study as it used secondary data that did not contain any identifiable patient information.

Study population

The study population comprised patients with breast cancer who met all the following criteria: 1) a diagnosis record of C50 (malignant neoplasm of the breast) based on the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) code [14]; and 2) at least one prescription record of anticancer therapies (defined by Anatomical Classification of Pharmaceutical Products code [Anatomical Therapeutic Chemical code] [15] L01), between April 2008 and June 2022. These anticancer therapies did not include the exclusive use of hormone therapy for cancer treatment.

Patients were excluded from the study if they met any of the following criteria: 1) no records within 60 days after the initiation of anticancer therapy, 2) determination of index date between 1 April 2008 and 31 March 2009 and 3) diagnosis record of malignant neoplasms other than breast cancer. The rationale behind these exclusion criteria is grounded in several considerations. Firstly, excluding early index dates was necessary to avoid attributing ILD to anticancer therapies administered prior to the study period. Secondly, ensuring a minimum observation period of 60 days post-therapy initiation was vital to adequately monitor for ILD development and to establish a robust association between the therapy and ILD. Lastly, omitting patients with diagnoses of nonbreast cancers was essential to prevent misattribution of ILD to therapies for other cancers and to mitigate potential misclassifications due to drug regimens for other cancer types.

Definition of treatment regimen

Administration of any of the drugs delineated in Supplementary Table 1 was identified as anticancer therapy for breast cancer. The first date of each anticancer therapy was defined as the index date. Anticancer agents administered within 21 days of the index date were incorporated and interpreted as a uniform regimen. Upon the introduction of new anticancer therapies, the existing treatment regimen was deemed discontinued, and a novel regimen was considered to have been initiated. Each drug combination within a treatment regimen was classified in alignment with Supplementary Table 2. The regimens were meticulously segregated into ten categories: antibody–drug conjugates, anti-HER2 antibodies, HER2 inhibitors, anti-VGFR antibodies, PARP inhibitors, immune checkpoint inhibitors, CDK4/6 inhibitors, mTOR inhibitors, cytotoxic drugs and others.

The duration of each treatment regimen was calculated from the initiation and termination dates of exposure. Onset of exposure was defined as the index date for each treatment regimen, whereas the termination of exposure was determined as the prescription date for injectable drugs, plus 21 days, or the prescription date plus the prescribed period for oral medications. The suspended treatment period was calculated as the difference between the initiation of exposure and last termination of exposure within the same treatment regimen. However, this was not incorporated into the duration of exposure.

Definition of an ILD event

Moderate-to-severe drug-induced ILD events were categorized when all the following criteria were met: 1) at least one ICD-10 code among J702, J704, J841 or J849 (‘interstitial pulmonary disease’ or ‘interstitial pneumonia’); and 2) at least one prescription record of corticosteroids, more than 20 mg/day of prednisolone or 80 mg/day of methylprednisolone, in accordance with the consensus statement for diagnosis and treatment of ILD [16]. Additionally, a condition was imposed that there must be a prescription record for criterion 2) in the same month as the month to which the diagnosis record for criterion 1) belongs. When multiple records of ILD were observed per patient, the first ILD in each patient was analyzed. ILD events were defined as outcome events for this study only if there was a prescription for anticancer therapies within 49 days from the date of occurrence of an ILD event. A treatment regimen was deemed as implicated in ILD development if it was administered within 49 days prior to the onset of ILD. This specific timeframe was determined in consideration of the typical cycles of anticancer regimens for breast cancer, allowing for scenarios in which a treatment hiatus might occur due to emerging signs of ILD. Consequently, corticosteroid treatment might be initiated in the following cycle or at the subsequent monthly visit, dependent upon symptomatic evaluation and progression.

Statistical analysis

Descriptive statistics were used to summarize the demographic and clinical characteristics of the study population. ILD incidence was calculated as the number of ILD events per 100 person-years of follow-up. The incidence rate of ILD and its 95% CI were calculated using Poisson distribution. To elucidate which regimen carries an augmented risk of ILD, the prevalence of ILD was calculated by delineating the number of ILD occurrences over the cumulative number of anticancer therapies for each regimen category. The 95% CI for the prevalence of ILD was calculated using the Clopper–Pearson method. The interval between the initiation of the most recent anticancer therapy and the occurrence of ILD was analyzed using the Kaplan–Meier method for patients who received a first-time diagnosis of ILD during the observation period. All analyses were conducted using the SAS software version 9.4 (SAS Institute Inc., NC, USA).

Results

Patient characteristics

A total of 81,601 patients with breast cancer who were treated with anticancer therapies between April 2009 and June 2022 were included in this study (Figure 1). Table 1A summarizes the characteristics of the study population. The demographic distribution of age at the time of initial diagnosis of breast cancer demonstrated a wide spread, with 6.0% falling under 40 years of age, 45.8% between 40 and 60 years of age, and the largest group, 48.3%, being 60 years or older. The median age of the patients was 56 years (range: 19–100 years). The index year at initial diagnosis showed a trend of increasing diagnoses with year.

Prescription patterns of anticancer therapies

Figure 2 shows the prescription patterns of the anticancer drug regimens for patients diagnosed with breast cancer between 2009 and 2022. The aggregate number of regimens demonstrated an annual escalation, culminating in 170,921 prescribed regimens during the defined period. ‘Cytotoxic drugs’ were the most recurrently prescribed regimen during the studied period, accounting for 62.4% of all prescriptions. Notably, the rate decreased over time, from 84.7% in 2009 to 56.0% in 2022.

Anti-HER2 antibody regimen prescriptions showed a semblance of consistency over the years, accounting for approximately 16–26% of the total prescriptions. A steady increase in prescription of anti-VEGFR antibodies was observed in 2012, with the proportion plateauing in 2015. There was an increase in the prescription proportion of mTOR inhibitors from 2014, which declined following the introduction of CDK4/6 inhibitors in 2018. The inception of CDK4/6 inhibitors occurred in 2018, leading to a drastic escalation, ultimately representing 13.0% of the total prescriptions in 2022. From 2020 onwards, the prescriptions for immune checkpoint inhibitors began to burgeon substantially; however, they remained a minority of the total prescriptions. Antibody–drug conjugates also showed an upward trend over the years, albeit they persisted as a minor fraction of the total prescriptions.

Incidence of ILD

Table 1B shows the characteristics of the 1042 patients who developed ILD requiring steroid treatment. The majority of the patients (67.1%, n = 699) were ≥60 years at the time of ILD diagnosis. The next largest group (29.9%, n = 312) belonged to 40–60 years of age, and a small proportion (3.0%, n = 31) was under the age of 40 years. The median age was 65 years. The median number of treatment regimens was two, with a range between one and 16. Most patients (40.9%, n = 426) had received a single treatment regimen before ILD diagnosis, followed by those who received two (31.4%, n = 327), three (11.9%, n = 124) and four or more treatment regimens (15.8%, n = 165). The annual incidence of ILD is shown in Table 2. Across the entire study period, the incidence rate was 1.41 per 100 person-years (95% CI: 1.33–1.50), based on 1042 ILD cases identified among 81,601 patients. The annual incidence varied over time. The incidence was relatively low in the early study period, with rates of 0.79 (95% CI: 0.10–2.87) in 2009 and 0.25 (95% CI: 0.05–0.73) in 2010. A gradual increase at a rate of 0.50 (95% CI: 0.20–1.02) was observed from 2011, reaching 0.98 (95% CI: 0.73–1.28) in 2014. Notably, a decrease was observed in 2018 with a rate of 1.11 (95% CI: 0.90–1.35). The incidence then escalated again, reaching 2.01 (95% CI: 1.72–2.34) in 2021. Although the data for 2022 only reflected the first half of the year, the highest rate was observed in this year at 2.87 (95% CI: 2.10–3.83).

Occurrence of ILD by regimen

Table 3 presents the ILD occurrence for each treatment regimen. The mTOR inhibitor regimen was associated with the highest prevalence of ILD, accounting for 6.60 events per regimen. This was followed by immune checkpoint inhibitors with a prevalence of 2.17 events per regimen, and antibody–drug conjugates with 1.48 events per regimen. CDK4/6 inhibitors were associated with a lower proportion of 0.96 events per regimen. On the other end of the spectrum, several regimens, such as anti-VGFR antibodies, PARP inhibitors, cytotoxic drugs and other therapies, showed an ILD prevalence of less than 1.00 events per regimen. Cytotoxic drugs constituted the largest percentage of regimens at 62.4%, and their corresponding ILD prevalence was relatively low (0.43).

Onset date of ILD

The median time duration from the initiation of anticancer therapy to the first occurrence of ILD was 69 days (interquartile range: 42–113 days), with an estimated 65.7% of the ILD incidents occurring within the initial 90-day period (Table 4). The frequency of ILD onset tapered considerably after 120 days, with less than 20% of patients experiencing ILD onset beyond this time point. Furthermore, the incidence of ILD plateaued beyond the 120-day mark (Figure 3).

Pretreatment regimens received by patients with ILD

The pretreatment regimens used for patients with ILD are listed in Table 5. Cytotoxic drugs were the most commonly used regimen, accounting for use in 43.8% (n = 456) of the patients. Anti-HER2 antibodies were used in 16.8% (n = 175) of the patients, followed by mTOR inhibitors in 15.8% (n = 165) of the patients. CDK4/6 inhibitors were used in 10.0% (n = 104) of the patients, followed by anti-VEGFR antibodies and antibody–drug conjugates, which were used in 6.1% (n = 64) and 5.4% (n = 56) of the patients, respectively. The remaining regimens, including HER2 inhibitors, immune checkpoint inhibitors, PARP inhibitors and others, were administered to a small percentage (<1%) of patients.

Discussion

To the best of our knowledge, this is the first study to provide detailed epidemiological data on moderate-to-severe drug-induced ILD incidence among patients with breast cancer using a large nationwide claims database. In addition, the observation period spanned over more than a decade, providing valuable insights into the evolving landscape of breast cancer treatment and its potential impact on ILD risk.

We evaluated the incidence of ILD requiring steroid treatment, thus focusing on moderate-to-severe cases of ILD among patients with breast cancer receiving various anticancer therapy regimens in Japan. The incidence of ILD in Japan has been reported to be substantially higher than that reported in clinical trials and observational studies conducted in other countries [7,10,11]. A Japanese subgroup analysis of the BOLERO-2 trial, which focused on everolimus' efficacy in patients with metastatic breast cancer, uncovered a significant disparity in the frequency of noninfectious lung-related adverse events between the Japanese cohort and the overall population. Specifically, the Japanese cohort experienced a 31.0% incidence, while the global population demonstrated a lower rate of 15.6% [17]. Although the reasons for this are not fully understood, potential influences could include the genetic makeup of the Japanese population and the specific ways in which ILD is diagnosed and reported in the country [11,18].

Our findings suggest a considerable fluctuation in the incidence of ILD among patients with breast cancer in Japan between 2009 and 2022. An overall increasing trend was observed. The gradual increase in the incidence of ILD from 2011, peaking in 2022, could be attributed to the incorporation of novel anticancer therapies into the treatment repertoire for breast cancer during the same period, which is supported by the observable prescription pattern shown in Figure 2. This trend is particularly concerning given the already higher incidence of drug-induced ILD in Japan than in other countries.

Observing the age distribution among 81,601 patients with breast cancer undergoing anticancer therapies and 1042 patients who developed ILD, a pattern emerged. The median age of the overall cohort was 56 years, whereas the median age of those developing ILD was 65 years. Given the observational nature of our study, which precludes the establishment of a causal relationship between age and ILD development, cautious interpretation of these findings is warranted. These preliminary observations may serve as a foundation for future controlled investigations into the potential age-related predisposition to ILD among anticancer therapies.

Considering the patterns of the incidence of ILD in specific years, an intriguing transient decrease was apparent, particularly in 2018. This decrease is conceivably related to a decline in the prescription of mTOR inhibitors, which are frequently employed in therapeutic contexts similar to CDK4/6 inhibitors, that is, in patients with hormone receptor-positive, HER2-negative metastatic or recurrent breast cancer. The introduction of CDK4/6 inhibitor after 2018 seems to have driven a shift in the treatment landscape as clinicians started replacing mTOR inhibitors with this newer class of drugs. Although our study showed that the highest ILD incidence rate was associated with mTOR inhibitors, the reduced use of these drugs due to the advent of CDK4/6 inhibitors likely impacted the overall ILD incidence rate in 2018. However, after 2018, we observed an increase in the incidence of ILD, which might have been associated with the rising prescription of CDK4/6 inhibitors and antibody–drug conjugates. Although the incidence rate of ILD associated with CDK4/6 inhibitors and antibody–drug conjugates in our study was approximately 1.0, an increase in their prescription could still contribute to an overall increase in ILD events, especially given the documented risk of ILD associated with these drugs [6,19,20]. However, these observations merely suggest correlations and do not establish a direct cause-and-effect relationship between the use of these drugs and the incidence of ILD. There may be other confounding factors at play, including underlying patient conditions or co-administered treatments, thus warranting careful interpretation of the data.

The incidence of ILD was associated with the type of treatment regimen used, with some regimens showing a higher risk than others. For instance, patients treated with mTOR inhibitors had the highest prevalence of ILD, followed by those who were treated with immune checkpoint inhibitors. This is consistent with the findings of previous studies, which indicated an increased risk of ILD associated with these therapies [7,21]. Further research is needed to elucidate the mechanisms underlying these observed differences and develop strategies to mitigate this risk. Moreover, our study revealed that cytotoxic chemotherapy was the most common regimen among patients who developed ILD, accounting for 0.43 events per regimen. This is consistent with previous reports suggesting that cytotoxic drugs are known risk factors for ILD [7]. However, despite being the most common regimen, cytotoxic chemotherapy was not associated with the highest prevalence of ILD. A detailed analysis of the various cytotoxic agents was not feasible because of the complexity of the treatment regimens, especially those involving combination treatments and concurrent use with molecularly targeted agents. Consequently, cytotoxic agents were grouped into a broad category, which is a limitation that we acknowledge in this study. Nevertheless, future research should explore this topic more thoroughly, potentially revealing additional nuances between specific cytotoxic agents and ILD incidence.

In the context of the time to ILD onset, the majority of the incidents occurred within the first 90 days of initiation of anticancer therapy. This emphasizes the need for vigilance and regular patient monitoring during the early stages of therapy. The rapid onset of ILD can seriously influence treatment plans, necessitating alterations in the therapeutic regimen or even treatment cessation. Recognizing the early symptoms of ILD and initiating appropriate management strategies can substantially improve patient outcomes.

The prescription patterns of anticancer therapies also reflect the dynamic landscape of breast cancer treatment, indicating that the introduction of novel therapies could influence the incidence of ILD. This highlights the importance of postmarketing surveillance for novel therapies, particularly in real-world settings, to identify and manage potential risks effectively.

Our study had a few limitations which need to be considered. First, the use of claims data may introduce potential misclassification, overestimation or underestimation of ILD events as diagnoses are based on ICD-10 codes and treatment records rather than clinical assessments. Certain biomarkers, such as oxygen saturation (SpO₂), KL-6 and SP-D, which are recognized for their potential utility in ILD diagnosis, could not be employed to define ILD events owing to their limited data availability in this database. Recognizing the inherent challenges in distinguishing ILD from other diseases with similar clinical presentations, we defined an ILD event as a condition that was diagnosed as ILD based on ICD-10 codes and was treated with high-dose steroids, a common therapeutic strategy for drug-induced ILD. Although the administration of steroids and the diagnosis of ILD were concurrent (within the same month), and while the steroid dosages administered align with those typically utilized in standard ILD treatment, it cannot be unequivocally affirmed that the steroids were employed specifically for ILD management, introducing a potential limitation to our findings. Second, this study was limited to the incidence of ILD requiring steroid treatment and did not assess mild ILD manifestations, such as instances in which patients recovered following drug cessation without systemic corticosteroid intervention. Due to the lack of detailed grading information, ILDs that were treated with steroids were considered to be of at least moderate severity and were defined as events. Third, the exact onset of ILD could not be identified. Therefore, we used the date of corticosteroid treatment initiation as a proxy marker for ILD onset. Fourth, considering the pathophysiological underpinnings of ILD onset and hitherto identified causative agents, this study predominantly focused on anticancer therapies that harbor potential cytotoxic agents, including targeted therapy. Consequently, the incidence of ILD in patients with breast cancer receiving exclusive hormonal therapy was not assessed. Finally, we were unable to account for other potential risk factors for ILD, such as smoking status, genetic predisposition and concomitant medications.

Conclusion

This large nationwide claims database study provides valuable insights into the incidence of ILD among patients with breast cancer treated with various anticancer therapy regimens in Japan. The observed increase in the incidence of ILD over time, along with the possibility of differences in ILD risk across treatment regimens, emphasizes the need for continued monitoring, particularly during the initial stages of treatment, and management of ILD in breast cancer patients receiving anticancer therapies. Further research is needed to elucidate the underlying mechanisms contributing to the development of ILD and to explore potential preventive and treatment strategies.