Breast cancer patient delay in Fukushima, Japan following the 2011 triple disaster: a long-term retrospective study

This study took place at Minamisoma Municipal General Hospital (MMGH) and Watanabe Hospital (WH), located in Minamisoma City (Fig. 1). MMGH and WH stopped outpatient services immediately after the disaster, yet both restarted these services in June 2011. WH was the only medical institution in So-so District with an attending physician specialized in breast cancer care before the disaster. However, it lost inpatient functioning post-disaster, and stopped surgical therapy and chemotherapy for breast cancer patients. In August 2011, the WH breast cancer specialist moved to MMGH to restart breast cancer care for local residents, and since then, MMGH has been the only medical institution with a breast cancer specialist in So-so District (as of July 2016). These two hospitals are therefore considered to be the core breast cancer centers of So-so District.

To evaluate the impact of the 3.11 triple disaster on patient delay among breast cancer patients in So-so District, we retrospectively assessed the records of symptomatic breast cancer patients, newly diagnosed based on pathological findings, with first presentation to either of MMGH or WH from 1 January 2005 to 10 March 2016. The study period of 2005–2016 was chosen in order to assess any long-term influence of the disasters. Because this study focused on the period before a confirmatory diagnosis, we included patients who were later referred to other academic institutions or cancer centers after a pathological diagnosis of breast cancer in the two study hospitals. Patients with recurrent breast cancer, male patients, and those from outside So-so District were excluded. We also excluded those who were referred to MMGH or WH in the study period after first medical consultation to other medical providers during the non-study period. In the post-disaster period, we excluded those who had moved into the area following the disaster. Through the above process, patients were identified and categorized into two groups: pre- (from 1 January 2005 to 10 March 2011) or post-disaster (from 11 March 2011 to 10 March 2016).

First, in order to assess the existence of patient delay, the date of first symptom discovery by patients or their families, and the date of the first medical consultation were collected from patient records. When a patient was referred from other medical providers, the first visit at the initial medical institution was considered as the first medical consultation.

Second, factors potentially associated with patient delay were extracted from the records. Sociodemographic factors at the first presentation were considered as follows: age [ 7, 8], cohabitant status (number of cohabitant family members, living with children or not, and living with a partner or not) [ 4, 8] and employment status (full-time job or not) [ 4, 8, 9]. Clinical characteristics were additionally included as follows; major symptoms (lump or non-lump) [ 9], hormone receptor (HR) status (positive or negative), stage at the diagnosis [ 5], American Society of Anesthesiologists (ASA) physical classification system (normal healthy patient, patient with mild systemic disease, and patient with severe systemic disease) [ 5, 26], body mass index (BMI) (<25 kg/m ², 25–30 kg/m ², or ≥30 kg/m ²) [ 5], history of benign breast disease (yes or no) [ 8, 9], history of breast cancer (yes or no), mammography screening within two years (yes or no) [ 5], and family history of any kind of cancer (yes or no) [ 27, 28]. It has been suggested that HR-positive breast cancer progresses more slowly compared to HR-negative cancer [ 29]. Additionally considering that slow-developing cancers can lead to delayed medical consultation [ 30], HR status may influence length of patient interval. HR status was defined to be positive if either estrogen receptor status or progesterone receptor status was positive according to a cut-off value of no less than 1% in immunohistochemical (IHC) analysis [ 31]. As a population-based breast cancer screening is conducted through biennial mammography in Japan [ 32], previous experience of screening was classified according to whether there was attendance of mammography within the past two years. The details of family history of cancer, including which relatives had cancer and what kind of cancer they had, could not be examined because of data limitations. Four variables related to health care access were taken into account [ 3, 4]: hospital of the first medical consultation (MMGH or WH) and its linear distance to each patient’s residence, a referral to MMGH or WH from other medical providers (yes or no), and the interval from first medical consultation to first breast cancer specific examination (i.e. mammography, ultrasonography, or biopsy) [ 33].

Lastly, factors possibly associated with patient delay in disaster settings were gathered from patient records. In general settings, psychosocial distress can lead to a longer delay in first medical consultation among breast cancer patients [ 34]. A recent study suggested that evacuation status and residential air dose rates may be associated with psychosocial stress of local residents in the aftermath of Fukushima nuclear disaster [ 35]. We therefore considered it valuable to examine residential address of each patient and the air dose rate of radiation at their post-disaster residence in exploring the potential influence of psychosocial stress on post-disaster patient delay. The residential address of each patient was categorized into mandatory evacuation zone, voluntary evacuation zone, or non-evacuation zone of So-so District. For pre-disaster patients, residential addresses at the time of medical consultation were extracted, and for post-disaster patients, residential address at the time of the disaster was extracted. Methods of identifying the air dose rate for each participant’s residence are summarized in the next section. Further, medical costs of post-disaster patients were classified as free or not, according to the original addresses of participants; although Japan has universal health care [ 36], those who originally lived in mandatory evacuation zone or voluntary evacuation zone of Fukushima are now completely exempt from paying any medical fees as a disaster relief measure by the central government [ 37]. Hypothesized relationships between patient delay and factors studied are summarized in Table 1.

Patient interval was defined as the period from first recognition of a breast cancer-related symptom to the first medical consultation. According to the most generally used definition of patient delay [ 3, 6], total patient delay was defined as an interval of three months or longer. In order to assess patient delay comprehensively, we also introduced an additional category; excessive patient delay, defined as an interval of twelve months or longer [ 4, 38, 39].

After the FDNPP disaster, the Japanese Ministry of Education, Culture, Sports, Science, and Technology (MEXT) has conducted airborne radiation monitoring inside the 80 km radius of the FDNPP. The methods of these measurements have been documented in detail in a previous study [ 40]. All monitored results are publicly available [ 41].

In the present study, the post-disaster air dose rate at each participant’s residence was calculated using the following approach. We considered the results of the fourth MEXT monitoring performed between 22 October and 5 November 2011; sixth monitoring between 31 October and 16 November 2012; eighth monitoring between 2 and 19 November 2013; ninth between 1 September and 7 November 2014; and tenth between 12 September and 4 November 2015, for the radiation levels of the first (2011–2012), second (2012–2013), third (2013–2014), forth (2014–2015), and fifth (2015–2016) year post-disaster, respectively [ 41]. These values were averaged by a 500-m ² mesh on the basis of the Japan Profile for Geographical Information Standards (elevation and slope angle fourth mesh data) [ 42], and each participant’s house was assigned to a mesh area. This approach enabled estimation of the air dose rate at the home of each participant. In cases where the patients had experienced evacuation, post-evacuation addresses were used.

The ethics committee of MMGH assessed and granted approval for this study (approval number: 27–03). The ethics committee agreed that written consent from the participants was not required as this study was a retrospective analysis of patient records. All data were anonymized prior to analysis.

In the study period, 298 patients were newly diagnosed with breast cancer; 167 pre-disaster patients and 131 post-disaster patients. All non-symptomatic patients, comprised of 44 pre-disaster patients and 34 post-disaster patients, whose cancers were identified by a breast cancer examination (i.e. physical examination, mammography or ultrasonography) or incidentally, were excluded. Resultantly, there were 220 symptomatic patients in total; 123 pre-disaster patients and 97 post-disaster patients. Pre-determined exclusion criteria were then applied. One patient with a first consultation before the study period was excluded from the pre-disaster population. No patients with recurrent cancer, male patients, or those from outside So-so District were observed pre- or post-disaster. Additionally, no post-disaster patients had moved into the area following the disaster. There were 4 patients with a history of breast cancer defined as metachronal contralateral breast cancer cases, rather than recurrent cases. In total, 219 female breast cancer patients (73.5% of all cases, 99.5% of all symptomatic cases) were included, with 122 pre-disaster patients (73.1% of all pre-disaster cases, 99.2% of all pre-disaster symptomatic cases) and 97 post-disaster patients (74.0% of all post-disaster cases, 100.0% of all post-disaster symptomatic cases). The process of patient selection is displayed in Fig. 2.

Table 2 shows characteristics of the patients included in the analysis. There was no significant difference in the age distribution between pre- and post-disaster patients (Chi-squared test, p = 0.74). Additionally, distributions of breast cancer stage were not significantly different pre- and post-disaster (Fisher’s exact test, p = 0.22). However, a significantly smaller median number of cohabitant family members (1 vs. 2, p < 0.05), and a significantly smaller proportion of those living with children (47.4% vs. 59.0%, p < 0.05) were observed among post-disaster patients compared to pre-disaster patients. As number of cohabitant family member did not follow a normal distribution, we reported a median for this variable instead of its mean. With respect to clinical characteristics, there was no significant difference in the proportions of patients presenting with a lump between pre- and post-disaster patients (86.9% vs. 93.8%, p = 0.09). Regarding access to health service, there was no significant difference in the median distance between patient residence and study institutions at the first presentation (3.6 km vs. 3.3 km, p = 0.68) and the median days from first medical consultation to first breast cancer specific examination (0 vs. 0, p = 0.71) pre- and post-disaster. In the post-disaster period, there was a smaller proportion of patients who had resided in areas which became classified as parts of the mandatory evacuation zone, compared to the proportions of pre-disaster patients living in these areas (12.4% vs. 31.2%, p < 0.01). Also, in the post-disaster group, 84.5% of the patients were exempt from paying medical fees, and a mean air dose rate at patient’s residence was 0.31 μSv/h.

RRs of experiencing patient delay post- vs. pre-disaster are shown in Table 3. When we compared the overall post-disaster population with the pre-disaster baseline, there was a significant increase in the age-adjusted RR for both total patient delay (RR: 1.66, 95% Confidence Interval (CI): 1.02–2.70, p < 0.05) and excessive patient delay (RR: 4.49, 95% CI: 1.73–11.65, p < 0.01). According to the analysis of respective year, the RR of experiencing total patient delay peaked in the fourth year post-disaster (RR: 2.05, 95% CI: 1.10–3.81, p < 0.05), and that of excessive patient delay exceeded 5.0 in the second (RR: 5.58, 95% CI: 1.77–17.56, p < 0.01), fourth (RR: 5.27, 95% CI: 1.73–16.03, p < 0.01), and fifth year (RR: 5.24, 95% CI: 1.64–16.78, p < 0.01) post-disaster.

Patient interval, instead of date of first symptom recognition, was reported in the majority of patient records. In such cases, the reported interval was directly collected from the patient records. We summarize sociodemographic factors, patient interval and date of first presentation for patients with excessive patient delay for both pre- and post-disaster in Table 4. Among the 18 patients with excessive patient delay in the post-disaster group, 27.8% (5/18) discovered their symptoms before the disaster.

The results of univariate logistic regression analysis for pre- and post-disaster total patient delay are shown in Tables 5 and 6, respectively. In the pre-disaster period, no associations between total patient delay and access-related factors, sociodemographic factors, or other clinical characteristics studied were observed. In the post-disaster period, none of access- and disaster-related factors and sociodemographic factors were significantly associated with experiencing total patient delay, however a significant association was observed with having a family history of any cancer (odds ratio: 0.38, 95% CI: 0.15–0.95, p < 0.50). Due to lack of significant variables, we did not conduct multivariate logistic regression analysis for total patient delay.

We attempted to conduct logistic regression analyses for excessive patient delay in respective pre- and post-disaster periods. However, it was difficult to establish a stable model of pre-disaster excessive patient delay owing to limited numbers of participants with this delay. We present characteristics of pre- and post-disaster patients stratified by with or without excessive and total patient delay in Additional files 1 and 2. Only 22.2% (4/18) of post-disaster patients with excessive patient delay lived with their children, compared to 53.2% (42/79) of those without excessive patient delay.

This study has several limitations. First, it is possible that selection bias may have influenced our results. A potential effect of post-disaster demographic changes, primarily due to mass-evacuation among young- and middle-aged generations, should be acknowledged as it may have resulted in differences between the pre- and post-disaster populations of So-So district. For example, if younger residents evacuated and did not return, an older population may naturally have led to larger proportions of patient delay [ 7, 8], which could have led to an overestimation of our results. However, aside from differences in areas of residential address and cohabitant status, we could find no significant differences in demographics of pre- and post-disaster participants, and it is particularly notable that there were no changes in the age distribution of patients in pre- and post-disaster periods. Second, although we included as many breast cancer patients as possible in the affected areas based on the data from MMGH and WH, there was still a limited sample size that provided low power for analyses and the observed results may have been affected by random error. Third, in most cases, patient intervals were reported instead of date of symptom recognition. As patient intervals lengthened, these variables may be more easily affected by recall biases. Finally, we could not evaluate some factors known to be associated with patient delay but not reported in patient records, including breast cancer knowledge, personalities, and sources of social support apart from family members [ 4, 8, 49].