A rapid rise in hormone receptor-positive and HER2-positive breast cancer subtypes in Southern Thai women: A population-based study in Songkhla.

The incidence of breast cancer is increasing in low- and middle-income countries, including Thailand. However, its molecular immunohistochemical (M-IHC) subtypes have not been summarized in a population-based cancer registry. Thus, we aimed to estimate the breast cancer incidence and trends based on the hormone receptor and human epidermal growth factor receptor 2 (HER2) status. This cross-sectional study included 2,883 women diagnosed with primary invasive breast cancer between 2009 and 2018 from the Songkhla Cancer Registry. After imputing the missing values of estrogen receptor (ER), progesterone receptor (PR), and HER2 status, the cases were classified into four subtypes: HR+/HER2-, HR+/HER2+, HR-/HER2-, and HR-/HER2+. The age-specific incidence rate of 5-year age groups and age-standardized incidence rate (ASR) were calculated. An age-period-cohort (APC) model was used to describe the effects of age, birth cohort, and period of diagnosis. Finally, the incidence trends were extrapolated to 2030 based on the APC and joinpoint models. The results showed, HR+/HER2- had the highest ASR in breast cancer. The incidence trends of HR+/HER2- and HR+/HER2+ increased with an annual percent change of 5.4% (95%CI: 2.5% to 8.3%) and 10.1% (95%CI: 4.9% to 15.5%), respectively. The rate ratio was high in the younger generation and recent period of diagnosis. The joinpoint and APC model projections showed that the ASR of HR+/HER2- would reach 30.0 and 29.2 cases per 100,000 women, while ASR of the HR+/HER2+ would reach 8.8 and 10.4 cases per 100,000 women in 2030. On the other hand, the incidence trends of the HR-/HER2- and HR-/HER2+ subtypes were stable. The rising trends of HR-positive and a part of HER2-positive breast cancer forecast a dynamicity of the future health care budgeting, resource allocation, and provision of facilities.

Introduction

Breast cancer is the most common cancer among women worldwide [1]. In 2020, the age-standardized incidence rate (ASR World) estimated by GLOBOCAN was 47.8 cases per 100,000 women, while the number of female patients with breast cancer was nearly a quarter of the total proportion of women with various types of cancers worldwide [1, 2]. The incidence of breast cancer was relatively high and stable in countries with very high and high human development index (HDI) scores, respectively [1]. By contrast, countries with medium and low HDI scores had a lower incidence [1], while the cancer trends increased [1].

The incidence of breast cancer has been rising in Thailand, especially among women aged >50 years [3, 4]. The proportion of luminal A-like subtype was significantly high among the molecular immunohistochemical (M-IHC) subtypes of breast cancer [5]. Nevertheless, triple-negative cancer was disproportionally high among Muslim women in Southern Thailand. Moreover, Muslim women with breast cancer, showed poor survival [6].

Since 2002, Thailand has implemented a universal health coverage (UHC) scheme [7]. This is the primary insurance scheme for Thai citizens. As of 2020, the UHC scheme has already provided health services to approximately 79% of the Thai population [8]. It covers health promotion, disease screening, treatment of basic and high-cost diseases under the Thai National List of Essential Medicines (NLEM), and palliative/supportive care.

The main goal of breast cancer treatment is to cure the disease. Surgery is the primary option for the tumor removal, while adjuvant therapy destroys the residual malignant cells. The following four main types of treatment modalities are used for breast cancer treatment: chemotherapy, hormonal therapy, targeted therapy, and radiotherapy. The use of hormone and targeted therapy is dependent on the female sex hormone receptor (HR) (estrogen receptors [ER] and progesterone receptors [PR]) and human epidermal growth factor receptor 2 (HER2) status in addition to the histologic type of breast cancer, grade, lymph nodes, and distant metastatic organ involvement [9, 10].

Trastuzumab (anti-HER2) was included in the NLEM in 2014 [11]. At a certified cancer clinic, patients with breast cancer who require targeted therapy can be registered in the national medical insurance funds and get access to the drug. Trastuzumab has proven to be cost-effective in the treatment of patients with early-stage breast cancer in Thailand [11]. Although targeted therapy is a cost-effective treatment for early-stage breast cancer, it costs 15,560 USD per patient to add a targeted therapy to the baseline treatment; meanwhile, the quality-adjusted life years (QALYs) expected to increase by 4.59 years [11]. Thus, the national health budget increases as the number of patients with early-stage breast cancer increase. However, longer life gain for women who benefited from a targeted therapy would make a return in terms of the gross domestic product (GDP) per capita, the benefit of which is greater than the budget spent by the UC scheme for the cost of treatment per patient.

Other targeted drugs approved by the Thai Food and Drug Administration for breast cancer treatment include pertuzumab and ribociclib. Currently, these are the only two drugs available to patients who are covered by the civil servant medical benefit scheme [12, 13]. Drug indications mainly depend on the HR and HER2 status. Therefore, determining the actual proportion and trends of breast cancer according to the HR and HER2 status is necessary. This data can help predict the need for targeted drug therapy as well as the survival in each breast cancer subtype. It also provides local information that helps increase the healthcare planning validity and assists in the proper allocation of healthcare resources.

The M-IHC subtype is not a mandatory reportable type of breast cancer according to the World Health Organization criteria. However, it is preferable for all cancer registries to record and report the incidence of M-IHC subtypes. The existing system of Songklanagarind Hospital reports both population-based and hospital-based statistics for histological but not M-IHC subtypes. The HR and HER2 status can be retrieved from the hospital information system and the pathology records to estimate the ASR and incidence trends based on the M-IHC subtypes of breast cancer. Thus, to guide breast oncologists and health insurance scheme decision-makers, this study aimed to estimate the breast cancer incidence and trends based on the HR and HER2 status. Furthermore, we aimed to project the number of cancer cases and the required expansion of healthcare budget for trastuzumab treatment in 2030.

Materials and methods

Study design and participants

This was a population-based study. The protocol was approved by the Human Research Ethics Committee, Faculty of Medicine, Prince of Songkhla University (REC.63-031-18-1). The requirement for obtaining an informed consent was waived by the ethics committee as the research poses no more than minimal risk to the participants, and the researchers agreed not to disclose the personal identification to third parties. Only patient identification information, which included hospital numbers, was used in this study to merge participants’ data files from two data sources. In the analysis and study report, all data was fully anonymous.

All patients with primary invasive breast cancers who were diagnosed between January 2009 to December 2018 were included in this study. The breast cancer cases were identified from the Songkhla Cancer Registry, a provincial population-based cancer registry, based on the 10th version of the International Classification of Diseases codes combined with the ICD, Oncology version 3 codes. The topography code for breast cancer (C50.x) and behavior code for malignancy (code 3) were selected. Patients with phyllodes tumors (morphology code: 9020) were excluded.

Study setting and data sources

Songkhla province is located in Southern Thailand, between 6°17′-7°56′ N latitude and 100°1′-101°6′ E longitude. The province has an area of approximately 7,394 square kilometers. As of 2019, Songkhla has a total population of 1.4 million people, of which 0.7 million (51.2%) are women [14].

The information on patients with breast cancer was collected from two data sources: the Songkhla Cancer Registry database and medical records of patients in Songklanagarind Hospital. The cancer registry staff extracted the data from the cancer registry database on March 25, 2020, and the researchers collected the data from the patient’s medical records from April 20, 2020, to October 18, 2020.

The Songkhla Cancer Registry is a population-based cancer registry that enrolled cancer patients residing in Songkhla Province. Case ascertainment began in 1988, and patients’ data were regularly supplied to the Cancer Incidence in Five Continents (CIV since vol. VIII) [15]. The cases were identified using active and passive case-finding methods and registered by trained staff. The sources of information included medical records and pathological records of all hospitals in the province. The vital status of the patients in the database is reported when deaths occur in the hospitals and healthcare network and is confirmed by the Bureau of Registration Administration Database, Department of Provincial Administration, Ministry of Interior.

The date of incidence was obtained from the most reliable source. The incidence date was firstly indicated as the date of biopsy, if not available, the date when the specimen was received at the pathological laboratory, or the report date. When the date of definitive pathological diagnosis was not indicated, the date of hospital admission due to this malignancy, radiological diagnosis, and other clinical diagnosis dates related to the occurrence of malignancy or death, death certificate notification only (DCO) were considered. The rules for reporting the incidence date were according to the European Network of Cancer Registries (ENCR) [16], which was also used by the Thai Cancer Registry Network and the International Agency for Research on Cancer (IARC). The IARC allowed missing pathological diagnosis in capturing cancer cases for population-based cancer registration as a low percentage of patients migrating out of the captive area of a cancer registry usually occurred in long-surviving diseases and death might occur before diagnosis in very short surviving diseases.

Songklanagarind Hospital is a super-tertiary hospital in Songkhla Province under the Prince of Songkla University. It has sophisticated diagnostic and therapeutic medical equipment and is the referral center for cancer patients, including those with breast cancer from other hospitals in Songkhla and nearby provinces. In the present study, we included only cases living in Songkhla province and registered in the population-based cancer registry.

Study variables

Most of the variables in this study were extracted from the Songkhla Cancer Registry database. The variables included patient’s demographic data and tumor characteristics. Data on the ER, PR, and HER2 status were collected from the electronic medical records of the Songklanagarind Hospital.

Immunohistochemical staining

The histopathology laboratory of the Songklanagarind Hospital interpreted the HR and HER2 status following the American Society of Clinical Oncology/College of American Pathologists guidelines. Invasive tumors with at least 1% positive nuclei staining for ER or PR were considered positive for HR.

The HER2 status is reported into three groups based on the immunohistochemical (IHC) staining levels: negative, equivocal, and positive. In this study, the equivocal and negative groups were combined into one category.

Population denominators

The Songkhla population projection was used as the population denominator to compute the age-specific incidence rate and ASR from 2010 to 2018. The expected populations were reported by the Office of the National Economic and Social Development Board of Thailand [17]. However, the report did not include population projections of 2009. Hence, we estimated the population denominator of 2009 based on the population censuses in 2000 and 2010, using a log-linear function. The two consecutive censuses were published by the National Statistical Office [18]. The projection data was deposit in the Data Archiving and Networking Services (DANS): https://easy.dans.knaw.nl/ui/datasets/id/easy-dataset:230753 [19].

Statistical analysis

Incidence rates, trends, and projection

To estimate the incidence rates of breast cancer, the disease was classified into four subtypes by performing an M-IHC: 1) ER- or PR-positive and HER2 negative (HR+/HER2−, luminal A-like), 2) ER- or PR-positive and HER2 positive (HR+/HER2+, luminal B-like), 3) ER- and PR-negative, and HER2 negative (HR−/HER2−, triple-negative) and 4) ER- and PR-negative and HER2 positive (HR−/HER2+, HER2-enriched). The age-specific incidence rates in each 5-year age group (0–4, 5–9, 10–14,…, 80–84, and ≥85 years) were calculated using the estimated Songkhla population (as described above), while the ASR was calculated using the world standard population [20] proposed by Segi (1960) and modified by Doll et al. (1966) following the direct method. The count of cases/ estimated Songkhla population in each 5-year age group might result in the age-specific rate of zero in very young age groups.

The effects of age, birth cohort, and year of diagnosis (period) were assessed by fitting a log-linear model to the observed data, assuming a Poisson distribution. The two Poisson models, AP-C and AC-P, where the cohort or period effects were modeled with age-period or age-cohort pairs, were set as an offset, and used to describe the impact of the time component on the incidence trends.

The joinpoint regression analysis was performed to identify the change in incidence trend [21]. The changes in incidence trends over time were described as annual percent changes.

The projection of the incidence rate was extrapolated to 2030 using the joinpoint and age-period-cohort (APC) models. In the joinpoint model, future were projected trends based on the estimated regression coefficients from 2009 to 2018. At the same time, linear interpolation was used to estimate the cohort and period effects of the APC model, which were extrapolated to the future. First, the age-specific incidence rates in each projection year were calculated using age, projected cohort, and projected period effect. Then, if the projected rates deviated from the overall trends, the average was used to smooth the outliers. Subsequently, the ASR of the 5-year interval was computed using the modified Segi world population based on the age-specific incidence rates of the middle-age group in each interval. In addition, the Holford method was used to extract the linear drift from the APC model. Finally, the cut trend concept was applied to our projection by 0% attenuation trends in the first 5 projection years, following the geometric dampening with the factor of 0.92 (1–0.08) per year both in the joinpoint and the APC models [22, 23].

Trastuzumab cost projection

The incremental cost of trastuzumab treatment was projected in 2030 based on three sources of information, including our projection of age-specific incidence rates of the HER2-positive subtype in 18 age groups (0–4, 5–9, 10–14,…, 80–84, and ≥85 years), Thai population projection of 18 age groups in 2030 [17], and pharmacoeconomic evaluation of trastuzumab treatment in Thailand [11].

Based on the APC model mentioned in the previous section, the age-specific incidence rates were projected in 2030. The total estimated number of new HER2-positive cases in 2030 referred to the summation of the new cases in each age group, calculated by multiplying the rate by the projection of Thai women in each stratum.

Then, the maximum number of eligible cases for trastuzumab treatment was estimated based on the proportion of patients with early-stage breast cancer, including those with local and regional stage, in our imputed datasets. Finally, the required additional budget in 2030 due to trastuzumab treatment was approximate based on the results of the budget impact analysis from the pharmacoeconomic study of trastuzumab treatment in Thailand. The results showed that the incremental cost was 475,921 THB per case per year, which was approximately 13,998 USD (1 USD = 34 THB) in 2021.

Management of missing data

Since the HR and HER2 status were not routinely reported together with the histopathological diagnosis of all breast cancer cases, we imputed the missing values of the receptor status using the multivariate imputation by chained equations (MICE) package in R-program [24]. As we predicted the receptor status variables to create the four subtypes of breast cancer, other demographic variables, including age, religion, year of diagnosis, and histopathologic variables, including morphology, grade and stage of tumor, were used to impute the missing HR and HER2 variables. The missing values in the variables were explored other than receptor status, such as grade and stage of the disease, and were found to be less than 22%. Thus, the missing values were set as the “unknown” class in the variables so that the “unknown” values could be used to predict the receptor status in the imputation. This process could avoid unclearly defined strata of the predictor variables, while the missing values in the predictor variables were also imputed during the cycles of chained equations. In addition, the values of one of the three receptor status were used to predict the missing values of the other missing receptor status. The imputation link was set to the logistic regression (logreg) model, and 1,000 datasets were generated. Some of them were known values from the original data, and the other were imputed data filled into the missing (blank) values. The dataset after imputation was deposited in the Data Archiving and Networking Services (DANS): https://easy.dans.knaw.nl/ui/datasets/id/easy-dataset:230753 [19].

The MICE package does not calculate the confidence interval (CI) around the mean of the imputed results for the outcomes of the categorical variables. Therefore, the percentile of the imputed results of the HR status and HER2 receptor status were computed; the results showed that the 2.5th and 97.5th percentiles simulated the 95% probability interval.

Results

Participants and characteristics

Among 2,909 women with breast cancer registered in the Songkhla Cancer Registry, 12 cases with in-situ tumors and 14 cases with phyllodes tumors were excluded. Hence, only 2,883 patients (99.1% of the initial dataset) with breast cancer diagnosed from January 2009 to December 2018 were included in the study.

The median age of the patients at diagnosis of breast cancer was 53.0 years (interquartile range, IQR: 46.0–62.0; range: 24–95), and most of them were diagnosed with regional tumors (57.2%). Ductal carcinoma was the most common tumor type found in 82.4% of the cases; meanwhile, lobular carcinoma was the second most common tumor type and was found in 4.6% of the cases.

HR status was reported in 54.4% of the patients. Of them 1,098 (38.1%) and 883 (30.6%) had ER-positive status and PR-positive status, respectively. HER2 status was reported in approximately half of the patients (50.2%). Positive HER2 status was observed in 319 (11.1%) patients, while the equivocal/negative status was reported in 1,127 (39.1%) patients. In the imputed data were 69.3% and 56.2% of the patients had ER-positive status and PR-positive status, respectively, while only 23.2% of patients had HER2-positive status (Table 1, S1 and S2 Tables).

Table 1

Demographic and tumor characteristics of observed and imputed data.

Variables	Observed data (all = 2883) n (%)	Imputed data (all = 2883) n (%)
Age (median: 53; IQR: 46–62; range: 24–95)
20–29	27 (1.0)	27 (1.0)
30–39	269 (9.3)	269 (9.3)
40–49	807 (28)	807 (28)
50–59	874 (30.3)	874 (30.3)
60–69	545 (18.9)	545 (18.9)
70 and over	361 (12.5)	361 (12.5)
Religion
Buddhist	2508 (87)	2508 (87)
Muslim	350 (12.1)	350 (12.1)
other	8 (0.3)	8 (0.3)
unknown	17 (0.6)	17 (0.6)
Morphology
ductal	2376 (82.4)	2376 (82.4)
lobular	134 (4.6)	134 (4.6)
mixed	70 (2.4)	70 (2.4)
others	93 (3.2)	93 (3.2)
unknown	210 (7.3)	210 (7.3)
Grade
well-differentiated	307 (10.6)	307 (10.6)
moderately- differentiated	1072 (37.2)	1072 (37.2)
poorly- differentiated	861 (29.9)	861 (29.9)
undifferentiated	8 (0.3)	8 (0.3)
unknown	635 (22)	635 (22)
Stage
local	449 (15.6)	449 (15.6)
regional	1648 (57.2)	1648 (57.2)
distant	243 (8.4)	243 (8.4)
unknown	543 (18.8)	543 (18.8)
Estrogen receptor
negative	470 (16.3)	886 (30.7)
positive	1098 (38.1)	1997 (69.3)
unknown	1315 (45.6)
Progesterone receptor
negative	684 (23.7)	1264 (43.8)
positive	883 (30.6)	1619 (56.2)
unknown	1316 (45.6)
HER2-receptor
negative/equivocal	1127 (39.1)	2214 (76.8)
positive	319 (11.1)	669 (23.2)
unknown	1437 (49.8)
Year at diagnosis
2009	206 (7.1)	206 (7.1)
2010	227 (7.9)	227 (7.9)
2011	293 (10.2)	293 (10.2)
2012	221 (7.7)	221 (7.7)
2013	269 (9.3)	269 (9.3)
2014	291 (10.1)	291 (10.1)
2015	296 (10.3)	296 (10.3)
2016	354 (12.3)	354 (12.3)
2017	360 (12.5)	360 (12.5)
2018	366 (12.7)	366 (12.7)

Proportions of breast cancer M-IHC subtypes

The proportions and confidence/probability intervals of the observed and imputed data were comparable for each breast cancer subtype. The HR+/HER2− or luminal A-like subtype was the predominant subtype, accounting for approximately 60% of all cases. The second most common subtype, HR−/HER2− or triple-negative, comprised 17% of all the cases, as shown in Table 2. The subtypes in which the proportion was computed from the observed data tended to regress towards 0.5 (50%) after the imputation process. The estimates and margin of error of the imputed data were calculated using three methods: median and 95% probability interval, mean and Rubin’s estimate of the 95% CI, and mean ±2 standard deviation margins.

Table 2

The proportion of breast cancer subtypes and confidence/probability intervals among the observed and imputed datasets.

Breast cancer Subtypes	Observed data		Imputed dataC
	Number of casesA (%) n = 2883	ProportionB (95%CI)D	Median number of cases (%); (95% PI)E, F	Median proportion and 95%PIE (Quantile)	Mean proportion and 95%CIG (Rubin’s)	Mean proportion ±2 SD
HR+/HER2− (luminal A-like)	880 (30.5)	0.609 (0.584 to 0.635)	1718(59.7); (1642–1789)	0.596 (0.570 to 0.621)	0.596 (0.571 to 0.620)	0.596 (0.570 to 0.621)
HR+/HER2+ (luminal B-like)	150 (5.2)	0.104 (0.088 to 0.120)	318(11.0); (277–374)	0.110 (0.096 to 0.130)	0.111 (0.094 to 0.127)	0.111 (0.094 to 0.128)
HR−/HER2− (triple-negative)	246 (8.5)	0.170 (0.151 to 0.190)	497 (17.3); (447–552)	0.172 (0.155 to 0.191)	0.172 (0.154 to 0.191)	0.172 (0.154 to 0.191)
HR−/HER2+ (HER2-enriched)	168 (5.8)	0.116 (0.100 to 0.133)	347(12.0); (306–400)	0.120 (0.106 to 0.139)	0.121 (0.104 to 0.138)	0.121 (0.104 to 0.138)

A 1,439 subjects with unknown subtype,

B proportion excluding the unknown subtype,

C subtypes estimation based on the known cases and the imputed receptor status cases,

D 95% CI is the confidence interval estimated from the proportion ± Zα/2 x standard error, where α is 0.05,

E 95% PI is the probability interval estimated from 0.025 to 0.975 quantiles.

F total cases are not equal to 2,883 cases because it’s the average number from 1,000 datasets

G 95% CI is the confidence interval estimated by Rubin’s rule for mean (Show the detail in S1 Appendix)

Age-specific incidence rates

Fig 1 shows the age-specific incidence rates per 100,000 women according to the M-IHC subtype. The incidence of all subtypes tended to increase in the 20–24-year age group. The highest peak of the age-specific incidence was observed in the HR+/HER2− (luminal A-like) subtype, reaching 60.2 cases per 100,000 women at the age of 61.4 years (95%CI: 52.3 to 70.6); followed by the HR–/HER2– (triple-negative) subtype with 17.2 cases at the age of 60.4 years (95%CI: 48.9 to 72.0), and the HR–/HER2+ (HER2-enriched) subtype with 13.4 cases at the age of 66.2 years (95%CI: 59.8 to 72.5); meanwhile, the lowest peak was observed in the HR+/HER2+ (luminal B-like) subtype, with 11.3 cases per 100,000 women at the age of 58.7 years (95%CI: 51.1 to 66.3). The rates declined in all subtypes in the older age groups after the age of 70 years. Thus, the incidence of the HR–/HER2+ subtype peaked later than the other three subtypes.

Fig 1

The 5-year mean age-specific incidence rates and 95% confidence interval per 100,000 women of the four breast cancer subtypes according to M-IHC.

HR = hormone receptor; HER2 = human epidermal growth factor receptor 2; HR+/HER2- = luminal A-like; HR+/HER2+ = luminal B-like; HR-/HER2- = triple-negative; HR-/HER2+ = HER2-enriched.

Age-cohort-period analysis of the age-standardized incidence rates

Fig 2 shows the APC analysis of breast cancer incidence by M-IHC subtype. The HR+/HER2– (luminal A-like) and HR+/HER2+ (luminal B-like) subtypes were affected by the birth cohort and time of diagnosis. In the AC-P models (blue lines in Fig 2A and 2B), the younger birth cohorts had a higher rate ratio than the older birth cohort (reference year: 1950). Near the end of the modeling spectrum, the AP-C models exhibited an increase in the rate ratio of the HR+/HER2– subtype since 2014, while the rate ratio of the HR+/HER2+ subtype later increased in 2017 (reference year: 2009). By contrast, the other two subtypes, HR–/HER2– (triple-negative) and HR–/HER2+ (HER2-enriched), were not affected by the birth cohort and year of diagnosis.

Fig 2

The age-period-cohort (APC) trend analysis of the four breast cancer subtypes according to M-IHC in A, B, C, and D.

The three curves in each subfigure, from left to right, represent the incidence rate by age, the rate ratio of incidence by birth-cohort, and year of diagnosis for the reference cohort in blue or reference period in red. The references were the cohort born in 1950 and the diagnosis period of 2009, respectively. Thick lines and the associated thin lines are the three coefficients mentioned previously and 95% confidence intervals. AP-C = Cohort effects modeled with Age-Period component as offset; AC-P = Period effects modeled with Age-Cohort component as offset; HR = hormone receptor; HER2 = human epidermal growth factor receptor 2.

Trends and projections of the age-standardized incidence rates

The HR+/HER2– (luminal A-like) subtype had the highest ASR of breast cancer (Fig 3). Overall, the rate of HR+/HER2– subtypes in each year was twice as high as that of the other subtypes. The ASR increased from 11.8 (95%CI: 9.5 to 14.1) cases per 100,000 women in 2009 to 20.8 (95%CI: 18.0 to 23.6) in 2018. From the joinpoint model, the incidence trends of the HR+/HER2– subtype significantly increased throughout the study period, with an annual percent change of 5.4% (95%CI: 2.5 to 8.3%). As a result, the ASR projection from the joinpoint model was 30.0 cases per 100,000 women in 2030 (Fig 3A), while the APC model projected a lower rate of 29.2 cases (Fig 3B).

Fig 3

The trends and projection of age-standardized rates per 100,000 women with two methods, joinpoint and age-period-cohort models in A and B of the four breast cancer subtypes according to M-IHC in four different colors.

The incidence rate (y-axis) is on a log scale. Dots connected with thin solid lines in subfigures A and B represent the calculated ASRs from 2009 to 2018. Thick solid lines represent the smooth/modeled ASRs. The extended dashed lines represent the projected trends of ASRs in 2019–2030. HR = hormone receptor; HER2 = human epidermal growth factor receptor 2.

Throughout the study period, the ASRs of other subtypes were not greater than 5.5 cases per 100,000 women. However, the rising trends in the incidence of HR+/HER2+ (luminal-B) breast cancer were remarkable. The HR+/HER2+ subtype surged from being the least to the second most common subtype, with the highest annual percent change rate of 10.1% (95%CI: 4.9 to 15.5%). The ASR of this subtype in 2009 was 2.6 per 100,000 women. It was projected to increase to 8.8 and 10.4 cases per 100,000 women in 2030 by the joinpoint and the APC models, respectively.

By contrast, the ASRs of HR–/HER2+ (HER2-enriched) and HR–/HER2– (triple-negative) subtypes were rather stable over time. The annual percent changes of the two subtypes were 1.3% (95%CI: –2.2% to 4.9%) and –2% (95%CI: –4.7% to 0.8%), respectively.

Additional cost projection of trastuzumab treatment

The overall ASR of HER2-positive (HR+/HER2+ and HR–/HER2+) breast cancer increased from 10–11 in 2020 to 13–14 per 100,000 women in 2030 (Fig 3). This would lead to an increase in the national healthcare budget for breast cancer treatment as a part of the adjuvant therapy recommended for HER2-positive patients [9, 10]. We postulated by applying our projected rates of the HER2-positive subtype to the 2030 Thailand population projection [17]. Our forecast expected an increase of 10.8% in HER2-positive cases per year. The number of cases would reach 5,020 patients in 2020 and 10,450 in 2030. Our imputed data showed that the local and regional stages accounted for 68% of all HER2-positive cases. Thus, the expected HER2-positive eligible patients for the NLEM indication in 2030 would be 7,100. The additional cost of trastuzumab treatment was calculated under the condition that all who need the drug are treated, using the cost information based on a study sponsored by the Health Intervention and Technology Assessment Program [11]. The result illustrated an incremental cost of at least 99.4 million USD (1 USD = 34 THB).

Discussion

In our population-based, cross-sectional study from 2009 to 2018, the most common tumor was the HR+/HER2− (luminal A-like) subtype; meanwhile, the HR+/HER2+ (luminal B-like) subtype was the less frequent subtype in the early years but became the second most common subtype at the end of the study period. The incidence of all subtypes except HR−/HER2+ (HER2-enriched) peaked at the age of 50 years, while the latter peaked at 60 years. The rates of all subtypes declined significantly after the age of 70 years.

The trends in the ASR of breast cancer were distinct according to the HR status. The incidence of HR-positive subtypes (HR+/HER2− and HR+/HER2+) increased over time. The most significant change was observed in those with HR+/HER2+ cancer. As shown in Fig 2A and 2B, the trends of HR-positive subtypes were affected by both birth cohort and period of diagnosis. The effect of period on the two subtypes increased beyond null (1) at the end of the study period in the AP-C models. The same finding was observed in the cohort effects in both subtypes. Such co-existence of both results in the same population would suggest an increased risk of being diagnosed in younger generations and later years.

Thailand has promoted breast cancer screening programs since 2002, covering over 90% of the Thai population [3]. An analysis of six provincial cancer registries in Thailand reported the effect of introducing a country-wide self-breast examination screening program [3]. It increased the detection of breast cancer among individual aged >80 years. However, the detectable rate declined several years after that because few prevalent cases were reported. Our study also found a decline in the age-specific incidence rates of all breast cancer subtypes after the age of 70 years, which is consistent with the report of a previous study.

Reproductive factors have differential effects on the trends of HR-positive breast cancer. An increase in the risk of HR-positive breast cancer was reported among women with an early age at menarche, a long period between menarche and the first delivery, fewer children, and older age at menopause, while the effect of breastfeeding history on HR-positive cancer remains controversial [25-27]. However, the 6-month exclusive breastfeeding rate in Thailand has changed over time. A series of national surveys reported that the rate was 14.5% in 2005 and increased to 23.1% in 2016, which was the highest; however, it sharply decreased to 14.0% in 2019 [28].

The reproductive characteristics of Thai women have been changing due to exposure to higher endogenous estrogen, such as younger age at menarche [29], later age at the first marriage [30], and a fall in the total fertility rate [31] from 4.9 in 1974–1976 to a fertility replacement rate of 2.1 in 1990 and down to 1.6 in 2011. These phenomena may explain the rise in the incidence of HR-positive breast cancer. Thai adolescents experienced a social dynamism in reproductive behavior, which possibly interfered with the overall change in internal estrogen exposure in this age group. From 2000 to 2012, the rate of teenage pregnancy has been increasing in Thai society [32], while the abortion rate due to unwanted pregnancies is high [33]. Such social effects on the reproductive age might have contributed to the incidence of HR-positive breast cancer.

Exogenous hormone exposure, particularly combined estrogen/progesterone hormone replacement therapy, may have also positively increased the risk of HR-positive subtypes [34, 35]. Two studies on female hormone use in Thailand demonstrated no association between exogenous hormonal use and breast cancer risk. However, these two studies did not specify the HR subtype of breast cancer [36, 37].

Elevated body mass index (BMI) is also associated with an increased risk of HR-positive breast cancer, especially in postmenopausal women [26, 38, 39]. The national surveys reported that the prevalence of overweight and obesity (BMI ≥ 25 mg/m2) among Thai women has increased [40-42]. The prevalence increased from 25.1 in 1994 to 34.4 and 41.8 kg/m2 in 2004 and 2014, respectively. In comparison, the prevalence trends of overweight and obesity in the Southern Thai women were 25.4, 36.3, and 43.7 kg/m2 in the same three years, respectively. Thus, the increasing trends of BMI in southern Thai women seemed to be higher than the average BMI of Thai women. This may partly explain the rising trends in HR-positive breast cancer in Thailand.

The increase in HR-positive subtypes, particularly in the young birth cohorts and recent periods, was also evident in Malaysian women [43] and Western populations [44-49]. The worldwide changes in reproductive factors demonstrated in our discussion, and the increased BMI in many countries are the possible underlying cause of the changes in the trends in HR-positive subtypes; however, the etiologic pathways are not well understood.

The change in the ASR percentage of breast cancer by M-IHC classification impacts human resources and medical infrastructure reallocation. For example, approximately 700 new cancer cases per medical oncologist (MO) were reported in Thailand in 2020 [50, 51], while the ratio in the United States and European countries was around 100–300 cases per MO since 2015 [52]. Therefore, to meet the ratio of fewer than 300 cases per MO, additional 365 MOs are needed in Thailand; however, as of 2020 only 268 MOs were available. The number of MOs also depends on the differences in sociocultural cost and health and medical care infrastructure in each country. Such differences imply a disparity in the healthcare system but not an inequity in people’s rights to healthcare access.

Our study has some limitations. First, there was a high level of missingness in the data on ER, PR, and HER2 status. The occurrence of missingness was primarily due to the non-intention to undergo further chemotherapy and radiotherapy; therefore, the degree of missingness had no backward correlation with the degree of positivity of the receptor status. We imputed missing data assuming missingness was at random and we properly modeled missing values. The missingness of receptor status was found in patients who underwent surgery at other hospitals and were referred to receive adjuvant therapy in Songklanagarind Hospital. Again, the missing data seemed to be random and not biased toward the risk status of treatable subtype. Second, inferring the results to other provinces in Southern Thailand seems possible because of the population demographics. However, the inference of this study to women in different regions of Thailand could be affected by the difference in the genetic background and associated risk factors of breast cancer that might be slightly non-homogeneous throughout the country.

In conclusion, the incidence trends of HR-positive and HER2-positive breast cancer in Thailand have been increasing, particularly in young birth cohorts and recent periods of diagnosis. The rising trends in breast cancer incidence direct the future health care budget, human resources, and medical care facilities. Their shortage in the field of cancer affects patient care quality and may influence the prognosis and survival, as well as the disparity and equity in healthcare access. Therefore, healthcare providers and administrators should prepare an appropriate plan to anticipate the situation of breast cancer shortly forecasted in this study.

Demographics and tumor characteristics of observed data stratified by receptor status.

(DOCX)

Click here for additional data file.

Demographics and tumor characteristics stratified by receptor status after imputation of unknown receptor status.

(DOCX)

Click here for additional data file.

The calculated margin of error of imputed datasets by three methods.

(DOCX)

Click here for additional data file.

6 Oct 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Sophie Pilleron, PhD

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf.

2. In your ethics statement in the Methods section and in the online submission form, please provide additional information about the data used in your retrospective study. Specifically, please ensure that you have discussed whether all data were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

3. Thank you for providing the date(s) when patient medical information was initially recorded. Please also include the date(s) on which your research team accessed the databases/records to obtain the retrospective data used in your study.

4. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

Additional Editor Comments (if provided):

Thank you for submitting your manuscript to Plos One. The topic is of high importance and the method seem adequate.

However I have some comments in addition to those of the reviewer.

Major points:

I would suggest authors to add more details about the cancer registry and how a cancer case is defined. Could you tell be more about the population covered by the registry? Does it cover all the province? The authors mentioned that the registry reports « both population-based and hospital-based statistics ». This is unclear. How do a cancer case is defined? What is the date of incidence considered? How did authors consider cancer diagnosis based on death certificate only (DCO)?

In this analysis, did authors apply an age criteria for inclusion?

Also, I really like the discussion part about number of new cancer cases and estimated costs (lines 335-344). This would deserve to be part of the results part and to be presented as a secondary objective. This is an interesting point that may be of interest for policy makers. I would highlight that point. If authors decide to do so, they need to explain about they estimate the nb of new cancer cases in 2030 and how they estimate the excess costs in the method section.

The paragraph at lines 345-353 is very interesting.

Why did not the authors use the Rubin’s rules to combine their estimates?

Because authors studied trends in cancer incidence, I would like to see a comment in the discussion on the possible change (or not) of recording system of new cancer cases. Also, authors showed that there is a decrease in incidence from the age of 70. Would not it be explained by a lower ascertainment of cancer cases at older ages? Older adults may be less likely to go through cancer diagnosis work-ups because of their age, the presence of other conditions, or frailty. This should be mentioned as possible reason.

Another possible reason for changes in incidence is breastfeeding practices? Have they changed over the period covered by the study?

Minor points:

I would suggest to replace 10e5 by 100,000 as it is easier to read.

I strongly encourage authors to have an English-native speaker to check the grammar and sentence structure as there are several parts of text that are unclear. For instance, lines 223-231, 243, 259, 264-266,  196-299, 331-333, 355-359, 363. I encourage authors to reformulate.

The introduction may be shortened, in particular, the paragraph on racial and cultural disparities. Because the study is about incidence, the paragraph on treatment does not seem relevant in the introduction.

Authors mentioned cancer stage several times but they did not define which staging system they used. I would then suggest to add the definition in the method section.

Lines 304-305: is not tautological?

Lines 311-312: authors mentioned an increase in teenage pregnancy. Would not it explain by a better reporting?

Lines 315-317: I would invite caution with this interpretation.

Lines 317: Are not there any data at all?

Lines 320-323: Could authors include some figures in the text? What are the prevalences?

Figures: I would recommend to add confidence intervals around curves.

Authors used both relative risk and rate ratios. I advise authors to pick one and stick with it throughout the manuscript. However, relative risk and rate ratios are not similar. Authors should probably clarify which one is estimated by APC models.

Line 109: what means stage 0 or extent 1?

Line 118: authors wrote « [the registry] collects cancer patients diagnosed or treated ». This is unclear. How does the cancer registry collect data on new cancer cases?

Line 161: Which period did authors consider as « recent »?

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors present the first population-based cancer registry data to summarize molecular-immune-histochemical subtypes. Overall, the authors provide a very rigorous analysis to interpret trends in hormone receptor status of primary breast cancer cases from 2009 to 2018. This is an important research topic, but I have some concerns, mainly related to the level of missing in the data.

Introduction

I would suggest a little reorganizing of this paragraph “Since 2002, Thailand has implemented the universal health coverage (UHC) scheme [10]. It is the basic scheme for Thai citizens, covered health promotion, disease screening, treatment of basic and high-cost diseases under the Thai national list of essential medicines 5 (NLEM), and palliative/supportive care. The UHC scheme contributed approximately 79% of the Thai population’s health insurance [11].”. Mainly, I would recommend mentioning that 79% of the population’s health insurance is the basic covered should be mentioned after the fact it is the basic scheme for Thai citizens. I would also recommend switching covered to covers as the UHC appears to still be in place.

The authors stated “Though a targeted therapy is cost-effective among early breast cancer cases, it increases the burden of cancer care costs as a whole and significantly contributes to the national health budget, especially when early patients are more detected“. I am not exactly following your logic on how it would significantly increase the burden of cancer care costs as a whole if it is a cost-effective treatment. Can you please elaborate?

The authors state “Other targeted therapeutic drugs for breast cancer have been or are in Thai FDA”. Can you please provide a few examples?

Methods

The authors mention that they excluded DCIS cases. Yet, DCIS can also be receptor positive. Can the authors explain why this did not include this group?

Management of missing data

The authors report using multiple imputation to account for missing outcome data, but more information would be valuable. Specifically, you mention that the missing in other variables is 22%, but what is the missing for your receptor statuses? This seems like a much bigger concern. It is also important to know how much change this imputation process had on your findings. Mainly, what does the results look like in the non-imputed dataset? How different is the population?

I am also a little confused by the process. Did you enter the other missing variables as missing or did you use the imputed predictor variables to determine?

Results

Can you please include the range of age at breast cancer diagnosis in addition to the IQR?

It would also be good to see some comparison of demographics by HR status.

From reading the results section, it appears that almost 50% of the receptor status was missing. This is of great concern, especially given that there was an additional 22% of other missing variables. There really needs to be comparisons in the outcome and characteristics of those with and without receptor status, and the new sample with the imputed data to understand the impact.

It is great to see that you included observed data versus imputed data in the results section. I would ask that you include percent in the column “Number of cases”. I still think there needs to be a comparison of how the demographics of the women looked before and after imputation.

Discussion

The sentence “In contrast, reproductive factors have an inconclusive influence on HR-negative cancers.” Requires a reference

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

23 Dec 2021

RESPONSE TO EDITOR AND REVIEWER

RESPONSE TO THE EDITOR:

MAJOR POINTS:

COMMENT: I would suggest authors to add more details about the cancer registry and how a cancer case is defined. Could you tell be more about the population covered by the registry? Does it cover all the province? The authors mentioned that the registry reports « both population-based and hospital-based statistics ». This is unclear. How do a cancer case is defined? What is the date of incidence considered? How did authors consider cancer diagnosis based on death certificate only (DCO)?

RESPONSE: We added more detail about the Songkhla cancer registry and revised the unclear sentences (Study setting and data sources paragraph: 3-4).

“The Songkhla Cancer Registry is a population-based cancer registry that enrolled cancer patients residing in Songkhla Province. Case ascertainment began in 1988, and patients’ data were regularly supplied to the Cancer Incidence in Five Continents (CIV since vol. VIII) [15]. The cases were identified using active and passive case-finding methods and registered by trained staff. The sources of information included medical records and pathological records of all hospitals in the province. The status of the patients in the database is reported when deaths occur in the hospitals and healthcare network and is confirmed by the Bureau of Registration Administration Database, Department of Provincial Administration, Ministry of Interior.

The diagnosis date of cancer patients was obtained from the most reliable source, if possible. The diagnosis date was indicated as the date of biopsy, the date when the specimen was received at the pathological laboratory, or the report date. When the date of definitive diagnosis provided by the pathological laboratory was not indicated, the date of hospital admission due to this malignancy and other dates related to the occurrence of malignancy or death were considered [16].”

COMMENT: In this analysis, did authors apply an age criteria for inclusion?

RESPONSE: We included all breast cancer cases who were diagnosed in the study periods. However, from our data, the youngest patient who met our criteria was 24 years old. That is why we set the age group for calculating the age-specific incidence from age 20 to ≥85 years.

COMMENT: Also, I really like the discussion part about number of new cancer cases and estimated costs (lines 335-344). This would deserve to be part of the results part and to be presented as a secondary objective. This is an interesting point that may be of interest for policy makers. I would highlight that point. If authors decide to do so, they need to explain about they estimate the nb of new cancer cases in 2030 and how they estimate the excess costs in the method section.

RESPONSE: We decided to set the projection of additional cost due to the trastuzumab treatment in 2030 as the secondary objective of our study. We explained how to estimate the excess cost in the methodology part as subtopic “Trastuzumab cost projection” and added the outcome in the result part under the subtopic “Additional cost projection of trastuzumab treatment.”

COMMENT: Why did not the authors use the Rubin's rules to combine their estimates?

RESPONSE: In our analysis, we calculated the estimates and margin of error of the imputed data by three methods, median and its 95% probability interval, mean, and Rubin’s estimate of the 95% confidence interval and mean ±2 SD margins. All three methods gave very close values of the estimates and the margins of error. Thus, in the previous draft, we decided to show only the median proportion and 95%PI, but we forgot to explain the similarity of the results. So, in this revision, we showed the results from all methods in Table 2. (Table 2 and Results part in subtopic: Proportions of breast cancer M-IHC subtypes)

COMMENT: Because authors studied trends in cancer incidence, I would like to see a comment in the discussion on the possible change (or not) of recording system of new cancer cases. Also, authors showed that there is a decrease in incidence from the age of 70. Would not it be explained by a lower ascertainment of cancer cases at older ages? Older adults may be less likely to go through cancer diagnosis workups because of their age, the presence of other conditions, or frailty. This should be mentioned as possible reason.

RESPONSE: One study from six provincial cancer registries in Thailand found an effect of introducing a country-wide self-breast examination screening program under the universal health insurance, which covered over 90% of the Thai population increase capturing of the hidden breast cancer cases aged over 80. The capture rate declined after some years of the program's start even it has continued as many cases have been captured. So, we think that we captured the decline in the ascertainment rate during our study period and is becoming stabilized after the end of our study. In the paragraph, we discussed that “Thailand has promoted breast cancer screening programs since 2002, covering over 90% of the Thai population [3]. An analysis of six provincial cancer registries in Thailand reported the effect of introducing a country-wide self-breast examination screening program [3]. It increased the detection of breast cancer among individual aged >80 years. However, the detectable rate declined several years after that because few prevalent cases were reported. Our study also found a decline in the age-specific incidence rates of all breast cancer subtypes after the age of 70 years, which is consistent with the report of a previous study.” (Discussion: paragraph 3).

COMMENT: Another possible reason for changes in incidence is breastfeeding practices? Have they changed over the period covered by the study?

RESPONSE: We added the detail about the change in the breastfeeding rate in Thailand over time in the revised version (Discussion: paragraph 4).

“However, the 6-month exclusive breastfeeding rate in Thailand has changed over time. A series of national surveys reported that the rate was 14.5% in 2005 and increased to 23.1% in 2016, which was the highest; however, it sharply decreased to 14.0% in 2019 [27].”

27. Topothai C, Tangcharoensathien V. Achieving global targets on breastfeeding in Thailand: gap analysis and solutions. Int Breastfeed J. 2021 Dec;16(1):38. doi: 10.1186/s13006-021-00386-0. PMID: 33962645.

MINOR POINTS:

COMMENT: I would suggest to replace 10e5 by 100,000 as it is easier to read.

RESPONSE: We agree with the editor's comment and change text "105" to "100,000" in our revised version.

COMMENT: I strongly encourage authors to have an English-native speaker to check the grammar and sentence structure as there are several parts of text that are unclear. For instance, lines 223-231, 243, 259, 264-266, 196-299, 331-333, 355-359, 363. I encourage authors to reformulate.

RESPONSE: Before resubmission, we sent the revised manuscript to the Editage for checking the grammar and sentence structure.

COMMENT: The introduction may be shortened, in particular, the paragraph on racial and cultural disparities. Because the study is about incidence, the paragraph on treatment does not seem relevant in the introduction.

RESPONSE: We decided to delete the paragraph that describes the racial variations in Thailand. However, we remained the paragraph that described the treatment option in breast cancer patients. It shows why the knowledge of breast cancer incidence in terms of subtype is essential. In addition, it relates to our secondary objective that we added in the revised version.

COMMENT: Authors mentioned cancer stage several times but they did not define which staging system they used. I would then suggest to add the definition in the method section.

RESPONSE: Songkhla cancer registry records the breast cancer stage according to the AJCC-TNM classification. In a long-running cancer registry, cancer staging may change over time. The registries usually reclassify the extent of disease to the summary stage according to the SEER staging system. It categorized tumor stage to in situ, localized, regionalized, distant, and unknown stage.

COMMENT: Lines 304-305: is not tautological?

RESPONSE: We revised the sentences in Line 304-305 and described this issue in more detail as “However, the 6-month exclusive breastfeeding rate in Thailand has changed over time. A series of national surveys reported that the rate was 14.5% in 2005 and increased to 23.1% in 2016, which was the highest; however, it sharply decreased to 14.0% in 2019 [27].” (Discussion: paragraph 4).

COMMENT: Lines 311-312: authors mentioned an increase in teenage pregnancy. Would not it explain by a better reporting?

RESPONSE: We changed the last part of the discussion on reproductive behaviors in the Thai population to lessen the stress on teenage pregnancy and HR-positive breast cancer. The detail is, “Thai adolescents experienced a social dynamism in reproductive behavior, which possibly interfered with the overall change in internal estrogen exposure in this age group. From 2000 to 2012, the rate of teenage pregnancy has been increasing in Thai society [31], while the abortion rate due to unwanted pregnancies is high [32]. Such social effects on the reproductive age might have contributed to the incidence of HR-positive breast cancer.”

In addition, we also changed reference no. 31 (previous version no.29) to “Sukrat B. Thailand Adolescent Birth Rate: Trend and Related Indicators. Thai J Obstet Gynaecol. 2014 Jan 20;15–21”. (Discussion: paragraph 5)

COMMENT: Lines 315-317: I would invite caution with this interpretation.

RESPONSE: We revised the sentences, described them in more detail. The new text is “Exogenous hormone exposure, particularly combined estrogen/progesterone hormone replacement therapy, may have also positively increased the risk of HR-positive subtypes [33,34]. Two studies on female hormone use in Thailand demonstrated no association between exogenous hormonal use and breast cancer risk. However, these two studies did not specify the HR subtype of breast cancer [35-36].”

In addition, we changed the reference to “Type and timing of menopausal hormone therapy and breast cancer risk: individual participant meta-analysis of the worldwide epidemiological evidence. The Lancet. 2019 Sep;394(10204):1159–68”. (Discussion: paragraph 6)

COMMENT: Lines 317: Are not there any data at all?

RESPONSE: Of course. We found only a few studies [35,36] on female hormone use and breast cancer risk. However, they are not specific on breast cancer subtypes or receptor status. They did not find the association between hormone use and breast cancer risk. We described more in the revised version. (Discussion: paragraph 6)

35. Poosari A, Promthet S, Kamsa-ard S, Suwanrungruang K, Longkul J, Wiangnon S. Hormonal Contraceptive Use and Breast Cancer in Thai Women. J Epidemiol. 2014 May 5;24(3):216–20.

36. Ratanawichitrasin A, Bhodhisuwan K, Reansuwan W, Kongpatanakul S, Ratanawichitrasin S. Risk of Breast Cancer in Post-Menopausal Women Using Hormone Replacement Therapy. J Med Assoc Thai. 2002 May;85(5):583-9.

COMMENT: Lines 320-323: Could authors include some figures in the text? What are the prevalences?

RESPONSE: We added more detail to the paragraph that described the BMI and positive-breast cancer risk and prevalence of obesity in Thai women as “The national surveys reported that the prevalence of overweight and obesity (BMI ≥ 25 mg/m2) among Thai women has increased [39-41]. The prevalence increased from 25.1 in 1994 to 34.4 and 41.8 kg/m2 in 2004 and 2014, respectively. In comparison, the prevalence trends of overweight and obesity in the Southern Thai women were 25.4, 36.3, and 43.7 kg/m2 in the same three years, respectively. Thus, the increasing trends of BMI in southern Thai women seemed to be higher than the average BMI of Thai women. This may partly explain the rising trends in HR-positive breast cancer in Thailand.” (Discussion: paragraph 7).

COMMENT: Figures: I would recommend to add confidence intervals around curves.

RESPONSE: We added the 95% confidence interval in Fig 1 following the editor’s suggestion and change y-scale from log- to normal scale.

COMMENT: Authors used both relative risk and rate ratios. I advise authors to pick one and stick with it throughout the manuscript. However, relative risk and rate ratios are not similar. Authors should probably clarify which one is estimated by APC models.

RESPONSE: We already clarified that the RR is the rate ratio like we notice in figure 2. We already revised the error text in our manuscript.

COMMENT: Line 109: what means stage 0 or extent 1?

Line 118: authors wrote « [the registry] collects cancer patients diagnosed or treated ». This is unclear. How does the cancer registry collect data on new cancer cases?

Line 161: Which period did authors consider as « recent »?

RESPONSE: Line 109: We included only invasive tumors, added “invasive” (subtopic: Study design and participants, paragraph 2) So, we can delete “in situ tumor (code as stage 0 or extent 1)”.

Line 118: We already added the detail about the cancer registry. (Setting and data sources: paragraph 3-4).

Line 161: we specified the period when we used the coefficients to project the future trend “coefficients from 2009-2018”. (Subtopic: Incidence rates, trends, and projection, paragraph 4)

RESPONSE TO REVIEWER 1:

COMMENT: The authors present the first population-based cancer registry data to summarize molecular-immune-histochemical subtypes. Overall, the authors provide a very rigorous analysis to interpret trends in hormone receptor status of primary breast cancer cases from 2009 to 2018. This is an important research topic, but I have some concerns, mainly related to the level of missing in the data.

RESPONSE: Our analysis found high percentages of missing data on ER, PR, and HER2 status, 45.6, 45.6, and 49.8%, respectively. Literature reviews found that multiple imputation could reduce bias and improved effect estimation at the high proportion of missing data. It is especially when imputation was conducted based on the auxiliary information, including variables predicting the complete set of the hormone and HER2 receptor statuses in the model. Thus, the replacement of the cases with missing values of the ER, PR, and HER2 statuses was random and conformed to the original construct of the association of the non-missing cases. The maximum of missing proportions that can hold by the multiple imputation method is different. However, the evidence shows that the multiple imputation by chained equation (MICE) method can handle up to 80% of missingness [3].

References.

1. Eisemann N, Waldmann A, Katalinic A. Imputation of missing values of tumour stage in population-based cancer registration. BMC Med Res Methodol. 2011 Dec;11(1):129.

2. Madley-Dowd P, Hughes R, Tilling K, Heron J. The proportion of missing data should not be used to guide decisions on multiple imputation. J Clin Epidemiol. 2019 Jun;110:63–73.

3. Souverein OW, Zwinderman AH, Tanck MWT. Multiple Imputation of Missing Genotype Data for Unrelated Individuals. Annals of Human Genetics. 2006;70(3):372–81.

COMMENT:

Introduction

I would suggest a little reorganizing of this paragraph "Since 2002, Thailand has implemented the universal health coverage (UHC) scheme [10]. It is the basic scheme for Thai citizens, covered health promotion, disease screening, treatment of basic and high-cost diseases under the Thai national list of essential medicines 5 (NLEM), and palliative/supportive care. The UHC scheme contributed approximately 79% of the Thai population's health insurance [11].".

Mainly, I would recommend mentioning that 79% of the population's health insurance is the basic covered should be mentioned after the fact it is the basic scheme for Thai citizens. I would also recommend switching covered to covers as the UHC appears to still be in place.

RESPONSE: We revised the sentence sequencing in the paragraph, “Since 2002, Thailand has implemented a universal health coverage (UHC) scheme [7]. This is the primary insurance scheme for Thai citizens. As of 2020, the UHC scheme has already provided health services to approximately 79% of the Thai population [8]. It covers health promotion, disease screening, treatment of basic and high-cost diseases under the Thai National List of Essential Medicines (NLEM), and palliative/supportive care.” following the suggestion from the reviewer. (Introduction: paragraph 3)

COMMENT:

The authors stated "Though a targeted therapy is cost-effective among early breast cancer cases, it increases the burden of cancer care costs as a whole and significantly contributes to the national health budget, especially when early patients are more detected ".

I am not exactly following your logic on how it would significantly increase the burden of cancer care costs as a whole if it is a cost-effective treatment. Can you please elaborate?

The authors state "Other targeted therapeutic drugs for breast cancer have been or are in Thai FDA". Can you please provide a few examples?

RESPONSE:

Though a targeted therapy is cost-effective among early breast cancer cases, it increases the sum of cancer care budget by 15560 USD per one patient to add a targeted therapy to the to the baseline treatment while the QALYs were expected to increase by 4.59 years. While an increased burden to the national health budget especially when early patients are more detected, in turn, it implies that the longer life gain to women benefited from a targeted therapy would make a return in terms of the GDP per capita greater than the budget spent by the UC scheme for the treatment cost per patient.

In addition, we added information about the other targeted drug that may include in the Thai national list of essential medicines (NLEM) in the future (Introduction: paragraph 6). It is the reason to show that why the estimated incidence in each breast cancer subtype is essential. The additional sentences are “Other targeted drugs approved by the Thai Food and Drug Administration for breast cancer treatment include pertuzumab and ribociclib. Currently, these are the only two drugs available to patients who are covered by the civil servant medical benefit scheme [12,13].”

COMMENT:

Methods

The authors mention that they excluded DCIS cases. Yet, DCIS can also be receptor-positive. Can the authors explain why this did not include this group?

RESPONSE: In our population-based cancer registry of Songkhla, it was not designed to collect in situ cases of breast cancer from the start of the registration since the standard of data collection for DCIS tumors is nonhomogeneous in hospitals in the province. It is possible in the future to collect this cancer at the best completeness after the province-wide data collection procedures are standardized.

COMMENT:

Management of missing data

The authors report using multiple imputation to account for missing outcome data, but more information would be valuable. Specifically, you mention that the missing in other variables is 22%, but what is the missing for your receptor statuses? This seems like a much bigger concern. It is also important to know how much change this imputation process had on your findings. Mainly, what does the results look like in the non-imputed dataset? How different is the population? I am also a little confused by the process. Did you enter the other missing variables as missing or did you use the imputed predictor variables to determine?

RESPONSE: In our study, we imputed only the missing values in outcome variables, which are ER, PR, and HER2 status. The proportion of missing values in each variable were 45.6, 45.6, and 49.8%, respectively. For the question that which of the missing values were addressed in the imputation, we transformed variables other than the three receptor statuses to factors and set the missing values to an ‘unknown’ class of that variable, so that the ‘unknown’ values have a predictive ability on the receptor statuses in the imputation. We also used the values of one of the three receptor statuses to predict the missing values of the other missing receptor statuses. The final set of the dataset consisted of 1000 imputed data set, a part of them were the known values from the original data, another part of them are imputed data filled into the missing (blank) values.

In Table 1, we added the characteristics of ER, PR, and HER2 status. For example, the numbers of HER2 negative/equivocal, positive, unknown were 1127, 319, 1437, after imputation, the mean of 1000 imputed data set were 2214, 669, and 0, while the SD of the imputed dataset were 38, 38 and 0. After we got the imputed values of the three receptor statuses, the Table 2 described the mean/median of the observed and imputed data with the range of prediction by the three methods. The difference in the observed and imputed proportion is on the third or second decimal places.

In addition, in our revised manuscript, we added the comparison baseline characteristics between observed and imputed data, as the Table 1 in the result part, and added a new table in the supplement (S1A and S1B Table).

COMMENT:

Results

Can you please include the range of age at breast cancer diagnosis in addition to the IQR?

RESPONSE: The range of age was 24-95 years. We added the range of age in table1.

COMMENT:

Results

It would also be good to see some comparison of demographics by HR status.

From reading the results section, it appears that almost 50% of the receptor status was missing. This is of great concern, especially given that there was an additional 22% of other missing variables. There really needs to be comparisons in the outcome and characteristics of those with and without receptor status, and the new sample with the imputed data to understand the impact.

It is great to see that you included observed data versus imputed data in the results section. I would ask that you include percent in the column "Number of cases". I still think there needs to be a comparison of how the demographics of the women looked before and after imputation.

RESPONSE: We added the percentage in column "Number of cases" of Table 2. The detail about the missing proportion and how we imputed the missingness is described in the previous comment. In addition, we added the baseline characteristic comparison between observed and imputed data according to the comment in table 1 and additional 2 tables (S1A S1B Table) in the supplement.

COMMENT:

Discussion

The sentence "In contrast, reproductive factors have an inconclusive influence on HR-negative cancers." Requires a reference

RESPONSE: We decided to revise this paragraph and delete the sentence according to the editor’s comment, making it a clearer comparison. From “In contrast, reproductive factors have an inconclusive influence on HR-negative cancers.” to “The reproductive characteristics of Thai women have been changing due to exposure to higher endogenous estrogen, such as younger age at menarche [28], later age at the first marriage [29], and a fall in the total fertility rate [30] from 4.9 in 1974–1976 to a fertility replacement rate of 2.1 in 1990 and down to 1.6 in 2011. These phenomena may explain the rise in the incidence of HR-positive breast cancer. Thai adolescents experienced a social dynamism in reproductive behavior, which possibly interfered with the overall change in internal estrogen exposure in this age group. From 2000 to 2012, the rate of teenage pregnancy has been increasing in Thai society [31], while the abortion rate due to unwanted pregnancies is high [32]. Such social effects on the reproductive age might have contributed to the incidence of HR-positive breast cancer.” (Discussion: paragraph 5)

Submitted filename: Response to reviewers.docx

Click here for additional data file.

25 Jan 2022

7 Feb 2022

Response to Reviewers

Comment for PONE-D-21-21989R1

Journal Requirements:

Suggestion: Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Response:

In manuscript version 2 (PONE-D-21-21989R1), revised from manuscript version 1(PONE-D-21-21989):

1. We removed the references from manuscript version 1 because the previous references were not relevant after we revised the text in the manuscript. The lists of removed references are as:

Hays J. People of Thailand: origin, different Thai groups and Siamese twins | facts and details. [cited 15 Jun 2021]. Available from: http://factsanddetails.com/southeast-asia/Thailand/sub5_8c/entry-3209.html

Srisontisuk DS, Katchamat P, Pakdee P. Poverty and the ethnic minority groups in Thailand. J Mekong Society. 2005;1: 151–189.

Reid LA. Benedict’s Austro-Tai hypothesis—an evaluation. Asian Perspect. 1984;26: 19–34.

Aekplakorn W, Inthawong R, Kessomboon P, Sangthong R, Chariyalertsak S, Putwatana P, et al. Prevalence and trends of obesity and association with socioeconomic status in Thai adults: national health examination surveys, 1991–2009. J Obes. 2014;2014: 410259. doi:10.1155/2014/410259. PMID: 24757561.

Sakboonyarat B, Pornpongsawad C, Sangkool T, Phanmanas C, Kesonphaet N, Tangthongtawi N, et al. Trends, prevalence and associated factors of obesity among adults in a rural community in Thailand: serial cross-sectional surveys, 2012 and 2018. BMC Public Health. 2020;20. doi:10.1186/s12889-020-09004-w. PMID: 32493314.

2. We added some references in manuscript version 2 because we added the new text in manuscript 2. The lists of added references are as:

Drug and Medical Supply Information Center, Ministry of Public Health. Medical Reimbursement Criteria for Cancer and Hematology Patients Who Need Expensive Drugs (Additional) (W 339, W 340) [Internet]. [cited 2021 Oct 29]. Available from: http://dmsic.moph.go.th/index/detail/7823

Drug and Medical Supply Information Center, Ministry of Public Health. Medical Reimbursement Criteria for Cancer and Hematology Patients Who Need Expensive Drugs (Additional) (W 278, W 279) [Internet]. [cited 2021 Oct 29]. Available from: http://dmsic.moph.go.th/index/detail/8221

Bray F, Colombet M, Ferlay J, Mery L, Piñeros M, Znaor A, et al. Cancer Incidence in Five Continents Volume XI [Internet]. [cited 2021 Oct 31]. Available from: https://publications.iarc.fr/Book-And-Report-Series/Iarc-Scientific-Publications

European Network of Cancer Registries (ENCR).pdf [Internet]. [cited 2021 Oct 31]. Available from: https://www.encr.eu/sites/default/files/pdf/incideng.pdf

Topothai C, Tangcharoensathien V. Achieving global targets on breastfeeding in Thailand: gap analysis and solutions. Int Breastfeed J. 2021 Dec;16(1):38. doi: 10.1186/s13006-021-00386-0. PMID: 33962645.

Poosari A, Promthet S, Kamsa-ard S, Suwanrungruang K, Longkul J, Wiangnon S. Hormonal Contraceptive Use and Breast Cancer in Thai Women. J Epidemiol. 2014 May 5;24(3):216–20.

Ratanawichitrasin A, Bhodhisuwan K, Reansuwan W, Kongpatanakul S, Ratanawichitrasin S. Risk of Breast Cancer in Post-Menopausal Women Using Hormone Replacement Therapy. J Med Assoc Thai. 2002 May;85(5):583-9.

Thai National Health Examination Survey 1994-1995, NHES I [Internet]. 1st ed. Nonthaburi: Health Systems Research Institute (HSRI); 1996 [cited 2021 Oct 31]. 271 p. Available from: https://www.hiso.or.th/hiso/picture/reportHealth/report/report5.pdf

Porapakkham Y, Boonyaratapan P, editors. Thai National Health Examination Survey 2003-2004, NHES III [Internet]. 1st ed. Nonthaburi: Health Systems Research Institute (HSRI); 2006 [cited 2021 Oct 31]. 267 p. Available from: https://www.hiso.or.th/hiso/picture/reportHealth/report/report2.pdf

Aekphakorn W, Pakjaroen H, Thaikla K, Satheannoppakao W. Thai National Health Examination Survey 2013-2014, NHES V [Internet]. 1st ed. Nonthaburi: Health Systems Research Institute (HSRI); 2016 [cited 2021 Oct 31]. 283 p. Available from: https://www.hiso.or.th/hiso/picture/reportHealth/report/report9.pdf

3. We change some of the references from manuscript version 1 as following lists:

Ref no 31. was changed from “UNFPA Thailand. [Where do teen mothers live in Thailand?]. [cited 22 May 2021]. Available from: https://thailand.unfpa.org/en/publications” to “Sukrat B. Thailand Adolescent Birth Rate: Trend and Related Indicators. Thai Journal of Obstetrics and Gynaecology. 2014 Jan 20;15–21.”.

The new reference was written in English, while the previous one was written in Thai. However, the information was similar.

Ref no 34. changed from “Kim S, Ko Y, Lee HJ, Lim J. Menopausal hormone therapy and the risk of breast cancer by histological type and race: a meta-analysis of randomized controlled trials and cohort studies. Breast Cancer Res Treat. 2018;170: 667–675. doi:10.1007/s10549-018-4782-2. PMID: 29713854.” to “Collaborative Group on Hormonal Factors in Breast Cancer. Type and timing of menopausal hormone therapy and breast cancer risk: individual participant meta-analysis of the worldwide epidemiological evidence. The Lancet. 2019 Sep;394(10204):1159–68. doi: 10.1016/S0140-6736(19)31709-X. PMID: 31474332.”.

The new reference is more relevant and more recent than the previous reference.

In manuscript version 3, (revised from manuscript version 2; PONE-D-21-21989R1):

We checked our references (from manuscript version 2). We revised the URL of reference no.18 from “http://web.nso.go.th/en/census/poph/cen_poph.htm” to “http://songkhla.old.nso.go.th/nso/project/search/index.jsp?province” because the previous URL cannot access presently. However, the information is the same as with the previous reference.

We revised the format of Ref no. 12, 13, 14, 40, 41, and 49 as the mark in “Revised Manuscript with Track Changes” file.

Additional Editor Comments:

Suggestion: I would advise deleting "and encourage all oncologists and health policymakers to manage breast cancer cases in Thailand." in line 20 as it seems not the right place to make this call.

Line 119: April 20, 2020, October 18, 2020. —> April 20, 2020, to October 18, 2020.

Line 125: The status of the patients —> the vital status of the patients

Line 128: The diagnosis date of cancer patients —> The date of incidence.

Response: We already revised the text in line No. 20, 119, 125 and 128 following the editor suggestion. (Revised version: Line 20, 119, 125, 129, and 130)

Suggestion: Lines 129-130: Because all these dates may not be the same, please, could you be more specific on how you choose the date? The earlier available?

Response: The incidence date was specified following the diagnosis date of cancer patients in the cancer registry database. The date of incidence was obtained from the most reliable source. The incidence date was firstly indicated as the date of biopsy, if not available, the date when the specimen was received at the pathological laboratory, or the report date. When the date of definitive pathological diagnosis was not indicated, the date of hospital admission due to this malignancy, radiological diagnosis, and other clinical diagnosis dates related to the occurrence of malignancy or death were considered. We already revised the paragraph. (Materials and Methods: Study setting and data sources, paragraph: 4)

Suggestion: Line 137: Does it means that the cancer registry can include patients from other province too? If so, the population covered is not only that of the Songkhla Province. Am I right?

Response: The Songkhla cancer registry is a population-based cancer registry. It includes only the cancer patients who are citizens of Songkhla province. The patients are identified from all hospitals in Songkhla province. Songklanagarind hospital is a hospital in Songkla province. The cancer patients of Songklanarind hospital who are the Songkhla citizen are registered to the Songkhla cancer registry and the hospital-based cancer registry of Songklanagarind hospital. While cancer cases, who are not Songkhla citizens, are not registered in the Songkla cancer registry. They are registered only in the hospital-based cancer registry of Songklanagarind hospital. We added more detail as “The hospital has been running both hospital-based CR the Songklanagarind hospital and population-based CR of Songkhla province. Subjects in this study were extracted from the population-based cancer registry.” (Materials and Methods: Study setting and data sources, paragraph: 5)

Suggestion: Line 431: what do you mean? Missing at random?

Response: Yes, we mean the missing at random. We already revised the sentence to more clear as “…In the MICE process, the missing process does not violate the missing at random assumption…”. (Discussion, paragraph 10)

Suggestion: Line 435: I am not sure to understand the link between this sentence and the following one.

Response: Southern Thai women slightly differ from women in other regions regarding genetic and cultural backgrounds. Though, we thought that the inference of this study to all Thai women was still relevant with no solid evidence to validate it.

Reviewers' comments:

Suggestion:

Introduction:

Please add a reference to the first line “Breast cancer is the most common cancer among women worldwide”.

Response: We already added the reference to that sentence. We wrote the sentence based on reference number 1. (Introduction, paragraph: 1)

Suggestion:

Methods:

If you one method to determine malignancy is based on death, wouldn’t this mean there may be an underestimation of the true prevalence of the disorder? I assume not everyone who dies undergoes an autopsy. This should be addressed in the limitations section.

Response: We would like to apologize not to address the diagnosis based on death certificate in the limitation section.

The percentage of the death certificate only (DCO) cases of Songkhla cancer registry was reported around 0.5% in the Cancer Incidence in Five Continents (CI5). Usually, the International Agency for Research on Cancer (IARC) doubts the quality of data when the DCO is zero as the registry might collect cases from pathology sources only. And none of the diagnosis was from autopsy. We are not sure how much pathology sources only bias occurred in diagnosis of breast cancer in the registry. However, the inclusion of cases diagnosed from death certificates ensures a better estimate of the incidence. The cancer registries model the prevalence of cancer based on the incidence and survival, not the count of existing cases in the registry. Since we were estimating the incidence of breast cancer by receptor status, the presence of DCO in cancer diagnosis would not affect the imputation process.

Reference: the International Agency for Research on Cancer (IARC). Indices of data quality of the Cancer Incidence in Five Continents (CI5) volume X. [cited 4 Feb 2022]. Available from: https://ci5.iarc.fr/CI5I-X/old/vol10/I_09.pdf

Suggestion:

You mention that your youngest case of breast cancer is in a women who is 24 years of age, yet use 5 year age groups to calculate the age-specific rates. What is the number of individuals who are 24 years of age? This group is likely very small, and unstable. How does the estimate change if you collapse the 24 year old into the upper age group?

Response: The number of individuals who are 24 years of age, were 2 cases from 2,883 cases. It estimated 0.1% of total cases.

The age-specific rates were calculated to the unit of cases per 100,000 population in a particular 5-year age group according to Boyle P. & Parkin D.M. in which age-specific rates were calculated from age group of 0–4 to 85+, which the formula as following.

a_i= (r_i⁄n_i )×100 000

Where ai is an age-specific rate per 100,000 population in each five-year age group, ri is the number of cases in the same five-year age group, and ni is the corresponding person-years of the observation.

The ultimate aim of stratifying into five-year strata is to calculate the age-standardized incidence rates (ASR) from the formula as follows:

ASR= ( ∑_(i=1)^A▒〖a_i w_i 〗)⁄(∑_(i=1)^A▒w_i )

Var (ASR) = (∑_(i=1)^A▒〖[a_i w_i^2 (100 000-a_i 〗)/n_i])⁄〖(∑_(i=1)^A▒w_i )〗^2

Where ai is an age-specific rate per 100,000 population in each five-year age group, wi is world standard population in the corresponding five-year age group, and ni is the person-years of sample population in the same five-year age group.

The variance of age-specific rate was still low in very young age groups in which the number of patients were small, while the person-years in the sample population were large, see Figure 1.

Reference: P. Boyle and D. M. Parkin, “Statistical Methods for Registries,” In: O. M. Jensen, D. M. Parkin, R. MacLennan, C. S. Muir and R. G. Skeet, Eds., Cancer Registration: Principles and Methods, IARC Scientific Publication No. 95, International Agency for Research on Cancer, Lyon, 1991, pp. 126-158.

Suggestion:

I was a little confused by this statement “including our projection of age-specific incidence rates of the HER2- 194 positive subtype in 18 age groups (0–4, 5–9, 10–14, ..., 80–84, and ≥85 years)”. How is this possible if you have the first case of cancer starting at 24 years? Wouldn’t the relative cost for each of these other groups be zero?

Response: The age-specific rates were calculated as described above. Even the age-groups with no case have their associated person-years (ni) and the 95%CI could be calculated by the following formula

ASR ± Z_(∝/2)× (s.e.(ASR))

s.e.(ASR)= √(Var (ASR))

The 95%CIs of age-specific rate in the very young age groups were narrow, since the number of patients were small, while the person-years in the sample population were large.

Suggestion:

Results:

Please include range of age in this sentence “The median age of the patients at diagnosis of breast cancer was 53.0 years 237 (interquartile range, IQR: 46.0–62.0)”.

Response: We already included the age range in this sentence. The new sentence is “The median age of the patients at diagnosis of breast cancer was 53.0 years (interquartile range, IQR: 46.0–62.0; range: 24–95),…” (Result: Participants and characteristics, paragraph: 2).

Suggestion:

Please also extend the table for age to show each 10 year age group from 20 to give an idea of the instability in estimates for the younger age groups.

Response: We already revised table1 to show each 10 ages group, starting from 20-29, 30-39, until 70 and over. (Table 1)

Suggestion:

Figure 3 title says 105, please change to 100,000. Same in the sentence “Over the study period, the ASRs of other subtypes were not greater than 5.5 cases per 105 women” And “. The ASR of this subtype in 2009 was 2.6 per 105 women.” And here “The overall ASR of HR-positive and HER2-positive (HR+/HER2+ and HR-/HER2+) breast cancer increase from 30-32 and 10-11 to 39 and 13-14 per 105 women in 2020 to 2030”

Response: We did not find the sentence that that reviewer suggested to revise in the second version (PONE-D-21-21989R1) of our manuscript. All of these sentences were in the first version (PONE-D-21-21989)? We already revised all of these sentences. (Figure 3 title, Result: Trends and projections of the age-standardized incidence rates, paragraph: 2)

Submitted filename: Response to Reviewers .docx

Click here for additional data file.

8 Feb 2022

A rapid rise in hormone receptor-positive and HER2-positive breast cancer subtypes in Southern Thai women: a population-based study in Songkhla.

PLOS ONE

Dear Dr. Sriplung,

Thank you for submitting your manuscript to PLOS ONE.

There are some additional minor points I would like you to address before accepting the manuscript for publication. Please see below.

Please submit your revised manuscript by Mar 25 2022 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

1 Mar 2022

Response to Reviewers

comment on PONE-D-21-21989R3

We added the information of our data deposit in the Data Archiving and Networking Services (DANS) in subtopic “Population denominators” and “Management of missing data, paragraph 1”, and we added a reference No. 19 in the manuscript as “19. Chuaychai, A. 1000 imputed data set of receptor status for breast cancer. 2021. Data Archiving and Networking Services (DANS). https://doi.org/10.17026/dans-xn5-5286”.

1. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Response: We already added captions of the supporting information files at the end of our manuscript.

comment on PONE-D-21-21989R2

Additional Editor Comments (if provided):

suggestion:

Line 178: (0–4, 5–9, 10–14, ..., 80–84, and ≥85 years) —> 20-25, …, ≥85 as you don’t have cases under 20 years old.

Same comment for line 207 as you don’t have cases under 20 so it will be zero anyway.

response: Yes, the cases under 20 year age group were zero cases in the result. However, in the method part, we wrote 18 age groups, including age group under 20 years, because our inclusion criteria included all cases and age groups. That is why we wrote 18 age-group in the methods part. We added the sentence “The count of cases/ estimated Songkhla population in each 5-year age group might result in the age-specific rate of zero in very young age groups.”. (Statistical analysis: Incidence rates, trends, and projection, paragraph: 1)

suggestion:

Lines 144-147: I appreciate that authors added details I suggested. However I would suggest another formulation: Instead « The hospital has been running both hospital-based of the Songklanagarind hospital and population-based cancer registry of Songkhla province. Subjects in this study were extracted from the population-based cancer registry. » I suggest: « In the present study, we included only cases living in Songkhla province and registered in the population-based cancer registry. »

response: We already revised following editor suggestion. (Study setting and data sources, paragraph: 5)

suggestion:

Line 224: instead of « manage », I would suggest using « impute ». This could give something such as « we imputed missing values of the receptor status using MICE R package. » I would also suggest mentioning clearly which variables you used to impute.

response: We already revised following editor suggestion. (Management of missing data, paragraph: 1)

suggestion:

Line 265-267: I would suggest deleting : « Since the raw dataset contained missing values for receptor status in approximately half of the cases, the MICE package was used to assign values to the “unknown” receptor status and classify the patients into four subtypes described in the Method section. » as it is not a result per se and was already described in the method section.

response: We already revised following editor suggestion. (Proportions of breast cancer M-IHC subtypes, paragraph: 1)

suggestion:

Line 441: I appreciate authors answered my comment. However, the sentence is not correct. You cannot say for sure that missing data pattern is at random. You can only assume it as it is not really verifiable. Author can consider something like: « We imputed missing data assuming missingness was at random and we properly modeled missing values. » instead of « In the MICE process, the missing process does not violate the missing at random assumption. »

response: We already revised following editor suggestion. (Discussion, paragraph: 10)

suggestion:

Authors did not really answer the comment of Reviewer 2 regarding diagnosis made via death certificate. The reviewer requested that authors acknowledge that some cancer cases may have been missed since not everyone has an autopsy. In addition, I would suggest authors to add %DCO in the method section where they mention it.

response: We added the information about the rules for reporting the incidence date of population-based cancer registry as “The rules for reporting the incidence date were according to the European Network of Cancer Registries (ENCR) [16], which was also used by the Thai Cancer Registry Network and the International Agency for Research on Cancer (IARC). The IARC allowed missing pathological diagnosis in capturing cancer cases for population-based cancer registration as a low percentage of patients migrating out of the captive area of a cancer registry usually occurred in long-surviving diseases and death might occur before diagnosis in very short surviving diseases.”. (Study setting and data sources, paragraph: 4)

Submitted filename: Response to Reviewers_26022022.docx

Click here for additional data file.

2 Mar 2022

A rapid rise in hormone receptor-positive and HER2-positive breast cancer subtypes in Southern Thai women: a population-based study in Songkhla.

PONE-D-21-21989R3

Dear Dr. Sriplung,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Sophie Pilleron, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

The manuscript would deserve to be check for English language.

Reviewers' comments:

9 Mar 2022

PONE-D-21-21989R3

A rapid rise in hormone receptor-positive and HER2-positive breast cancer subtypes in Southern Thai women: a population-based study in Songkhla.

Dear Dr. Sriplung:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Sophie Pilleron

Academic Editor

PLOS ONE