Seasonal Variation in Reporting of Venous Thromboembolism in a Spontaneous Adverse Event Reporting System: Analysis of the Japanese Adverse Drug Event Report Database

Abstract

Background

Venous thromboembolism (VTE) is a major cardiovascular disorder associated with high morbidity and mortality. Several epidemiological studies have reported seasonal variations in VTE, with higher rates during winter. However, it remains unclear whether reports of VTE adverse events in spontaneous reporting systems exhibit similar seasonal patterns.

Material and Methods

A retrospective pharmacovigilance study was conducted using data from the Japanese Adverse Drug Event Report (JADER) database. Reports of VTE, including deep vein thrombosis and pulmonary embolism, were identified from the adverse event records of the JADER database. Monthly reporting rates were calculated by dividing the number of VTE reports by the total number of adverse event reports. Seasonal variations were evaluated using negative binomial regression models with monthly indicator variables and harmonic seasonal terms. The association between monthly VTE reporting rates and the mean monthly temperature in Tokyo was assessed using Pearson’s correlation analysis.

Results

VTE adverse event reports showed a clear seasonal pattern. In the harmonic seasonal model, the estimated peak occurred in December and the nadir in July, with a peak-to-nadir rate ratio of 1.29. The harmonic seasonal model showed a better fit than the monthly indicator model (Akaike information criterion: 1696.8 vs 1712.6). Monthly VTE reporting rates were inversely correlated with mean monthly temperature in Tokyo (Pearson’s r = −0.769, p = 0.003).

Conclusion

VTE adverse event reports in the JADER database showed a clear seasonal pattern, with higher and lower reporting rates in winter and summer, respectively. Seasonal and environmental factors may be associated with seasonal variation in VTE reporting. It should be noted, however, that spontaneous reporting data cannot be used to estimate the absolute incidence of VTE or to establish causal relationships between drugs and adverse events. Further studies using complementary data sources are required to confirm these findings.

Introduction

Venous thromboembolism (VTE), which includes deep vein thrombosis (DVT) and pulmonary embolism (PE), is a major cardiovascular disease associated with substantial morbidity and mortality worldwide. The annual incidence of VTE in the general population is estimated to be approximately 1–2 per 1,000 person-years in Europe and the United States,1 whereas its incidence in Asia is generally considered to be lower. In South Korea, VTE incidence was estimated at approximately 0.2 per 1,000 person-years.2 Furthermore, a nationwide epidemiological survey in Japan identified 16,096 newly diagnosed cases of PE and 24,538 cases of DVT in 2011, corresponding to annual incidences of 12.6 and 19.2 per 100,000 population, respectively.3 These observations are consistent with previous reports suggesting lower VTE incidence in East Asian populations than in Western populations. Several explanations have been proposed for the lower reported VTE rates in Asia relative to Western populations, including underdiagnosis, limited clinical recognition, and the rarity of thrombophilic conditions prevalent among Caucasians, notably factor V Leiden and prothrombin gene mutations.4 VTE is associated with a substantial risk of recurrence and long-term complications, such as post-thrombotic syndrome and chronic thromboembolic pulmonary hypertension. The established risk factors for VTE include advanced age, immobility, surgery, malignancy, and inherited thrombophilia.5 In addition to these clinical risk factors, several medications have been reported to increase the risk of VTE. For example, hormonal therapies, certain anticancer agents, and some immunomodulatory drugs contribute to thrombotic events through various mechanisms, including alterations in coagulation pathways, endothelial dysfunction, and pro-inflammatory effects.6 Therefore, understanding the reporting patterns and characteristics of VTE adverse events is important for pharmacovigilance and drug safety monitoring.

Accumulating epidemiological evidence suggests that the incidence of VTE varies seasonally. Several population-based studies have reported higher incidence rates in winter and lower incidence rates in summer.7–10 For example, a meta-analysis by Zhao et al8 reported that the incidence of VTE was significantly higher in winter than in summer (relative risk = 1.12, 95% confidence interval [CI]: 1.01–1.24). The pooled seasonal incidence was 27.2% in winter and 24.6% in summer. Although the mechanisms underlying this seasonal pattern are not fully understood, environmental and physiological factors, such as lower ambient temperature, reduced physical activity, hemoconcentration, and seasonal changes in coagulation and inflammatory responses have been proposed as potential contributors.7–10 Despite these observations, whether VTE adverse event reports in spontaneous reporting systems exhibit similar seasonal patterns remains unclear. Identifying seasonal variations in VTE adverse event reports may provide insights into the potential interactions between pharmacological effects and environmental influences on thrombotic risk.

Spontaneous adverse event reporting systems are valuable resources for evaluating drug safety at the population level. The Japanese Adverse Drug Event Report (JADER) is a national pharmacovigilance database that collects reports of suspected adverse drug reactions in Japan, and it has been widely used to detect and evaluate drug safety signals.11 Although spontaneous reporting systems are subject to underreporting and reporting bias, previous analysis of the JADER database has reported seasonal variations in certain adverse events,12–14 suggesting that long-term aggregated pharmacovigilance data may allow the detection of temporal and seasonal patterns. However, seasonal patterns in VTE adverse event reports have not been systematically investigated. Therefore, the present study aimed to evaluate seasonal variations in VTE adverse event reports using the JADER database.

Material and Methods

Data Source

Data were obtained from the JADER database maintained by the Pharmaceuticals and Medical Devices Agency (PMDA). Adverse event report data have been publicly available through the PMDA website (https://www.pmda.go.jp) since April 2004 and can be accessed free of charge. The JADER is a spontaneous reporting system that collects adverse event reports submitted by healthcare professionals and pharmaceutical companies in Japan. It consists of four tables (“DEMO,” “DRUG,” “REAC,” and “HIST”), which include patient demographic information, drug information, adverse events, and primary diseases, respectively. In the present study, data from the REAC table were used for analysis.

Report Selection and Outcome Definition

This study included adverse event reports registered in the JADER database between April 2004 and August 2025. Reports were included if the onset date of the adverse event was available, at least at the monthly level. Reports with onset dates before April 2004 were excluded. The unit of analysis was adverse event reports recorded in the REAC table, rather than individual cases. VTE events were identified using two specific MedDRA/J preferred terms (PTs): Deep vein thrombosis (PT code: 10051055) and Pulmonary embolism (PT code: 10037377), according to MedDRA/J version 28.1J. These PTs were selected a priori because VTE is clinically defined as DVT and PE. Broader Standardized MedDRA Queries for embolic and thrombotic events were not applied, as this study specifically targeted VTE defined as DVT and PE. Each REAC record corresponding to either PT was counted as a separate VTE report. Therefore, multiple VTE-related PTs reported within the same case, including concomitant DVT and PE reports, were counted independently and were not deduplicated at the case level. The monthly number of VTE reports was calculated as the total number of eligible DVT and PE reports. The total number of adverse event reports submitted to JADER during the same month was also calculated.

Statistical Analysis

All statistical analyses were performed using R software (ver. 4.5.2., R Foundation for Statistical Computing, Vienna, Austria). Negative binomial regression models were fitted using the glm.nb function in the MASS package. Statistical significance was defined as a two-sided p-value of <0.05.

Monthly Reporting Rate

The total number of adverse event reports per month was included as an offset variable to account for fluctuations in the total number of adverse event reports over time. This approach allowed the estimation of the reporting rate of VTE relative to the total number of adverse event reports, rather than the incidence of VTE in the underlying population. Using the total number of adverse event reports as an offset is a commonly used approach in pharmacovigilance studies to adjust for variations in the overall reporting volume over time. As reporting patterns may have been influenced by the coronavirus disease 2019 (COVID-19) pandemic,15–17 a binary indicator variable representing the COVID-19 period (January 2020 to May 2023) was included in the regression models. This period was defined based on the declaration of a public health emergency of international concern by the World Health Organization in January 2020 and the reclassification of COVID-19 to a Category V infectious disease under Japanese law in May 2023, which marked a significant transition in the pandemic response in Japan.

Seasonality in VTE reporting was evaluated using negative binomial regression models with a log-link function to account for overdispersion in the count data. The monthly number of VTE reports was modeled while adjusting for temporal trends and the overall volume of adverse event reports. The logarithm of the total number of adverse event reports (AE_total_t) in each month was included as an offset term to account for fluctuations in reporting volume. The exponentiated coefficients were interpreted as reporting rate ratios (RRRs) with the corresponding 95% CIs.

Model 1: Month Indicator Model

Seasonality was first evaluated using categorical indicator variables for each calendar month. June was selected as the reference category because preliminary descriptive analyses indicated the lowest reporting frequency around this month. The reference month was selected based on descriptive examination of the data and should therefore be regarded as data driven. The model included calendar time, month indicator variables, the COVID-19 period indicator, and the logarithm of the AE_total_t as an offset term (Supplementary Methods S1).

Model 2: Harmonic Seasonal Model

To characterize seasonality using a smooth periodic function, harmonic terms based on sine and cosine functions with a 12-month periodicity18 were incorporated into the negative binomial regression model. The model additionally included calendar time, the COVID-19 period indicator, and the logarithm of the AE_total_t as an offset term (Supplementary Methods S1). Residual autocorrelation diagnostics were examined using the autocorrelation function of Pearson residuals from the harmonic seasonal model.

Estimation of the Seasonal Peak and Nadir

The timing of the seasonal peak and nadir was estimated from the coefficients of the harmonic seasonal model. Seasonal amplitude was quantified using the peak-to-nadir rate ratio derived from the fitted harmonic curve (Supplementary Methods S1).

Sensitivity Analysis

A sensitivity analysis was conducted using only reports with adverse event onset dates up to December 2019, excluding the COVID-19 period.

Association with Ambient Temperature

To explore the relationship between seasonal variations in VTE reporting and ambient temperature, the mean monthly temperature in Tokyo was obtained from publicly available meteorological data provided by the Japan Meteorological Agency (https://www.jma.go.jp/jma/indexe.html). Because nationwide temperature data covering the entire study period were not available in a consistent monthly format, the mean monthly temperature in Tokyo was used as a representative indicator of ambient temperature. The monthly mean temperatures from 2004 to 2024 were averaged for each calendar month. The monthly VTE reporting rate was calculated for each calendar month by dividing the number of VTE reports by the total number of adverse event reports. The association between the monthly reporting rate of VTE and mean monthly temperature was evaluated using Pearson’s correlation coefficient.

Ethics Approval

Ethics approval was waived because this study was based on a publicly available database containing fully anonymized data and did not involve direct access to identifiable personal information, in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects issued by the Ministry of Health, Labour and Welfare of Japan.

Results

Study Population and Reporting Trends

During the study period from April 2004 to August 2025, 1,008,591 adverse event reports were included in the analysis. Among these, 4,825 VTE reports were identified. These reports originated from 4,080 individual cases, as multiple VTE-related preferred terms, including concomitant DVT and PE reports, could be reported within the same case (Figure 1). Among these 4,080 individual cases, 55.22% were aged 60 years or older, and 65.25% were female. Detailed demographic characteristics are presented in Supplementary Table S1.

Figure 1.

Flowchart of report selection for the analysis of venous thromboembolism reports using the Japanese Adverse Drug Event Report database.

Long-Term Trend

In the regression models, the VTE reporting rate increased slightly over time (RRR = 1.002, 95% CI: 1.002–1.003, p = 6.68×10⁻¹¹).

Impact of the COVID-19 Period

During the COVID-19 period, the VTE reporting rate was lower than that during the pre-pandemic period (RRR = 0.847, 95% CI: 0.736–0.974, p = 0.017).

Model 1: Month Indicator Model

In the negative binomial regression model, which treated calendar month as a categorical variable with June as the reference category, several winter months exhibited higher VTE reporting rates. In particular, the reporting rate was significantly higher in January (RRR = 1.339, 95% CI: 1.067–1.680, p = 0.012) and December (RRR = 1.350, 95% CI: 1.076–1.694, p = 0.010). Higher reporting rates were also observed in February, March, April, and November (Table 1).

Table 1.

Rate Ratios for Venous Thromboembolism Reporting by Calendar Month Estimated from a Negative Binomial Regression Model (with June as the Reference Month)

Month	Reporting Rate	RRR (95% CI)	P-Value
January	0.00535	1.339 (1.067–1.680)	0.012
February	0.00512	1.298 (1.033–1.613)	0.025
March	0.00504	1.268 (1.011–1.591)	0.040
April	0.00491	1.273 (1.017–1.595)	0.036
May	0.00410	1.068 (0.850–1.341)	0.573
June	0.00407	Reference
July	0.00427	1.048 (0.834–1.316)	0.689
August	0.00465	1.132 (0.902–1.421)	0.287
September	0.00472	1.163 (0.924–1.463)	0.199
October	0.00494	1.241 (0.989–1.558)	0.062
November	0.00527	1.286 (1.026–1.614)	0.030
December	0.00542	1.350 (1.076–1.694)	0.010

Abbreviations: RRR, reporting rate ratio; CI, confidence interval.

Model 2: Harmonic Seasonal Model

The harmonic seasonal model, including sine and cosine terms, demonstrated a seasonal pattern in VTE reporting. The estimated seasonal peak occurred in December, whereas the seasonal nadir occurred in July (Figure 2). The peak-to-nadir rate ratio was 1.29, indicating that the VTE reporting rate during the seasonal peak was approximately 29% higher than that during the seasonal nadir.

Figure 2.

Seasonal pattern of venous thromboembolism reporting estimated using the harmonic seasonal model. The solid line represents the predicted monthly venous thromboembolism reporting rate estimated using a negative binomial regression model with harmonic terms (sine and cosine with a 12-month period), showing a peak in December and a nadir in July. The shaded area indicates the 95% confidence interval of the predicted rate. Filled circles represent the observed venous thromboembolism reporting rates aggregated by calendar month across the study period.

Model Comparison

Model fit was compared using the Akaike information criterion (AIC). The harmonic seasonal model showed a better fit than the monthly indicator model (AIC = 1696.8 vs 1712.6), suggesting that a smooth seasonal function adequately described seasonal variation in VTE reporting.

Model Diagnostics

For the harmonic seasonal model, the estimated overdispersion parameter (theta) was 11.68 (standard error (SE) = 1.75). The estimated sine and cosine coefficients are provided in Supplementary Table S2. Residual autocorrelation diagnostics based on Pearson residuals are presented in Supplementary Figure S1.

Sensitivity Analysis

In the sensitivity analysis restricted to reports with adverse event onset dates up to December 2019, seasonal patterns remained largely unchanged.

Association with Ambient Temperature

The monthly reporting rate of VTE showed a significant inverse correlation with the mean monthly temperature in Tokyo (Pearson’s r = −0.769, 95% CI: −0.932 to −0.35, p = 0.003), indicating higher reporting rates during colder months (Figure 3).

Figure 3.

Correlation between mean monthly temperature in Tokyo and monthly venous thromboembolism reporting rates. The solid line represents the fitted linear regression line, and the shaded area indicates the 95% confidence interval. Each point represents the mean monthly venous thromboembolism reporting rate and the corresponding mean monthly temperature.

Discussion

In this study, we investigated the seasonal patterns of VTE adverse event reports using JADER, a large-scale spontaneous adverse event reporting database. Our analysis revealed a clear seasonal variation in VTE adverse event reports, with a peak in December and a nadir in July. We found that the reporting rate at the seasonal peak was approximately 29% higher than that at the nadir. In addition, an inverse correlation was observed between the monthly reporting rate of VTE adverse events and ambient temperature, with higher reporting rates during colder months.

Seasonal variations in VTE have been reported in numerous epidemiological studies, which consistently show a winter predominance and a summer nadir in VTE incidence.7–10 The seasonal pattern observed in the present study closely mirrors previously reported findings. This consistency suggests that VTE adverse event reports may reflect some of the same seasonal influences observed in the general population. VTE occurs preferentially in specific high-risk clinical populations, including patients with malignancy, those undergoing surgery, individuals with prolonged immobility, and patients receiving certain pharmacotherapies associated with increased thrombotic risk.5 Several biological and behavioral mechanisms have been proposed to explain seasonal variation in VTE. Cold temperatures are associated with increased blood viscosity, enhanced platelet activation, and peripheral vasoconstriction, all of which may contribute to a hypercoagulable state.19 In addition, reduced physical activity during colder months may increase venous stasis, an important component of the VTE pathophysiology.20 Consistent with these observations, we observed a significant inverse correlation between VTE reporting rates and the mean monthly temperature in Tokyo. Although dehydration during hot weather may theoretically increase thrombotic risk, epidemiological studies have consistently demonstrated a winter predominance of VTE incidence,7–10 suggesting that seasonal variation in VTE is likely influenced by multiple interacting biological and behavioral factors rather than by ambient temperature alone. However, because this analysis was based on a nationwide spontaneous reporting database and a single-city temperature measure, the temperature analysis should be considered as exploratory, and should not be interpreted as evidence of a direct causal relationship between ambient temperature and VTE reporting. Previous studies have demonstrated the feasibility of evaluating seasonal patterns in adverse event reports using the JADER database, including analyses of adverse events associated with SGLT2 inhibitors and diuretic-induced dehydration.13,14 However, the seasonal patterns observed in those studies differed from those identified in the present analysis, suggesting that the mechanisms underlying seasonal variation may vary across adverse event types. Therefore, seasonal patterns observed in spontaneous reporting databases should be interpreted within the biological and clinical context of each adverse event.

Although spontaneous reporting systems have not been designed to estimate disease incidence, the seasonal pattern observed in this study is unlikely to be explained solely by fluctuations in reporting activity. Seasonal differences in healthcare utilization and reporting practices may also contribute to the observed patterns. The analysis was adjusted for the total number of adverse event reports by using an offset term, thereby accounting for temporal changes in the overall reporting volume. Moreover, the seasonal pattern identified in the present analysis closely corresponds to the winter predominance of VTE reported in population-based epidemiological studies. These findings suggest that spontaneous reporting databases can still capture the underlying seasonal signals of adverse events, although they are not suitable for direct measurement of incidence in the exposed population. Furthermore, findings derived from spontaneous reporting databases should not be interpreted as evidence of treatment effects or comparative effectiveness, as such databases are intended primarily for signal detection rather than causal inference. The lower VTE reporting rate observed during the COVID-19 period may reflect changes in healthcare utilization and adverse-event reporting behavior during the pandemic rather than a true reduction in VTE occurrence. Increased attention to COVID-19-related conditions, changes in reporting priorities, and disruptions in routine clinical practice may have influenced reporting patterns in spontaneous reporting systems.

In recent years, analyses using spontaneous adverse event reporting databases have increasingly used time-to-onset analyses to evaluate the temporal relationship between drug exposure and adverse events.21–23 The seasonal variations observed in this study suggest that seasonality may influence the occurrence or reporting of VTE adverse events. Therefore, seasonal factors should be considered when interpreting time-to-onset analyses of VTE adverse event reports in pharmacovigilance studies. Furthermore, the observed association between VTE reporting and ambient temperature suggests that environmental conditions may influence thrombotic risk. Considering the ambient temperature when monitoring adverse events may provide useful information for clinical pharmacovigilance and patient management. In particular, increased vigilance during colder months may be beneficial for early detection and prevention of VTE adverse event reports.

Our study has several strengths. First, we analyzed a large number of adverse event reports collected over more than two decades, enabling the evaluation of long-term seasonal patterns. Second, we applied a harmonic seasonal model to estimate the timing and magnitude of seasonal variations in order to provide a smooth and biologically plausible representation of seasonal trends. Third, we explored the association between seasonal variations and ambient temperature as an exploratory analysis, providing additional context for the observed seasonal pattern in VTE reporting. In addition, seasonal variation was evaluated based on adverse event onset dates rather than reporting dates, which may better reflect the actual timing of VTE occurrence and reduce the potential bias related to reporting delays in spontaneous reporting systems.

However, this study also had some limitations. Spontaneous reporting systems, including the JADER database, are subject to underreporting, reporting bias, duplicate entries, variability in data quality, and temporal changes in reporting behavior.24 In addition, the number of adverse events reported does not necessarily reflect the true incidence of VTE. A major limitation of the present study is that the analysis was based on reported adverse events rather than on individual patient cases. As spontaneous reporting databases record adverse events as individual reports, multiple reports may originate from the same patient or a clinical episode. Therefore, the number of reported events may not directly correspond to the number of unique cases of VTE. Although we attempted to mitigate this issue by analyzing the reporting rates, adjusting for temporal trends, and including the total number of adverse event reports as an offset term, this limitation is inherent in pharmacovigilance studies that use spontaneous reporting databases. In addition, duplicate reporting and variability in reporting behavior over time may further influence the observed reporting patterns. Residual autocorrelation was detected in the Pearson residuals of the harmonic seasonal model, indicating that some temporal dependence remained unexplained by the current model. Future studies incorporating more flexible time-series modeling approaches may help address this limitation. Furthermore, although nationwide epidemiological data in Japan indicate a DVT-to-PE ratio of approximately 4:1, DVT and PE were reported in similar numbers in the present study, suggesting possible underreporting of DVT and preferential reporting of more severe events such as PE.25 Another important limitation is that information on drug exposure in the underlying population was not available in spontaneous reporting databases. Such databases are designed primarily to collect reports of suspected adverse drug reactions rather than to monitor all patients receiving a particular medication. Consequently, the total number of exposed patients (ie., the denominator population) is not captured. As a result, the reporting rates observed in this study should not be interpreted as incidence rates. In addition, our analysis did not evaluate the specific medications associated with VTE. Therefore, the differences in thrombotic risk across individual drugs or drug classes could not be assessed in the present study. Furthermore, the present analysis evaluated VTE reports collectively and did not investigate whether seasonal patterns differed across therapeutic categories. Because the mechanisms and timing of reported VTE events may vary substantially among drug classes, future studies stratified by specific medications or therapeutic categories may provide more clinically relevant insights into seasonal variation in VTE reporting. Finally, the analysis was based on aggregated data, and information on individual-level risk factors for VTE, such as comorbidities and immobility, was not available in the present analysis.

Conclusion

VTE adverse event reporting shows a clear seasonal pattern, with higher reporting rates during winter and lower rates during summer. The inverse association between VTE reporting and ambient temperature suggests that environmental factors may be associated with this seasonal reporting variation; however, these findings should be interpreted cautiously, as spontaneous reporting data reflect reporting patterns rather than true incidence or causal relationships. Nevertheless, these findings highlight the importance of considering seasonal and environmental factors when interpreting pharmacovigilance data for VTE adverse event reports. Further studies using complementary data sources and study designs are warranted to confirm these findings and better understand the mechanisms underlying seasonal variations in VTE adverse event reports.

Funding Statement

This research received no specific grants from any funding agency in the public, commercial, or not-for-profit sectors.

Generative AI Statement

The authors declare that Generative AI was used in the creation of this manuscript. The authors acknowledge the use of ChatGPT (OpenAI, GPT-5, https://chat.openai.com/) to support English grammar correction. In addition, the text was subsequently reviewed by a native speaker.

Data Sharing Statement

The datasets generated and/or analyzed in the current study are available from the JADER database (https://www.info.pmda.go.jp/fukusayoudb/CsvDownload.jsp, accessed on January 13, 2026).

Ethics Approval and Informed Consent

Ethics approval was not applicable to this study, and informed consent was not required based on the Ethical Guidelines for Medical and Biological Research Involving Human Subjects issued by the MHLW, since only completely anonymized data were used.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report no competing interests in this work.