The chemoprotective effect of anti-platelet agents on cancer incidence in people with non-alcoholic fatty liver disease (NAFLD): a retrospective cohort study.

We performed a retrospective cohort study using the TriNetX (TriNetX LLC, Cambridge, MA, USA) platform. The TriNetX research platform is a global collaborative network providing access to real-time anonymised electronic medical records. TriNetX has data usage and publication agreements in place with all health care organisations (HCOs). The TriNetX Global Collaborative network composes of over 135 million individuals across over 100 health care organisations (HCOs), primarily, secondary and tertiary units in North America and Europe. Data contained with the network includes demographics, diagnosis, procedures, medications and health care utilisation. We conformed to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Additional File 1: Table S1). The data used in this study was collected on 30 March 2024. We identified all adults, aged 18 or over, with NAFLD, defined by the presence of International Classification of Diseases 10th revision (ICD-10) codes K75.8 (other specified inflammatory liver diseases) or K76.0 (fatty change of the liver not elsewhere classified). We excluded individuals with other causes of chronic liver disease (Additional File 1: Table S2). Two groups were created: (1) anti-platelet use and (2) no anti-platelet use. Anti-platelet use was defined as aspirin, ticagrelor, cangrelor, clopidogrel or prasugrel use for at least 1 year. Individuals must have had a diagnosis of NAFLD prior to commencement of antiplatelet use. We adopted an active comparator new user design where analysis was of new starters of antiplatelets. Individuals in the ‘no antiplatelet use’ group must not have had any coding of antiplatelet use in their electronic medical record. The index event for antiplatelet users was defined at 1 year post initiation of medication, and the index event for non-users was defined at 1 year post the first diagnosis of NAFLD (Fig.  1 ). In addition, we performed an analysis of people using aspirin monotherapy as compared to people not prescribed any anti-platelet agents. Fig. 1 Timeline of included individuals for the main cohort analysis Timeline of included individuals for the main cohort analysis Groups were propensity score matched (PSM) for age at index event, sex, ethnicity, presence of type 2 diabetes (T2D) (ICD-10 E11), obesity (defined as BMI ≥ 30 kg/m 2 ), all neoplasms (ICD-10 C00-D49), ischaemic heart diseases (ICD-10 I20-25), cerebrovascular disease (ICD-10 I60-69) and other peripheral vascular disease (ICD-10 I73). Primary outcome was incidence of new HCC [ 13 ] (ICD-10 C22). Secondary outcomes included incidence of other cancers known to be associated with obesity according to the literature (oesophageal [ 14 ], gastric [ 15 ], colorectal [ 16 ], gallbladder [ 17 ], pancreatic [ 18 ], breast [ 19 ], ovarian [ 20 ], uterine [ 21 ], thyroid [ 22 ], multiple myeloma [ 23 ]) (ICD codes in Additional File 1: Table S3). Individuals with a history of an outcome of interest were excluded from prospective analysis of that respective outcome only. Individuals were followed up for 5 years post index event. Participants that died during the study period or who were lost to follow-up (e.g. moving to another HCO not included within the network) were censored at that time point. Sensitivity analysis was conducted for 3- and 5-year aspirin exposure, with outcomes analysed from the point of drug initiation. Due to the nature of the data source, this dataset may face some typical data quality challenges of EMRs such as incomplete or inaccurate data entries, under-reporting of certain conditions, limited granularity and exclusion of data not integrated into the HCO’s EMR. Nevertheless, TriNetX employs data validation processes to ensure the accuracy and reliability of its data. These processes include regular data quality checks to identify and correct discrepancies, validation against external benchmarks to ensure consistency and accuracy and collaboration with data contributors to resolve any identified issues and improve data quality continuously. Statistical analysis is conducted within the TriNetX platform using the R survival package as a backbone. Groups were 1:1 propensity score matched (PSM). We used greedy nearest neighbour matching with a caliper of 0.1 pooled standard deviations. The proportional hazard assumption was tested using the generalised Schoenfeld approach built into the platform. Hazard ratios alongside 95% CI and p values are reported for the prospective analysis and mean and standard deviation for baseline characteristics. Kaplan–Meier curves were calculated for survival probability. If the last data entry (outcomes of interest, date of death, end of data collection, or loss to follow-up) in the patient’s record was in the time window for analysis, the patient was censored on the day after the last fact in their record. We calculated the E -value for each outcome of interest. The E -value is defined as the minimum strength of association, on the risk ratio scale that an unmeasured confounder would need to have with both treatment and outcome to fully explain away a specific treatment-outcome association [ 24 ]. There is no threshold of significance for the E -value and should be interpreted in context with the size of the HR [ 25 ]. Variables with strictly standardised mean difference (SSMD) < 0.1 is well matched between groups. Statistical significance is set at the 5% level.

We identified a total of 1,086,684 individuals. 42,649 had been issued an antiplatelet agent for at least 1 year post diagnosis of NAFLD, and 1,044,035 had no coding of any anti-platelet agent use ever. Post PSM, there were 42,192 in each group. Aspirin monotherapy was the most frequent group consisting of 66.9% of all individuals ( n  = 28,220). Clopidogrel, prasugrel, ticagrelor and cangrelor monotherapy made up 2.6%, 0.5%, 0.4% and < 0.1% respectively. The remaining 29.0% of individuals used a combination of antiplatelet therapy. For individuals prescribed ‘any’ antiplatelet, 93% included a prescription for aspirin. Table 1 summarises the baseline demographics. Table 1 Baseline patient demographics and characteristics post propensity score matching Anti-platelet use ( n  = 42,192) No antiplatelet use ( n  = 42,192) Strictly standardised mean difference Demographics  Age at index event (years) 62.8 ± 12.6 63.2 ± 12.7 0.029  Sex (female) [%] 58 57 0.007  Race (White/Black or African American/Asian) [%] 71/9/4 70/9/3 0.011/0.006/0.012 Anthropometrics  Body mass index (kg/m 2 ) 33.4 ± 7.6 33.9 ± 7.5 0.062 Comorbidities [%]  Type 2 diabetes 33.8 33.4 0.009  Ischaemic heart disease 20.1 19.8 0.008  History of any neoplasm 14.4 14.8 0.012  Cerebrovascular disease 6.7 6.4 0.010  Other peripheral vascular disease 3.3 3.0 0.019 Baseline patient demographics and characteristics post propensity score matching Overall, there were 796 cancer events in people prescribed anti-platelets and 1033 events for people not prescribed anti-platelets respectively. Antiplatelet use was associated with statistically significant reduction in all obesity-related cancers (HR 0.71, 95% CI 0.65–0.78, p  < 0.001, E -value 1.85) in people with NAFLD (Table  2 ). Antiplatelet use was associated with a significantly reduced risk of HCC (HR 0.52, 95% CI 0.40–0.68, p  < 0.001, E -value 2.52) compared to non-users (Fig.  2 ). Benefit was observed after a year of prescription and increased over time (Fig.  2 ). Table 2 Summary of outcomes stratified for all antiplatelet and aspirin only users All antiplatelets ( n  = 84,304) Aspirin only ( n  = 56,440) Sample size Outcome ( n ) HR (95% confidence interval) p value E value Sample size Outcome ( n ) HR (95% confidence interval) p value E value All obesity-related cancers a  Antiplatelet(s) 37,757 796 0.71 (0.65–0.78)  < 0.001 1.85 15,417 280 0.64 (0.55–0.74)  < 0.001 2.06  No antiplatelets 38,536 1033 15,421 407 Hepatocellular carcinoma  Antiplatelet(s) 41,799 89 0.52 (0.40–0.68)  < 0.001 2.52 27,953 48 0.46 (0.32–0.64)  < 0.001 2.80  No antiplatelets 41,941 156 28,061 99 Colorectal carcinoma  Antiplatelet(s) 41,292 162 0.68 (0.56–0.84)  < 0.001 1.94 27,564 93 0.63 (0.49–0.82) 0.001 2.10  No antiplatelets 41,406 217 27,710 138 Pancreatic carcinoma  Antiplatelet(s) 41,963 101 0.61 (0.47–0.78)  < 0.001 2.16 28,054 63 0.61 (0.45–0.84) 0.002 2.16  No antiplatelets 41,883 152 27,997 96 Oesophageal carcinoma  Antiplatelet(s) 41,119 26 0.70 (0.42–1.17) 0.174 1.00 28,169 16 0.58 (0.31–1.08) 0.082 2.27  No antiplatelets 41,093 34 28,150 26 Gastric carcinoma  Antiplatelet(s) 42,079 43 0.76 (0.50–1.13) 0.171 1.00 28,144 29 0.73 (0.45–1.19) 0.207 1.00  No antiplatelets 42,075 52 28,137 37 Gallbladder carcinoma  Antiplatelet(s) 42,162 10 b 0.66 (0.29–1.48) 0.305 1.00 28,197 10 b 0.62 (0.22–1.75) 0.365 1.00  No antiplatelets 42,156 14 28,195 10 b Ovarian carcinoma  Antiplatelet(s) 41,979 58 0.87 (0.61–1.25) 0.443 1.00 28,061 41 0.85 (0.56–1.30) 0.457 1.00  No antiplatelets 42,004 61 28,089 45 Uterine carcinoma  Antiplatelet(s) 41,718 93 0.75 (0.57–0.98) 0.034 1.74 27,881 71 0.81 (0.59–1.11) 0.191 1.00  No antiplatelets 41,751 114 27,877 82 Breast carcinoma  Antiplatelet(s) 17,535 255 0.78 (0.66–0.92) 0.003 1.66 26,885 173 0.80 (0.65–0.97) 0.024 1.61  No antiplatelets 17,502 302 27,236 206 Multiple myeloma  Antiplatelet(s) 41,910 63 0.98 (0.68–1.39) 0.889 1.00 28,007 36 0.79 (0.50–1.22) 0.283 1.00  No antiplatelets 42,051 59 28,146 43 Thyroid carcinoma  Antiplatelet(s) 41,802 71 0.82 (0.60–1.13) 0.225 1.00 27,933 48 0.92 (0.62–1.36) 0.665 1.00  No antiplatelets 41,949 79 28,044 49 a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Fig. 2 Kaplan–Meier estimates of time to HCC diagnosis Summary of outcomes stratified for all antiplatelet and aspirin only users a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Kaplan–Meier estimates of time to HCC diagnosis Antiplatelet agent use was associated with statistically significant reduction in incident breast carcinoma (HR 0.78, 95% CI 0.66–0.92, p  = 0.003, E -value 1.66), pancreatic carcinoma (HR 0.61, 95% CI 0.47–0.78, p  < 0.001, E -value 2.16) and colorectal carcinoma (HR 0.68, 95% CI 0.56–0.84, p  < 0.001, E -value 1.94). For women, there was a significant reduction in risk of uterine carcinoma (HR 0.75, 95% CI 0.57–0.98, p  = 0.034, E -value 1.74) (Table  2 ). There was no significant difference between users and non-users of anti-platelet agents in incidence of gallbladder carcinoma (HR 0.66, 95% CI 0.29–1.48, p  = 0.305), gastric carcinoma (HR 0.76, 95% CI 0.50–1.13, p  = 0.171), oesophageal carcinoma (HR 0.70, 95% CI 0.42–1.17, p  = 0.174), ovarian carcinoma (HR 0.87, 95% CI 0.61–1.25, p  = 0.443), multiple myeloma (HR 0.98, 95% CI 0.68–1.39, p  = 0.889) and thyroid carcinoma (HR 0.82, 95% CI 0.60–1.13, p  = 0.225) (Table  2 ). A reduced incidence of all obesity-related cancer according to anti-platelet use was observed for both men (HR 0.69, 95% CI 0.58–0.82, p  < 0.001, E -value 1.91) and women (HR 0.78, 95% CI 0.70–0.88, p  < 0.001, E -value 1.66) (Table  3 , Fig.  3 ). For men, antiplatelet use was associated with reduced risk of HCC (HR 0.49, 95% CI 0.34–0.72, p  < 0.001, E -value 2.66) and colorectal carcinoma (HR 0.70, 95% CI 0.52–0.96, p  = 0.026, E -value 1.88). For women, a benefit was observed for colorectal carcinoma (HR 0.71, 95% CI 0.54–0.94, p  = 0.018, E -value 1.85), pancreatic carcinoma (HR 0.45, 95% CI 0.31–0663, p  < 0.001, E -value 2.86) and uterine carcinoma as described previously (Table  3 , Fig.  3 ). Table 3 Summary of outcomes stratified by sex for all antiplatelet users Male ( n  = 35,630) Female ( n  = 44,240) Sample size Outcome ( n ) HR (95% confidence interval) p value E value Sample size Outcome ( n ) HR (95% confidence interval) p value E value All obesity-related cancers a  Antiplatelets 16,749 224 0.69 (0.58–0.82)  < 0.001 1.91 19,003 525 0.78 (0.70–0.88)  < 0.001 1.66  No antiplatelets 16,937 293 19,639 638 Hepatocellular carcinoma  Antiplatelets 17,589 43 0.49 (0.34–0.72)  < 0.001 2.66 21,967 44 0.68 (0.46–1.01) 0.053 1.00  No antiplatelets 17,667 78 22,035 60 Colorectal carcinoma  Antiplatelets 17,391 71 0.70 (0.52–0.96) 0.026 1.88 21,691 85 0.71 (0.54–0.94) 0.018 1.85  No antiplatelets 17,420 89 21,779 111 Pancreatic carcinoma  Antiplatelets 17,707 54 0.72 (0.50–1.03) 0.070 1.00 22,012 39 0.45 (0.31–0.66)  < 0.001 2.86  No antiplatelets 17,693 67 21,963 81 Oesophageal carcinoma  Antiplatelets 17,768 14 0.54 (0.28–1.05) 0.066 1.00 22,096 10 b 0.84 (0.36–1.98) 0.690 1.00  No antiplatelets 17,760 23 22,106 11 Gastric carcinoma  Antiplatelets 17,760 21 0.60 (0.35–1.05) 0.068 1.00 22,070 20 0.97 (0.52–1.83) 0.935 1.00  No antiplatelets 17,759 31 22,081 19 Gallbladder carcinoma  Antiplatelets 17,787 10 b 1.32 (0.22–7.91) 0.759 1.00 22,109 10 b 0.69 (0.24–2.00) 0.494 1.00  No antiplatelets 17,802 10 b 22,098 10 b Ovarian carcinoma  Antiplatelets N/A 21,921 58 0.87 (0.61–1.25) 0.443 1.00  No antiplatelets 21,932 61 Uterine carcinoma  Antiplatelets N/A 21,685 93 0.75 (0.57–0.98) 0.034 1.74  No antiplatelets 21,715 114 Breast carcinoma  Antiplatelets 17,783 10 b 1.04 (0.32–3.41) 0.947 1.00 20,448 235 0.79 (0.66–0.94) 0.006 1.63  No antiplatelets 17,791 10 b 20,878 281 Multiple myeloma  Antiplatelets 17,689 25 0.85 (0.49–1.47) 0.553 1.00 10,755 36 0.98 (0.61–1.57) 0.934 1.00  No antiplatelets 17,754 26 10,799 34 Thyroid carcinoma  Antiplatelets 17,714 29 1.07 (0.62–1.83) 0.819 1.00 21,852 36 0.82 (0.52–1.28) 0.369 1.00  No antiplatelets 17,756 24 21,975 41 a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Fig. 3 Forest plot of all clinical outcomes at 5 years in all antiplatelet users with NAFLD, sub-stratified by sex Summary of outcomes stratified by sex for all antiplatelet users a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Forest plot of all clinical outcomes at 5 years in all antiplatelet users with NAFLD, sub-stratified by sex Overall, 28,220 individuals were issued aspirin only post PSM. Aspirin monotherapy was associated with statistically significant reduction in incident obesity-related cancers combined (HR 0.64, 95% CI 0.55–0.74, p  < 0.001, E -value 2.06). For individual cancers, aspirin use was found to be protective for HCC (HR 0.46, 95% CI 0.32–0.64, p  < 0.001, E -value 2.80), colorectal carcinoma (HR 0.63, 95% CI 0.49–0.82, p  = 0.001, E -value 2.10), pancreatic carcinoma (HR 0.61, 95% CI 0.45–0.84, p  = 0.002, E -value 2.16) and breast carcinoma (HR 0.80, 95% CI 0.65–0.97, p  = 0.024, E -value 1.61). There was no significant difference between aspirin users and non-users of anti-platelets in incidence of oesophageal carcinoma (HR 0.58, 95% CI 0.31–1.08, p  = 0.082), gastric carcinoma (HR 0.73, 95% CI 0.45–1.19, p  = 0.207), gallbladder carcinoma (HR 0.62, 95% CI 0.22–1.75, p  = 0.365), multiple myeloma (HR 0.79, 95% CI 0.50–1.22, p  = 0.283) or thyroid carcinoma (HR 0.92, 95% CI 0.62–1.36, p  = 0.665) (Table  2 ). For women, aspirin prescription was associated with reduced incidence of all obesity-related cancers (HR 0.77, 95% CI 0.67–0.88, p  < 0.001, E -value 1.69). Specifically, a significant risk reduction was observed for colorectal carcinoma (HR 0.65, 95% CI 0.45–0.94, p  = 0.019, E -value 2.03), pancreatic cancer (HR 0.41, 95% CI 0.25–0.68, p  < 0.001, E -value 3.09) and breast carcinoma (HR 0.76, 95% CI 0.62–0.93, p  = 0.009, E -value 1.71). No statistically significant association was observed for other obesity-related cancers (Table  4 , Fig.  4 ). For men, aspirin use was associated with reduced incidence of all obesity-related cancers (HR 0.71, 95% CI 0.56–0.90, p  = 0.004, E -value 1.85) and HCC (HR 0.41, 95% CI 0.24–0.68, p  < 0.001, E -value 3.09), (Table  4 , Fig.  4 ). Table 4 Summary of outcomes of aspirin users stratified by sex (people prescribed aspirin monotherapy vs non-users of any antiplatelets) Male ( n  = 21,964) Female ( n  = 31,584) Sample size Outcome ( n ) HR (95% confidence interval) p value E value Sample size Outcome ( n ) HR (95% confidence interval) p value E value All obesity-related cancers a  Aspirin 10,271 125 0.71 (0.56–0.90) 0.004 1.85 13,469 346 0.77 (0.67–0.88)  < 0.001 1.69  No antiplatelets 10,441 162 13,999 437 Hepatocellular carcinoma  Aspirin 10,846 21 0.41 (0.24–0.68)  < 0.001 3.09 15,671 26 0.63 (0.38–1.03) 0.064 1.00  No antiplatelets 10,888 47 15,739 39 Colorectal carcinoma  Aspirin 10,690 39 0.71 (0.47–1.08) 0.103 1.00 15,463 49 0.65 (0.45–0.94) 0.019 2.03  No antiplatelets 10,738 50 15,546 71 Pancreatic carcinoma  Aspirin 10,909 36 0.82 (0.53–1.29) 0.397 1.00 15,709 22 0.41 (0.25–0.68)  < 0.001 3.09  No antiplatelets 10,902 40 15,678 50 Oesophageal carcinoma  Aspirin 10,949 10 b 1.01 (0.41–2.50) 0.979 1.00 15,776 10 b 0.67 (0.21–2.11) 0.488 1.00  No antiplatelets 10,947 10 b 15,780 10 b Gastric carcinoma  Aspirin 10,604 13 0.70 (0.34–1.43) 0.321 1.00 15,754 15 0.83 (0.41–1.66) 0.595 1.00  No antiplatelets 10,593 17 15,769 17 Gallbladder carcinoma  Aspirin 10,969 10 b 0.46 (0.04–5.06) 0.514 1.00 15,783 10 b 0.78 (0.24–2.56) 0.681 1.00  No antiplatelets 10,973 10 b 15,775 10 b Ovarian carcinoma  Aspirin N/A 15,643 38 0.73 (0.48–1.11) 0.140 1.00  No antiplatelets 15,647 49 Uterine carcinoma  Aspirin N/A 15,473 66 0.92 (0.66–1.30) 0.642 1.00  No antiplatelets 15,475 67 Breast carcinoma  Aspirin 10,959 10 b 2.70 (0.28–29.92) 0.371 1.00 14,558 160 0.76 (0.62–0.93) 0.009 1.71  No antiplatelets 10,970 10 b 14,915 202 Multiple myeloma  Aspirin 10,892 15 1.23 (0.57–2.68) 0.603 1.00 15,676 19 0.90 (0.48–1.68) 0.731 1.00  No antiplatelets 10,943 11 15,759 20 Thyroid carcinoma  Aspirin 10,920 13 1.08 (0.48–2.40) 0.859 1.00 15,584 32 1.00 (0.61–1.65) 0.988 1.00  No antiplatelets 10,953 11 15,682 30 a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Fig. 4 Forest plot of all clinical outcomes at 5 years in aspirin monotherapy users with NAFLD, sub-stratified by sex Summary of outcomes of aspirin users stratified by sex (people prescribed aspirin monotherapy vs non-users of any antiplatelets) a Individuals were censored at the first coding of a constituent obesity-related malignancy composite outcome. The total number of individuals experiencing the composite outcome differ than that of the sum of the individual events because, to better ascertain the primary preventative effect of aspirin on all obesity-related malignancies, individuals with a history of any of the constituent events were excluded from analysis of the composite outcome b TriNetX implements several safeguards to minimise the risk of patient reidentification. To avoid the risk that a series of individual queries could identify small subsets of cohorts, when a query returns a patient count on an outcome where the patient count is ≤ 10 but greater than 0, the count is obfuscated to 10. The reported HR is calculated without this obfuscation present Forest plot of all clinical outcomes at 5 years in aspirin monotherapy users with NAFLD, sub-stratified by sex For individuals aged over 60, aspirin use was associated with reduced incidence of all obesity-related cancers (HR 0.72, 95% CI 0.63–0.82, p  < 0.001, E -value 1.82), HCC (HR 0.41, 95% CI 0.28–0.59, p  < 0.001, E -value 3.09), colorectal carcinoma (HR 0.67, 95% CI 0.51–0.90, p  = 0.006, E -value 1.97), pancreatic carcinoma (HR 0.70, 95% CI 0.49–0.99, p  = 0.040, E -value 1.88), gastric carcinoma (HR 0.59, 95% CI 0.36–0.97, p  = 0.035, E -value 2.24) and breast carcinoma (HR 0.74, 95% CI 0.59–0.92, p  = 0.005, E -value 1.77). Cancer incidence was significantly less in individuals aged 60 or less and aspirin was not associated with any reduced incidence of cancers (Additional File 1: Table S4, Additional File 1: Fig. S1). Aspirin exposure for at least 3 and 5 years was associated with a statistically significant reduced incidence of all obesity-related cancers, HCC and pancreatic carcinoma (Additional File 1: Table S5, Additional File 1: Fig. S2). Aspirin exposure for 3 years, with follow-up beginning from the point of drug initiation, was associated with a statistically significant reduced incidence of obesity-related cancers (HR 0.79, 95% CI 0.69–0.89, p  < 0.001, E -value 1.63) and HCC (HR 0.55, 95% CI 0.39–0.78, p  < 0.001, E -value 2.39). Additionally, aspirin exposure for 5 years, from the point of initiation, was associated with a greater reduced incidence of obesity-related cancers (HR 0.74, 95% CI 0.61–0.89, p  = 0.001, E -value 1.77) and HCC (HR 0.44, 95% CI 0.26–0.76, p  = 0.003, E -value 2.91).

Obesity is a chronic complex disease associated with a wide range of health complications including medical (cardiovascular disease, type 2 diabetes, non-alcoholic fatty liver disease (NAFLD)), musculoskeletal, mental health complications and risk of multiple malignancies (oesophageal, gastric, colorectal, liver, pancreatic, gall bladder, breast, uterine, ovarian, thyroid, meningioma, multiple myeloma) [ 1 ]. Prevalence of obesity has doubled from 1990 to 2022, according to World Health Organisation data, with 43% of adults living with been overweight or obese (60% in Europe and 67% in the Americas) [ 2 ]. NAFLD has emerged as the leading cause of chronic liver disease, affecting up to one third of the global population [ 3 ]. NAFLD occurs where there is excessive hepatic fat accumulation with secondary inflammation and potentially fibrosis. NAFLD is now the main driver for increased chronic liver disease incidence [ 4 ] and is associated with cirrhosis and hepatocellular carcinoma (HCC). As a multisystem disorder, it is also independently associated with cardiovascular disease [ 5 ], chronic kidney disease [ 6 ] and extra-hepatic cancer [ 7 ]. A recent meta-analysis including 10 cohort studies (182,202 people, median follow-up 5.8 years) identified that NAFLD was significantly associated with a 1.5- to 2.0-fold increased risk of incident gastrointestinal cancers (oesophagus, stomach, pancreas, colorectal cancers) and a 1.2- to 1.5-fold increased risk of lung, breast, gynaecological or urinary system [ 7 ]. These risks were independent of age, sex, smoking, obesity and diabetes status, although many patients will have multiple common metabolic-related oncogenic risk factors. Obesity also may adversely affect/limit oncological treatment options, increase the risk of cancer-related mortality and increase rates of disease recurrence [ 8 ]. Prevention and early detection/timely treatment of cancer are therefore critical for people living with obesity. There has been increasing interest in the use of aspirin, an inhibitor of cyclooxygenase (COX)−2, for primary prevention of cancer and to improve cancer survival post diagnosis. Evidence is strongest for colorectal and other gastrointestinal tract [ 9 ]. Aspirin has also been shown to be protective against HCC in the general population and people with chronic liver disease in recent meta-analyses [ 10 , 11 ], although no benefit was seen for people with cirrhosis who are at the highest risk of HCC [ 11 ]. For individuals with NAFLD, aspirin has recently been shown to halve the risk of HCC in a large retrospective study using Taiwan’s National Health Insurance Research Database (145,212 NAFLD patients, 33,484 received daily aspirin for 90 days or more and 55,543 patients did not receive any antiplatelet therapy, adjusted hazard ratio, HR 0.48, 95% CI 0.37–0.63) [ 12 ]. To our knowledge, the benefits of either aspirin or other anti-platelet agents has not been explored in people with NAFLD for prevalence of other cancers associated with obesity. This is highly clinically relevant given the high burden of metabolic related (oncogenic) risk factors in this population. Using a large international cohort of patients, we therefore explored the chemoprotective role of aspirin, and other anti-platelet agents, for the prevention of HCC and other cancers associated with obesity in the literature.

Using a large global federation health research network, we demonstrate that use of anti-platelet agents (for which over 90% included a prescription for aspirin) is associated with reduced incidence of HCC and other cancers known to be linked to obesity, namely colorectal, breast, pancreatic and uterine cancer in people with NAFLD following propensity score matching for confounders. For aspirin monotherapy, there was a reduction in risk for HCC, colorectal, breast cancer and pancreatic cancer. This protection was lost for people under 60 years where cancer incidence was reduced. This is the first study in a Western population to examine the association between aspirin and incident HCC in people with NAFLD and the first to analyse the wider chemoprotective role of aspirin for other obesity-related cancers in this population. Observational data supports a chemoprotective role of aspirin for HCC. In a recent systematic review and meta-analysis, aspirin use was associated with a lower incidence of HCC in a dose-dependent and duration-dependent manner and was also found to be associated with reduced recurrence and mortality [ 26 ]. In a further meta-analysis (12 cohort studies, 4 case–control studies; 822,680 aspirin users, 20,626 HCC cases), Abdelmalak et al. demonstrated that aspirin use reduces incident HCC by 30%, although this protection was not demonstrated for people with cirrhosis, a leading risk factor for HCC [ 11 ]. This association has not been studied in the setting of a randomised control trial (RCT) however. Few studies have examined a chemoprotective role for aspirin in people with NAFLD. This is of particular interest as the prevalence of NAFLD-associated HCC [ 27 ]. Lee et al. conducted a large retrospective cohort study to assess the relationship between aspirin and reduction in risk of HCC in people with NAFLD using Taiwan’s National Health Insurance Research database, involving 145,212 NAFLD patients, 33,484 patients continuously receiving a daily dose of aspirin for at least 90 days and 55,543 patients who did not receive any antiplatelet therapy [ 12 ]. Aspirin therapy was associated with a reduced HCC risk (adjusted HR 0.48 [95% CI 0.37–0.63]), and aspirin use for greater than 3 years was associated with the lowest risk of HCC. In this study, we observe a similar reduction in risk for both aspirin monotherapy and any platelet use for people with NAFLD. Following stratification for sex, protection against incident HCC in this study was only observed for men, which may be related to the higher incidence of primary liver cancer in this group. A sensitivity analysis, with follow-up starting from the point of antiplatelet initiation, demonstrated a reduction in incidence of obesity-related cancers and HCC, with a greater effect seen with 5-year aspirin exposure compared to 3 years. We also demonstrate a reduction in incident colorectal cancer in addition to breast, uterine and pancreatic cancer for women and colorectal cancer for men for people with NAFLD prescribed any anti-platelet therapy. While observational data supports a protective role of aspirin use for colorectal cancer [ 28 ], meta-analysis of RCT data has shown conflicting results. Ma et al. concluded that while aspirin use overall did not reduce colorectal cancer incidence, a pooled analysis of studies which used low dose aspirin showed moderate benefit (relative risk 0.84) [ 29 ]. Aspirin was also found to reduce recurrence and cancer-related mortality [ 29 ]. Conversely, Shah et al. concluded that only high-dose aspirin is protective [ 30 ]. Ghaddaf et al. reported that aspirin use only reduces the risk of advanced lesions for up to 5 years [ 31 ]. Indeed, an updated evidence report and systemic review for the US Preventative Services Task Force (USPSTF) concluded that while low-dose aspirin was associated with small absolute risk reductions in major cardiovascular disease, ‘colorectal cancer results were less robust and highly variable’ [ 32 ]. With respect to the role of aspirin and incidence of other cancers, most data available for analysis is observational. Meta-analysis data do suggest a chemo-preventative benefit for aspirin for incident pancreatic cancer, in particular for people taking high-dose aspirin with a longer duration of use [ 33 , 34 ]. Similarly, meta-analyses have reported a modest reduced incidence of gastric cancer (33 studies, risk ratio 0.89) [ 35 ] and breast cancer (42 studies, relative risk 0.92) [ 36 ], although there was significant heterogeneity between studies. Pooled analyses of 12 observational studies have identified a 13% reduction in incident ovarian cancer in all subgroups other than women with endometriosis [ 37 ]. This benefit is not borne out in RCT data however. In the Women’s Health Study, a randomised 2 × 2 factorial trial of aspirin 100 mg daily and aspirin placebo (39,876 US women) for 10 years did not reveal any difference in incidental cancer at any site other than non-melanoma skin cancer [ 38 ]. While other studies have identified reduced incidence of oesophageal cancer (metanalysis, 9 studies) [ 39 ], this data is observational only. In common with this study, no benefit has been demonstrated for aspirin against incidence myeloma [ 40 ] and minimal data exists for gallbladder cancer. In this cohort, we identify that any anti-platelet use is associated with reduced incidence of HCC, colorectal cancer, pancreatic cancer, uterine cancer and breast cancer, and aspirin monotherapy was only found to be protective against all the above but not uterine cancer. It is unclear whether this is related to a lower number of people at risk and lower number of events in the aspirin monotherapy group or a compound effect of dual or consecutive anti-platelet use. Aspirin inhibits cyclooxygenase-2, which promotes inflammation and cell proliferation and inhibits nuclear factor kappa light chain induction of apoptosis. In terms of HCC, aspirin may influence carcinogenesis via reduction in hepatic fat content, recently demonstrated in a preliminary phase 2 trial [ 41 ]. Antiplatelet therapy (aspirin/clopidogrel) has also been shown to reduce intrahepatic platelet accumulation and platelet–immune cell interaction, limiting hepatic immune cell trafficking leading to attenuated intrahepatic cytokine and chemokine release, macrovesicular steatosis and hepatic inflammation [ 42 ]. The clinical implications of our findings are hugely significant considering NAFLD represents a major public health challenge [ 3 ]. Breast and colorectal carcinoma represent two of the leading causes of cancer in the UK. Although HCC is less common, survival rates are poor at it is the only cancer for which incidence and mortality rates are increasing [ 43 ]. Our study supports previous findings from observational research that anti-platelet therapy may be beneficial for primary prevention of common obesity-related cancers. However, these findings have not yet been convincingly reproduced in RCTs, the level of evidence required for clinical recommendations. In 2016, the US Preventive Services Task force (USPSTF) endorsed for the first time low-dose aspirin for prevention of colorectal cancer, in addition to cardiovascular disease, for individuals aged 50–59 years with a 10% 10-year cardiovascular risk [ 44 ]. However, in 2022, the USPSTF revised its recommendations regarding aspirin for primary prevention of CVD and withdrew its recommendation regarding colorectal cancer, citing ‘inadequate’ evidence to support aspirin’s reduction of colorectal cancer risk [ 32 ]. Therefore, future proof-of-concept trials should be performed to confirm, or refute, our findings, to help inform clinical guidelines. Such clinical guidelines would need to detail a practical and cost-effective approach to identify those at high risk of obesity-related complications (without imaging or biopsy studies) best placed for anti-platelet-based cancer chemoprotection. Non-invasive markers of fibrosis, e.g. fibrosis-4 score could be used to triage patients into more specific investigations. Our study has several strengths, including being the first and largest real-world study performed in a Western population to assess the impact of anti-platelet therapy on protection against obesity-related cancers in people with NAFLD. The topic is highly clinically relevant given the rising rates of obesity in adults and children [ 2 ], and subsequently NAFLD, and the burden of hepatic and extrahepatic cancers in this group. Of note, the composite outcome of ‘obesity-related’ cancer used in this paper was chosen given the clinical relevance of this topic and strong evidence in the literature linking certain cancers with obesity. It does not translate that all the incident cancer cases reported here were directly related to obesity and instead will have occurred as a result of a complex interplay of metabolic risk, lifestyle factors and genetics. We must acknowledge some limitations. Firstly, these are real-world data, and comparisons are not randomised nor controlled. Therefore, we cannot comment on causation. Second, resulting from data being extracted from electronic health records of an administrative database, there is potential for a lack of data completeness. For example, data may not be recorded by the HCO, such as the dose, or duration, of treatment, or recorded in free text that we are unable to extrapolate. We attempted to mitigate against challenges faced in determining treatment duration by ensuring that repeat coding for anti-platelet therapy was present. NAFLD may have resolved in some participants over the study time course. We were unable to identify which patients had experienced NAFLD resolution as such assessment would require serial biopsies or imaging data that was not available to us. Related, we could not analyse the impact of individual anti-platelets, aside from aspirin, on obesity-related cancer outcomes, as the sample size was too small. Moreover, as with any retrospective database study, despite thorough covariate adjustment through PSM at baseline, it is possible that minimal residual bias confounding remains. We attempted to reduce risk of unidentified residual confounding through calculation of E -values as a quantitative bias analysis to assist readers in the interpretation of the strength of our results [ 24 ]. The 1-year time lag may have introduced an element of immortality bias, but this was consistent between groups. In addition, a sensitivity analysis from the point of drug initiation was performed which further supports our findings. Individuals adhering to long-term medication may have higher socioeconomic status and healthier lifestyles which are better controlled for and evaluated through RCTs. We could not evaluate outcomes over a longer period of follow-up (e.g. 10 years) due to the loss of signal. Finally, although ICD-10 revision coding is a validated method for identifying disease outcomes, variability in diagnostic and coding practices might influence its accuracy. We used diagnostic codes that pertain to the diagnosis of ‘NAFLD’, despite the recent update in the nomenclature to metabolic dysfunction associated steatotic liver disease (MASLD) as there are significant gaps in metabolic data measured which is necessary to make a diagnosis of MASLD.

In summary, our study highlights the novel potential of anti-platelet therapy, notably aspirin, in reducing the incidence of several hepatic and extra-hepatic obesity-related cancers, in individuals with NAFLD, a high-risk population. Randomised, controlled studies should explore their potential primary cancer prevention role.

Additional File 1: Table S1. STROBE checklist. Table S2. ICD-10 codes used to exclude people with other aetiologies of chronic liver disease when creating the study cohorts. Table S3. ICD-10 codes of other obesity-related carcinomas. Table S4. Summary of outcomes of aspirin users stratified by age (people prescribed aspirin monotherapy vs non-users of any antiplatelets). Table S5. Summary of outcomes of aspirin users stratified by length of exposure (people prescribed aspirin monotherapy vs non-users of any antiplatelets). Fig. S1. Forest plot of all clinical outcomes at 5 years in aspirin users with MASLD, sub-stratified by age. Fig. S2. Forest plot of all clinical outcomes at 5 years in aspirin users with MASLD, sub-stratified by length of aspirin exposure. Additional File 1: Table S1. STROBE checklist. Table S2. ICD-10 codes used to exclude people with other aetiologies of chronic liver disease when creating the study cohorts. Table S3. ICD-10 codes of other obesity-related carcinomas. Table S4. Summary of outcomes of aspirin users stratified by age (people prescribed aspirin monotherapy vs non-users of any antiplatelets). Table S5. Summary of outcomes of aspirin users stratified by length of exposure (people prescribed aspirin monotherapy vs non-users of any antiplatelets). Fig. S1. Forest plot of all clinical outcomes at 5 years in aspirin users with MASLD, sub-stratified by age. Fig. S2. Forest plot of all clinical outcomes at 5 years in aspirin users with MASLD, sub-stratified by length of aspirin exposure.