Studies retrieved and characteristics
According to the previous search strategy, 3546 citations were obtained from the online database to Feb 1th, 2019. A total of 3512 articles were excluded by viewing the titles and abstracts. The full texts of 34 records were read. Ultimately, 17 full-text studies4-7, 11-23 were obtained and assessed according to the eligibility criteria, including 1 case-control study and 16 cohort studies, comprising more than 71,644 patients. The detailed literature search and screening process are shown in supplement Figure 1. The characteristics included in the study are shown in Table 1, including the first author’s name, year of publication, study design, region, journal, sample size, period of patient recruitment patients, follow-up time and definition of aspirin use.
The quality of 17 studies was assessed using NOS; four studies achieved a score of 6, five a score of 7 and eight studies a score of 8 (Table 2). Thirteen studies stated a clear follow-up time. The longest median follow-up period was 10.8 years. Six studies reported a clear definition of the use of PPIs. Seven studies compared the risk of gastric cancer between PPI users and non-PPI users. Twelve studies evaluated the association between prediagnosis aspirin use and colorectal cancer survival. Twelve studies evaluated the association between postdiagnosis aspirin use and colorectal cancer survival.
- Association between postdiagnosis aspirin use and survival (OS and CSS) in esophageal and gastric cancers
Three studies involving 6,797 patients compared the overall survival of esophageal cancer among aspirin users compared with non-aspirin users. The estimated pooled HRs showed no significant differences between the two groups [HR= 1.009,95%CI(0.847, 1.202)] (figure 1A).
Two studies involving 4,589 patients compared the overall survival of gastric cancer among aspirin users compared with non-aspirin users, and the estimated pooled HRs indicated no significant differences between the groups [HR= 0.870,95%CI(0.470, 1.610)] (figure 1A).
Three studies involving 11,380 patients compared the overall survival of upper digestive cancer among aspirin users compared with non-aspirin users, with no significant differences between the two groups based on estimated pooled HRs [HR= 0.831,95%CI(0.679, 1.016)] (figure 1A).
One study involving 946 patients compared the cancer-specific survival of esophageal cancer among aspirin users with non-aspirin users; based on HRs, the use of aspirin postdiagnosis was associated with longer cancer-specific survival [HR= 0.34,95%CI(014 , 0.69)] (figure 1B). One study involving 750 patients compared the cancer-specific survival of gastric cancer among aspirin users with non-aspirin users, and the HRs revealed no significant differences between the groups [HR= 0.70, 95% CI (0.29, 1.69)] (figure 1B).
- Association between postdiagnosis aspirin use and survival (OS and CSS) in colorectal cancer
Nine studies involving 64,260 patients compared the overall survival of colorectal cancer among aspirin users compared with non-aspirin users. According to the estimated pooled HRs, the use of aspirin postdiagnosis was associated with longer overall survival [HR= 0.82,95%CI(0.72, 0.94)] (figure 2A).
The result of cumulative meta-analysis showed that the significant difference supporting PPI use was first found in the latest study in Gray et al. [HR=0.93, 95% CI(0.89–0.96)], with the CI narrowing and the effect size becoming stable (Figure 2B).
Seven studies involving 49,370 patients compared cancer-specific survival in colorectal cancer among aspirin users and non-aspirin users. The estimated pooled HRs showed that the use of aspirin postdiagnosis was associated with longer overall survival [HR=0.73,95%CI(0.59, 0.91)] (figure 2C).
The result of cumulative meta-analysis indicated that the significant difference supporting PPI use was first found in the latest study by Gray et al. [HR=0.822, 95% CI (0.718-0.941], with the CI narrowing and the effect size becoming stable (Figure 2D).
- Association between prediagnosis aspirin use and survival (OS and CSS) in colorectal cancer
With regard to overall survival in colorectal cancer, five studies involving 6,202 patients compared among aspirin users compared with non-aspirin users. The estimated pooled HRs demonstrated no significant differences between the two groups [HR= 1.01,95%CI(0.96, 1.06)] (figure 3A).
Five studies involving 45,101 patients compared the cancer-specific survival of colorectal cancer among aspirin users compared with non-aspirin users, and according to the estimated pooled HRs, there were no significant differences between the groups [HR= 0.93,95%CI(0.84, 1.03)] (figure 3B)
- Association between both pre- and postdiagnosis aspirin use and survival (OS and CSS) in colorectal cancer
Four studies involving 2,350 patients compared the overall survival of colorectal cancer among aspirin users compared with non-aspirin users. The estimated pooled HRs revealed that the use of aspirin both pre- and postdiagnosis was associated with longer overall survival [HR=0.75,95%CI(0.61, 0.92)] (figure 4A).
Three studies involving 1,849 patients compared cancer-specific survival in colorectal cancer among aspirin users compared with non-aspirin users, and the estimated pooled HRs indicated that the use of aspirin both pre- and postdiagnosis was associated with longer overall survival [HR=0.78,95%CI(0.73, 0.85)] (figure 4B).
- Subgroup analysis according to the PIK3CA gene status
Four studies involving 4,346 patients compared the overall survival of colorectal cancer among aspirin users compared with non-aspirin users among those with PIK3CA gene mutation. Based on the estimated pooled HRs, the use of aspirin postdiagnosis was associated with longer overall survival [HR= 0.70,95%CI(0.50, 0.99)] (figure 5A).
For overall survival in colorectal cancer, three studies involving 8,490 patients compared among aspirin users compared with non-aspirin users among patients with a wild-type PIK3CA gene, and the estimated pooled HRs showed no significant differences between the groups [HR= 0.79,95%CI(0.53, 1.13)] (figure 5A).
Two studies involving 2,451 patients compared the cancer-specific survival in colorectal cancer among aspirin users compared with non-aspirin users among patients with a mutated PIK3CA gene. The estimated pooled HRs showed that the use of aspirin postdiagnosis was associated with longer overall survival [HR= 0.27,95%CI(0.08, 0.91)] (figure 5B).
- Subgroup analysis according to the PTGS2 (COX-2) expression status
Two studies involving 560 patients compared overall survival in colorectal cancer among aspirin users compared with non-aspirin users among patients with strong PTGS2 (COX-2) expression. According to the estimated pooled HRs, the use of aspirin postdiagnosis was associated with longer overall survival [HR= 0.65,95%CI(0.54, 0.83)] (figure 5C).
Regarding the overall survival of colorectal cancer, two studies involving 4,328 patients compared aspirin users with non-aspirin users among patients with weak PTGS2 (COX-2) expression. The estimated pooled HRs showed no significant differences between the two groups [HR= 0.75,95%CI(0.43, 1.30)] (figure 5C).
- Subgroup analysis according tumor stage
Four studies involving 28032 patients compared overall survival in colorectal cancer among aspirin users compared with non-aspirin users among patients. The estimated pooled HRs showed no significant differences between the groups (Supplement Figure 3A).
Five studies involving 32826 patients compared cancer specific survival in colorectal cancer among aspirin users compared with non-aspirin users. The estimated pooled HRs showed no significant differences between the groups in stage I, stage III and stage IV patients. While the use of aspirin was associated with longer cancer specific survival in stage II patients [HR= 0.65,95%CI(0.54, 0.83)] (Supplement Figure 3B).
Sensitivity analysis
Sensitivity analysis was performed to test the stability of the results by excluding each study successively. The results were not affected by sequential exclusion of any particular trial, except for one study (Bains et al, 2016). The detailed sensitivity analysis results are depicted in figure 6.
Publication bias
In a meta-analysis with few studies (less than 10), the power of asymmetrical tests is too low to distinguish chance from real asymmetry. Because of the limited number of included studies, it was difficult to confirm the existence of publication bias in the current meta-analysis.