Nonsteroidal Anti-inflammatory Drugs But Not Aspirin Are Associated With a Lower Risk of Post-colonoscopy Colorectal Cancer

Ka Shing Cheung; Lijia Chen; Esther W. Chan; Wai Kay Seto; Ian C.K. Wong; Wai K. Leung


Aliment Pharmacol Ther. 2020;51(9):899-908. 

In This Article


Although NSAIDs have been shown to be associated with a lower risk of CRC,[15] studies on the role of NSAIDs in PCCRC (which accounts for up to 9% of all diagnosed CRCs) are lacking.[8] To our knowledge, this is the first study involving more than 180 000 subjects to demonstrate the potential chemopreventive effects of NSAIDs on PCCRC. We showed that NSAIDs were associated with a 47% lower risk of PCCRC-3y and the benefits were observed for both proximal and distal cancers.

Although the magnitude of protection of NSAIDs against PCCRC-3y in our study was similar to that reported in studies on NSAIDs against all CRCs,[15] previous studies failed to stratify cancers into detected CRC and PCCRC. Results from this study would therefore shed new light onto the potential chemopreventive effects of NSAIDs on CRC development according to the timing of NSAID uses and colonoscopy. Intuitively, NSAIDs appear to inhibit the growth of pre-existing neoplastic lesions that are either missed or residual lesions left after polypectomy,[18,19,47] as well as reducing the number and size of colonic adenomas.[48,49] The effect of NSAIDs on PCCRC>3y, however, was nonsignificant on the main analysis. Further studies with even larger sample size may be needed to avoid possible underpower of the subgroup analysis in this study.

When individual NSAIDs were analysed, diclofenac and naproxen were both associated with reduced PCCRC-3y risk. In contrast, a recent study showed that nearly all non-aspirin NSAIDs were associated with reduced CRC risk.[16] However, our results on individual NSAID analysis should also be interpreted with caution due to the small number of events in patients using NSAIDs other than diclofenac and naproxen.

Interestingly, aspirin was not found to be associated with a lower PCCRC-3y risk in this study. First, while COX-2 inhibition is believed to play an important role in the chemopreventive effect of NSAIDs,[22,23] aspirin is more selective for COX-1 inhibition. Our finding is in line with a recent network meta-analysis which showed that NSAIDs are superior to low-dose and high-dose aspirins in preventing advanced metachronous neoplasia (advanced adenoma and CRC) in patients with previous colorectal neoplasia (odds ratio of 0.37 and 0.71 for NSAIDs and aspirin respectively).[24] Second, chemopreventive effect of aspirin depends on the duration of use and latency period. Both post hoc analysis of clinical trials and prospective studies have shown that chemopreventive effect of aspirin is evident only after more than 5 years of use with a latency period of at least 10 years.[25,26] On the other hand, multiple observational studies have demonstrated a much shorter duration of NSAID use might be enough for the chemopreventive effect to exert.[15] Collectively, these data might hint that NSAIDs appear to be more potent than aspirin on CRC prevention, particularly over a relatively short period of time as in PCCRC prevention.

In this study, the beneficial effect of NSAIDs was found to be similar in both proximal and distal cancers. In contrast, the chemopreventive effect of statins on PCCRC-3y is limited to proximal cancer as demonstrated in our recent study.[39] The current subgroup analysis also shows that the beneficial effect of NSAIDs was observed in both sex, but may be higher in females. However, it was only significant among those aged 60 years or above, which may be explained by lower burden of both adenomatous[50] and serrated polyps[51] in younger patients, and hence a lower risk of missed colonic polyps or incomplete resection of lesions. The beneficial effect of NSAIDs was also limited to nondiabetic patients and those without the history of colonic polyps. Hyperinsulinaemia in diabetic patients, which promotes cancer growth, may override the beneficial effect of NSAIDs. However, cautions should be undertaken in interpreting these results due to possible underpower from subgroup analysis, in particular those with the history of colonic polyps in which borderline significance was noted. As NSAIDs are associated with GIB and cardiovascular diseases,[15] subgroup analysis provide insights into which subgroup of patients may benefit more from NSAID use. Further studies are warranted to determine whether there are subgroups in which a favourable risk-benefit profile exists. Concomitant use of proton pump inhibitors (PPIs) to reduce the risk of upper GIB may also be considered to increase the benefit-risk ratio in at-risk groups, as a recent study showed that the chemopreventive effect of NSAIDs on CRC was not modified by PPIs.

There are several strengths of this study. First, the use of territory-wide healthcare database, which captured all diagnoses, drug prescription and dispensing history, would limit some of the biases common to traditional observational studies including selection and recall biases.[41] Importantly, 'reverse causality' was minimised by defining NSAID exposure as baseline drug use prior to index colonoscopy. This is well illustrated by another study showing a possible 'reverse causality' mainly occurred when NSAID use within 1–6 months before CRC diagnosis was considered.[47] Immortal time bias was negligible in this study as the primary analysis focussed on NSAID use before index colonoscopy. Second, no previous studies specifically investigated the effect of NSAIDs in patients who had prior colonoscopy and negative for CRC.[15]

Certain limitations of this study exist. First, data on some of the risk factors for CRC, like family history and lifestyle factors, were unavailable in the electronic database. However, the prevalence of positive family history of CRC would unlikely to differ between the NSAID users and nonusers as they shared similar baseline characteristics, in particular history of colonic polyps and polypectomy. Although true prevalence of smoking and alcoholism may be underestimated by diagnosis coding, cardiovascular risk factors and diseases were similar between NSAID users and nonusers (Table 1). Second, drug compliance and over-the-counter NSAID use could not be ascertained, although this is likely a nondifferential misclassification bias attenuating result to null. Third, some quality measures related to index colonoscopy, such as individual endoscopist's adenoma detection rate, quality of bowel preparation, polyp characteristics (eg number, size, histology), were not available in the database. Instead, the centre's colonoscopy volume and polypectomy rates, two surrogate markers of centre's performances, were considered. It is also unlikely that these characteristics determined NSAID use. Fourth, the causes of PCCRC-3y could not be defined, which prevent further delineation of the exact chemopreventive mechanisms of NSAIDs. Fifth, as inherent to all observational studies, residual/unmeasured confounding is possible, although a large number of variables were included to minimise this risk. Sixth, as the majority of our patients are Chinese, generalisability should be corroborated by studies on different ethnic groups.