Impact of hormone replacement therapy on all‐cause and cancer‐specific mortality in colorectal cancer: A systematic review and dose‒response meta‐analysis of observational studies

The effect of hormone replacement therapy (HRT) on colorectal cancer (CRC) mortality and all‐cause mortality remains unclear. We conducted a systematic review and dose–response meta‐analysis to determine the effects of HRT on CRC mortality and all‐cause mortality.


INTRODUCTION
Colorectal cancer (CRC) is the third most common cancer worldwide and the third leading cause of cancer-related deaths in women in the United States. 1--3Most CRC patients have sporadic cancer, and approximately 75% of rectal cancer patients and 80% of sporadic colon cancer patients are ≥ 60 years old. 4,5The incidence rate of colorectal cancer is lower in women (14.3 per 100,000 individuals) than in men (20.6 per 10,000). 6The cumulative lifetime risk of developing CRC in the United States is approximately 6%. 7 Despite advances in treating this disease, the 5-year survival rate in the United States is only 62%. 8 The lower incidence of CRC in women indicates that female hormones may exhibit protective effects. 9[12][13][14][15] Despite the well-established association between HRT use and a lower risk of CRC, the influence of postmenopausal HTR use on the survival of patients with established CRC remains uncertain.Previous literature has explored this relationship and indicated better survival after CRC diagnosis in some cases, but the results are still conflicting. 16,17The dose-response relationship between the duration of HRT use and all-cause mortality has not been reported.Therefore, we conducted a comprehensive systematic review and dose-response meta-analysis of observational studies to estimate summary hazard ratios for the associations between HRT exposure and the risk of CRC mortality and all-cause mortality.

Data sources and search strategy
According to the Cochrane Handbook for Systematic Reviews 18 and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines, 19 we searched the electronic databases of PubMed, Embase, and The Cochrane Library for all relevant studies published until January 2024 to investigate the effects of HRT exposure on survival rates for CRC patients (Supplementary Information).The following search terms were used in combination: (colorectal neoplasms OR colonic neoplasms OR rectal neoplasms) AND ("hormone replacement therapy" OR estrogens OR progestins OR progesterone OR testosterone) AND (mortality OR survival OR death).We also searched the reference lists of included studies to identify additional research that could have been missed during the literature search and selection.
We restricted our language to studies published in English.We also searched Google Scholar using similar keywords.Our protocol was registered with the OSF (Registration DOI: https://doi.org/10.17605/OSF.IO/38JS7).

Study selection and inclusion criteria
Two independent investigators read all the citations identified by the literature search for potentially relevant studies.We cross-checked the relevant articles selected for inclusion, and disagreements were resolved through discussion or conferring with senior investigators.
Articles were included for meta-analysis if they met the following inclusion criteria: (1) they were observational studies (including prospective or retrospective cohorts) that explored the association between HRT and prognostic outcomes for patients with CRC; (2) they reported hazard ratios and 95% confidence intervals or provided indirect data for their calculations; (3) the outcomes were mortality from all causes or CRC.We did not exclude non-English language studies or studies that were abstracts.Articles (i.e., reviews, meta-analyses, commentaries, case series, editorials, and letters) without original survival data were excluded.

Data extraction and quality assessment
Data extraction was performed by two independently trained reviewers using standardized forms.The following variables were extracted from each study: first author, publication year, research country, number of cases, CRC stage, follow-up period, sample origin, type of HRT, survival endpoints, adjusted variables, hazard ratios (HRs), and 95% CIs with and without adjustment for confounding factors.The primary outcome measures of our study were CRC mortality and all-cause mortality.The study quality was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies. 20,21Studies with scores less than 4 were considered to have a high risk of bias, those with scores of 4 to 6 an intermediate risk of bias, and those with scores of 7 or more a low risk of bias. 22,23Conflicts between reviewers were resolved by discussion.

Data synthesis and analysis
All statistical analyses were performed with Stata version 14.0 (Stata Corp LP, College Station, TX, USA).Due to the inherent heterogeneity in our study, we applied DerSimonian and Laird random effects models. 24Statistical heterogeneity was estimated using the I 2 statistic, with a p value less than 0.1.An I 2 value above 50% indicated significant heterogeneity. 25 assumed the width of the category to be the same as the adjacent category when the highest category was open-ended and set the lower boundary to zero when the lowest category was openended.The midpoint of the range was used when the duration of HRT use was reported.In addition, studies that perform doseresponse meta-analyses must include HR values, the corresponding confidence intervals, and the duration of HRT use.To examine a possible nonlinear relationship between the duration of HRT use and CRC mortality, we performed a two-stage random-effects dose-response meta-analysis. 26,27We modeled the year of HRT use using restricted cubic splines limited by four knots at fixed percentiles (5%, 35%, 65%, and 95%) of the distribution.In the first stage of our analysis, acknowledging the correlation within each set of published HRs, we fitted a restricted cubic spline model with three spline transformations, equating to four knots minus one.In the second stage, we combined the three regression coefficients estimated in each study and the variance/covariance matrix.Subsequently, we applied the multivariate extension of the method of moments to conduct a multivariate random-effects meta-analysis. 28We used a likelihood ratio to test for nonlinearity and checked the goodness of fit for the nonlinear model compared to the linear model.The overall p value was calculated by simultaneously testing whether the three regression coefficients were zero.The nonlinear p value was calculated by testing if the coefficient of the third spline equaled zero. 29 explore sources of heterogeneity, we carried out a series of subgroup analyses based on study design (prospective cohort or retrospective cohort), region (USA, Sweden), patient age (<60 years or >60 years), duration of HRT use (≤5 years or > 5 years), location (colon, rectal), research center (single, multicenter), type of HRT (estrogen, estrogen+ progestin, contraceptive), and duration of HRT (former, current).We assessed publication bias using Begg's rank correlation test 30 and Egger's linear regression test, 31 with a p value < 0.05 indicating the existence of publication bias.We performed sensitivity analyses to apply the leave-one-out method for both all-cause and CRC mortality.All p -values were two-sided, and a p -value < 0.05 was considered statistically significant.

RESULTS
In the preliminary analysis, the search strategy yielded 4277 studies.After removing duplicates, 3915 studies remained for the title and abstract screening.We excluded 3886 irrelevant citations during this process, and 29 potentially relevant studies remained for full-text review.After excluding reviews, meta-analyses, or studies that did not report outcome data, ten studies were found eligible for inclusion in this systematic review and dose-response meta-analysis.
Figure 1 shows the results of the study selection.Data from 10 studies were included in the final analysis, which involved 480,628 participants. 32--41All studies were population-based.Seven studies were from the United States, and three were from Sweden.The median follow-up period was 5.4 to 13 years (two studies did not provide the median follow-up period).Details of the included studies are provided in Table 1.In one study, the exposed population was nurses, and the exposed population was underrepresented, while the other studies were general population.The control population in all studies was drawn from the same population as the exposure group.Outcome measures in all studies were derived from rigorous questionnaires.Four studies were comparable and adjusted for confounding factors.In all the studies, outcomes were measured by means of reliable records, with adequate follow-up for the disease under study, and with good completeness.All studies had a low risk of bias (NOS score ≥ 7).Detailed results of the risk of bias assessments for each study are provided in Appendix 2 of the Supplementary Information.

Associations between HRT use and the risk of CRC mortality
Ten cohort studies 32--41 were included in our analysis of the association between HRT use and CRC mortality.The meta-analysis showed that HRT was associated with a 23% reduction in the risk of CRC mortality (HR = 0.77, 95% CI (0.68, 0.87), I 2 = 69.5%,p < 0.05, Figure 2).A dose-response analysis showed strong evidence of a linear (p for nonlinearity = 0.34) association between HRT and CRC mortality based on five cohort studies, such that each additional year of HRT use was associated with a 3% reduction in the risk of CRC mortality; this decrease was statistically significant (HR = 0.97, 95% CI (0.94−0.99), p < 0.05, Figure 3).

Subgroup analysis
Given the significant heterogeneity in the meta-analysis of all included studies, we performed subgroup analyses to identify the sources of the heterogeneity.We conducted subgroup analyses based on study

Associations between HRT use and the risk of all-cause mortality
Meta-analyses for all-cause mortality included seven cohort studies. 33,36--41HRT was associated with a 29% reduction in the risk of all-cause mortality (HR = 0.71, 95% CI (0.54, 0.92), I 2 = 89.6%,p < 0.05, Figure 4).A linear dose-response analysis (p for nonlinearity = 0.88) of three cohort studies showed that each additional year of HRT use was associated with a 2% decrease in the risk of all-cause mortality; this finding was not statistically significant (HR = 0.98, 95% CI (0.97, 1.01), p = 0.08, Figure 5).

Subgroup analysis
Given the significant heterogeneity in the meta-analysis of all included studies, we performed subgroup analyses to better understand the heterogeneity.The decrease in risk remained constant among other subgroups stratified by study design, study region, research center, study setting, time of HRT usage, age, and type of HRT.Furthermore, the results of the subgroup analyses showed significant differences between groups by study design, study region, research center, HRT use time, and HRT type (p for between-group difference < 0.05).The results are shown in Table 3.

Sensitivity analyses and publication bias
The results of the sensitivity analyses for the outcomes of CRC mortality and all-cause mortality were consistent with those of the primary analyses based on cohort studies (Figure 6A

DISCUSSION
This systematic review and dose-response meta-analysis represents the most comprehensive review of the association between HRT use and CRC.We included ten cohort studies involving 480,628 individuals.The primary objective was to determine whether HRT use was associated with survival outcomes in CRC.The pooled analysis suggested that HRT was associated with a 29% reduction in the risk of total all-cause mortality and a 23% reduction in the risk of CRC-specific mortality.However, the subgroup analysis showed inconsistent results when considering different study types and HRT estimates.Several extensive observational studies suggest that using hormones can reduce CRC mortality.Specifically, the current use of HRT was significantly associated with a reduced risk of CRC mortality.In contrast, the former use of HRT did not correlate with a reduced risk of CRC mortality.Furthermore, HRT use can reduce the risk of CRC mortality but not the risk of rectal cancer mortality.Finally, our dose-response meta-analysis suggests that HRT use significantly decreased CRC and all-cause mortality rates over time.
The mechanism of the effect of HRT on CRC is unknown.HRT may reduce colorectal neoplasia, including direct or indirect reductions in secondary bile acid production and insulin-like growth factor I inhibition. 42There are also several other mechanisms through which hormone use may protect against the development of CRC and improve survival after diagnosis.Exogenous estrogens could lead to slower disease progression.Several studies have confirmed that estrogen inhibits cell growth in human CRC cell lines and that loss of estrogen results in increased proliferation of normal colonic cells. 43--45Furthermore, exogenous estrogens have been associated with changes that prevent methylation, inactivate estrogen receptors, and increase cell growth. 46 contrast, insulin-like growth factor I appears to stimulate epithelial cell proliferation. 42,47Our conclusions show that HRT is significantly associated with CRC mortality in colon cancer, but not in rectal cancer.
However, the mechanism of this difference needs to be clarified, which is the direction of future research.
A systematic review 48 published in 2019 showed that the current use of HRT was associated with a lower risk of CRC and overall mortality.An earlier systematic review 49 published in 1999 showed that the summary relative risk of colon cancer death in HRT users was 0.72 (95% CI 0.64, 0.81) compared to nonusers.Our findings show that HRT TA B L E 3 Subgroup analyses in a subset of included studies for the risk of all-cause mortality.

Limitations
The present meta-analysis has several limitations.First, we observed significant heterogeneity among the included studies, which was anticipated and may arise in part from differences in the baseline characteristics of the populations (such as age, location, HRT type, duration of HRT use, lifestyle, and genetic predispositions) and the statistical methods used (particularly adjustments for confounders).
Although several techniques were used to adjust the results, moderate to high heterogeneity remained.However, the consistency of most of the subgroup and sensitivity analyses with the leading results suggests that this heterogeneity has a limited impact on the study's primary conclusions.Second, the publication bias analysis was limited by the limited number of included studies, reducing its statistical power.
Third, some subgroup analyses yielded nonsignificant results, which could be attributed to the relatively small sample sizes and low statistical power.More evidence from high-quality prospective cohort studies on the effect of HRT on CRC mortality would be beneficial.
Last, our meta-analyses were restricted to studies published in English in peer-reviewed journals.Consequently, relevant articles published in other languages or journals beyond the three databases we searched (PubMed, Embase, and the Cochrane Library) might have been overlooked.We did not include unpublished grey literature.These three main databases have more than 80-90% of all available reports.
Including unpublished, nonpeer-reviewed literature could introduce bias.
The clinical implications of our study indicate that hormone therapy, administered both before and after a CRC diagnosis, is associated with a reduced risk of CRC mortality and all-cause mortality.This underscores the need for additional research to elucidate the mechanisms by which HRT affects colorectal carcinogenesis and cancer progression.

CONCLUSIONS
This dose-response meta-analysis of cohort studies indicates an inverse association between HRT use and all-cause and CRC mortality, demonstrating a significant reduction in mortality rates over time.However, the extent to which postmenopausal hormone therapy reduces CRC mortality requires further investigation.
and B).Sensitivity analyses were performed using the leave-one-out method to further examine TA B L E 1 Study characteristics of the included studies.

F I G U R E 2
Forest plots for meta-analysis of the effect of HRT use on the risk of CRC mortality.F I G U R E 3 Dose-response meta-analysis of HRT use time and risk of CRC mortality based on cohort studies.the stability of the results.We found that no individual study significantly altered the summary HR (lowest RR = 0.75, 95% CI (0.69, 0.80); highest RR = 0.78, 95% CI (0.68, 0.89)).The funnel plots for CRC mortality demonstrate specific evidence of publication bias determined by the Begg's test (p = 0.82) and Egger test (p = 0.38) (Figure 7A).The funnel plots for all-cause mortality also demonstrate evidence of publication bias as determined by the Begg's test (p = 0.92) and the Egger test (p = 0.25) (Figure 7B).However, there was no evidence of significant publication bias according to Begg's test and Egger's test.
Alternatively, estrogens may inhibit cancer progression through other mechanisms.HRT can reduce the occurrence of colorectal tumors, including through direct or indirect reduction of secondary bile acid production and inhibition of insulin-like growth factor I. In F I G U R E 4 Forest plots for meta-analysis of the effect of HRT use on the risk of all-cause mortality.F I G U R E 5 Dose-response meta-analysis of HRT use time and risk of all-cause mortality based on cohort studies.

F I G U R E 6
Sensitivity analysis based on cohort studies.can reduce the risk of CRC mortality.Furthermore, our study included more articles and performed a subgroup analysis.Our study has several strengths.First, this is the first dose-response meta-analysis incorporating cohort studies exploring whether hor-mones can reduce CRC and all-cause mortality.We investigated the linear relationship between the duration of hormone use and CRC mortality and all-cause mortality by dose-response meta-analysis.Second, the large sample size of 480,628 allowed us to assess the F I G U R E 7 Assessment of publication bias based on cohort studies.relationship between HRT and CRC quantitatively, making this the most potent and comprehensive evidence synthesis.Third, all the included studies were from nationwide or population-based cohorts, thereby minimizing the risk of selection bias originating from the study design.Fourth, we developed a detailed and simultaneous search strategy for each database with the help of a professional librarian, thus maximizing the possibility of identifying all relevant research.Fifth, several approaches, including subgroup and sensitivity analyses, have been applied to thoroughly determine sources of heterogeneity based on the abstracted baseline characteristics at the study level.
Subgroup analyses in a subset of included studies for the risk of CRC mortality.