SEARCH

SEARCH BY CITATION

Keywords:

  • Bootstrap approach;
  • colon cancer;
  • epidemiology;
  • pelvic inflammatory disease

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Objective

Inflammation is an important risk factor for the development of colorectal cancer (CRC). Pelvic inflammatory disease (PID) comprises a spectrum of upper genital tract infections and inflammatory diseases. We aimed to evaluate the association between CRC and PID.

Design

Matched cohort study using the National Health Insurance Research Database.

Setting

Women with PID in Taiwan.

Population and sample

From the Longitudinal Health Insurance Database 2005 (LHID2005) in Taiwan, we obtained data on women from 13 to 45 years of age who were diagnosed with PID. The women with PID were matched 1:4 to selected members of the population without PID based on age and year of first entry into the LHID2005.

Methods

A Cox proportional hazards model was used to evaluate the hazard ratio for CRC during the 5-year follow-up period, after adjusting for sociodemographic characteristics and selected comorbid medical disorders.

Main outcome measures

Colorectal cancer.

Results

Of the 19 029 women with PID, 30 were diagnosed with CRC during the 78 965 person-year follow-up period. Of the 76 116 control women, 66 were diagnosed with CRC. The CRC hazard ratio during the 5-year follow-up period was 2.00 (95% CI 1.30–3.08) for women with PID relative to control women. Similarly, after adjusting for age, Charlson comorbidity index score, urbanisation level and monthly income, the adjusted CRC hazard ratio was 1.71 (95% CI 1.10–2.65) for the women with PID relative to the women in the comparison cohort.

Conclusions

Here we show a weak association between PID and CRC. Additional studies are needed to further evaluate this association and examine plausible mechanisms, including the influence of specific microorganisms.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Colorectal cancer (CRC) is the third most common malignancy and the fourth most common cause of mortality worldwide.[1] Surprisingly, only approximately 20% of CRC incidence has a familial basis.[2] The largest fraction of CRC cases has been linked to environmental causes rather than heritable genetic changes; inflammation is also likely to be involved with other forms of sporadic and heritable colon cancer.[3] Pelvic inflammatory disease (PID) typically results from an ascending infection through the endocervix, from the lower to the upper genital tract and into the pelvic organs. Inflammation is an important risk factor for the development of CRC.[3] People with long-standing inflammatory bowel disease have an increased risk of developing CRC.[4] Because PID is a pelvic infection, it would be interesting to determine whether pelvic inflammation increases the risk for CRC. To date, no studies have evaluated the relationship between PID and CRC. We therefore examined the association between CRC and a history of PID in women of reproductive age.

Furthermore, PID comorbidity was evaluated in this study. The Charlson comorbidity index (CCI) is a system for classifying severity that uses recorded data on secondary diagnoses to assign a weight to morbidity.[5] The comorbidity score was measured by the CCI using diagnoses recorded in the Taiwan National Health Insurance Database before the index date. The CCI has demonstrated excellent predictive validity for a variety of clinical outcomes and numerous malignancies. Therefore, the CCI score for women with PID was also evaluated in this study.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

A matched cohort study was performed. The data used in this study were obtained from the Longitudinal Health Insurance Database 2005 (LHID2005), which is part of the Taiwan National Health Insurance Database. The LHID2005 contains claims data on ambulatory care, inpatient care and claimed medical expenses, as well as demographic information on 1 000 000 beneficiaries sampled randomly from the 25.68 million enrolees in Taiwan. The study comprised all women who visited ambulatory care centres for the treatment of PID (International Classification of Diseases, 9th revision, clinical modification codes 614 and 615) between 1 January 2004 and 31 December 2005. The diagnosis of PID was made on the basis of the women's medical history, clinical features at presentation and findings on bimanual palpation, as well as on ultrasonography, culture of vaginal samples and laboratory data (= 33 118). There were 33 118 women diagnosed with PID. In addition, to ensure the validity of the PID diagnosis, only women who had at least two consensus diagnoses of PID were selected (11 332 women were excluded). Women who were younger than 13 years of age and older than 45 years of age were excluded from the cohort (2027 women were excluded). Furthermore, women with missing data (= 706) and those who had a previous history of colon cancer (= 24) were excluded. Our cohorts included 19 029 women with PID and 76 116 (1:4) control women. A minimum of 6 months follow up was required, and the women with PID could not have had colon cancer in the previous 6 months.

The comorbidity score was measured based on the 17 Charlson comorbidities[6] using the diagnoses recorded in the Taiwan National Health Insurance Database before the index date. Baseline variables, including age, urbanisation level, monthly income and CCI score, were obtained for all participants.

Statistical analysis

Pearson's chi-square tests and Fisher's exact test were used to compare differences, as indicated in Table 1. We used the Cox proportional hazards model to examine the hazard ratio for colon cancer between the two cohorts after adjusting for age, sex, urbanisation level, CVD, chronic liver disease, chronic renal failure and autoimmune disease. To meet the proportional hazards assumption, the proportionality of each dichotomous variable in the model was checked using investigative diagnostic log–log survival plots.

Table 1. Baseline variables, including demographic characteristics and comorbid medical disorders of the two cohorts (= 95 145)
Baseline variableWomen with PID n = 19 029Control women = 76 116P value
n % n %
Characteristics
Age (years)
13–30757639.830 30439.8 
31–40646234.025 84834.0
41–45499126.219 96426.2
The year of entry
200411 76961.847 07661.8 
2005726038.229 04038.2
Urbanisation level
Urban11 57660.848 19963.3<0.001
Suburban559529.421 08227.7
Rural18589.868359.0
Comorbid medical disorders
Endometriosis
Yes5833.19711.3<0.001
No18 44696.975 14598.7
CCI
Myocardial infarction
Yes50.0180.00.835
No19 024100.076 098100.0
Congestive heart failure
Yes660.31490.2<0.001
No18 96399.775 96799.8
Peripheral vascular disease
Yes1050.62250.3<0.001
No18 92499.475 89199.7
Cerebrovascular disease
Yes1540.83870.5<0.001
No18 87599.275 72999.5
Dementia
Yes40.0230.00.051
No19 025100.076 093100.0
Chronic pulmonary disease
Yes17929.451806.8<0.001
No17 23790.670 93693.2
Connective tissue disease-rheumatic disease
Yes3271.79571.3<0.001
No18 70298.375 15998.7
Peptic ulcer disease
Yes306216.161618.1<0.001
No15 96783.969 95591.9
Mild liver disease
Yes18149.541685.5<0.001
No17 21590.571 94894.5
Diabetes without complications
Yes4782.512241.6<0.001
No18 55197.574 89298.4
Diabetes with complications
Yes460.21450.20.158
No18 98399.875 97199.8
Paraplegia and hemiplegia
Yes1360.73650.5<0.001
No18 89399.375 75199.5
Renal disease
Yes1420.73920.5<0.001
No18 88799.375 72499.5
Cancer
Yes2211.27671.00.061
No18 80898.875 34999.0
Moderate or severe liver disease
Yes70.0250.00.791
No19 022100.076 091100.0
Metastatic carcinoma
Yes80.0370.00.709
No19 021100.076 079100.0
AIDS/HIV
Yes00.040.00.317
No19 029100.076 112100.0
CCI score (SD)0.440.740.270.58<0.001

To ensure the validity of the results of our Cox proportional hazards model, we performed sensitivity analyses using the bootstrap approach. The bootstrap technique is used to estimate the accuracy of an estimator, such as the standard error, a confidence interval, or the bias of an estimator. The technique is useful for analysing small expensive-to-collect datasets where previous information is sparse, distributional assumptions are unclear, and further data may be difficult to acquire.[7] The advantage of the bootstrap resampling technique is that it avoids distributional assumptions and system error bias, and it provides validation of the Cox proportional hazards model. The bootstrap hazard ratio for CRC corresponded to the median hazard ratio observed among 1000 replications, and the 95% CI was derived by using the bootstrap bias-corrected and accelerated confidence interval.[8] All of the data analyses were performed with the SAS statistical package (v.9.1.3; SAS Institute, Cary, NC, USA). Two-tailed significance level of α = 0.05 was used to determine the statistical significance.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

We assessed the data collected from 19 029 women with PID and 76 116 control women. The baseline characteristics are shown in Table 1. Comorbidity was measured using the CCI using the diagnoses recorded in the NHI before the index date. Women with PID had higher rates of comorbid endometrial disease, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, connective tissue disease, rheumatic disease, peptic ulcer disease, mild liver disease, diabetes without complications, paraplegia and hemiplegia, and renal disease (all P < 0.001). The CCI score was significantly higher in women with PID than in the women in the comparison cohort (0.44 ± 0.74 versus 0.27 ± 0.58, P < 0.001).

Of the 19 029 women with PID, 30 (a rate of 37.9 per 100 000 person-years) were diagnosed with CRC during the 78 965-person-year follow-up period. Of the 76 116 control women, 66 (a rate of 19.7 per 100 000 person-years) were diagnosed with CRC during the 334 156-person-year follow-up period. The hazard ratio for CRC during the follow-up period was 2.00 (95% CI 1.30–3.08) for the women with PID relative to the women in the comparison cohort. After adjusting for age, CCI score, urbanisation level and endometriosis, the adjusted CRC hazard ratio was 1.71 (95% CI 1.10–2.65) for the women with PID relative to the women in the comparison cohort (shown in Table 2).

Table 2. Incidence, crude and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for colon cancer among the women with PID during the 5-year follow-up period
Presence of colon cancerControl womenWomen with PID
  1. Adjustments were made to hazard ratio for age, CCI scores, endometriosis and urbanisation: *P < 0.05, **< 0.001.

Follow-up period
Yes/Total66/76 11630/19 029
Person-years334 15678 965
Incidence per 100 000 person-years (95% CI) 19.7 (15.5–25.1)37.9 (26.6–54.2)
Crude HR (95% CI) 1.002.00** (1.30–3.08)
Adjusted HR (95% CI) 1.001.71* (1.10–2.65)

To ensure the validity of the results of our Cox proportional hazards model, we performed sensitivity analyses using the bootstrap approach. The crude and adjusted bootstrap hazard ratios and the 95% confidence intervals for CRC for the women with PID relative to the women in the comparison cohort during the 5-year follow-up period (validation for the Cox model) were calculated. The CRC hazard ratio during the follow-up period was 1.69 (95% CI 1.11–2.59) for the women with PID, which was greater than the CRC hazard ratio for the women in the comparison cohort. Similarly, after adjusting for age, CCI score, urbanisation level and monthly income, the adjusted CRC hazard ratio was 1.55 (95% CI 1.01–2.38) for the women with PID relative to the women in the comparison cohort. The CRC hazard curves calculated using the Kaplan–Meier method for women with PID relative to the comparison cohort during the 5-year follow-up period are shown in Figure 1 (P < 0.01).

image

Figure 1. A plot of CRC Kaplan–Meier hazard ratios for women with PID relative to control women during the 5-year follow-up period.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Main findings

In this study, women with PID had a CRC hazard ratio of 2.00 (95% CI 1.30–3.08) relative to controls over a 5-year follow-up period. After adjusting for age, CCI score, urbanisation level and endometriosis, our results showed that the risk of CRC increased by up to 71% in women with PID compared with the control group. The CCI is a scoring system that includes weighting factors based on disease severity. To assess the exact association of PID with CRC, CCI scores were calculated and adjusted in the two groups: women with PID and control women. We calculated the hazard ratio after adjusting for age, CCI score, urbanisation level and monthly income. The risk for CRC was still significantly increased in women with PID (hazard ratio 1.55, 95% CI 1.01–2.38).

Strengths and weaknesses

To the best of our knowledge, this is the first large-scale, nationwide cohort study to examine whether PID increases the risk of CRC. The strengths and limitations are mainly determined by the database that was used. The data used in this study were obtained from the LHID2005, which did not contain some important information regarding the women who were studied; therefore, variables such as body mass index, history of inflammatory bowel disease, and dietary habits could not be analysed. Furthermore, the relatively short follow-up time (5 years) for evaluating the development of CRC is another limitation of this study. However, the large size of the database used for this study might compensate for the above-mentioned limitation. Our cohorts included 19 029 women with PID and 76 116 (1:4) control participants (all women).

Interpretation

The link between inflammation and the development of colorectal cancer is becoming increasingly clear. It has long been recognised that people with inflammatory bowel disease are at an increased risk of colon cancer.[9] The risk for colon cancer increases with the duration and anatomical extent of colitis and the presence of other inflammatory disorders, whereas it decreases with the use of anti-inflammatory medications.[10] PID comprises a spectrum of inflammatory disorders in women that result from an infection ascending from the cervix into the involved pelvic organs. PID has been associated with ovarian cancer[11, 12] and cervical cancer.[13] However, an association between PID and CRC has never been reported.

Ulcerative colitis and Crohn's disease are the two major types of inflammatory bowel disease. Bernstein and colleagues reported increased incidence rates and rate ratios of colon carcinoma in both people with Crohn's disease (2.64; 95% CI 1.69–4.12) and those with ulcerative colitis (2.75; 95% CI 1.91–3.97).[14] However, although Jess and colleagues recently reported that a diagnosis of ulcerative colitis or Crohn's disease no longer seems to increase risk of CRC, the author also suggested that this finding might be due to improved therapies for people with inflammatory bowel disease.[15]

The connection between inflammation and tumorigenesis is well established, and a large amount of genetic, pharmacological and epidemiological data that support this link have been collected.[3] PID is a polymicrobial infection resulting from the acquisition of sexually transmitted pathogens, and it results in the ascending spread of aerobic and anaerobic vaginal bacteria, which leads to inflammation of the endometrium, uterus, fallopian tubes, adnexal structures and pelvic peritoneum.[16] The pelvic organs that can be affected by inflammatory conditions in women with PID include the colon and rectum. Chronic inflammation is believed to promote carcinogenesis. If microbiota are involved in the development of cancer, the colon must be their major site of action. The predominantly anaerobic microbiota of the distal ileum and colon contains an extraordinarily complex variety of metabolically active bacteria and fungi that closely interact with the host's epithelial cells and mucosal immune system.[17] Chronic inflammation may stimulate the production of cytokines and chemokines that contribute to the development or activation of malignant disease in vitro.[18] Activated inflammatory cells produce reactive oxygen species and reactive nitrogen intermediates that can induce DNA damage and mutation.[19] Reactive oxygen species can affect the regulation of genes that encode factors that prevent carcinogenesis.[10] Inflammatory cytokines and chemokines promote tumour cell growth, perturb their differentiation, and support cancer cell survival.[20] Cells from inflamed colonic mucosa demonstrate these genetic alterations before there is any histological evidence of dysplasia or cancer.[10] Persistent inflammation facilitates tumour promotion by activating the proliferation and anti-apoptotic properties of premalignant cells, as well as tumour progression and metastasis.[21] Individual components of the innate and adaptive immune response have also been implicated in carcinogenesis.[10] To explain the association between PID and CRC—the risk for developing cancer increases substantially as a result of poorly regulated inflammatory responses to pathogenic bacterial infections might be one of the reasons.[22]

The Bradford Hill criteria,[14] otherwise known as Hill's criteria for causation, are a group of minimal conditions that are necessary to provide adequate evidence of a causal relationship between an incidence and a consequence. Briefly, the Bradford Hill criteria are as follows: (1) strength of association, (2) consistency, (3) specificity, (4) temporal relationship, (5) biological gradient, (6) plausibility, (7) coherence, (8) experiment and (9) analogy. This study, which demonstrates a weak association between PID and CRC, adds further support to the hypothesis that inflammation or inflammatory responses play a causal role in colorectal carcinogenesis. Further studies are now required to investigate the specific role of infections of the lower genital tract in the aetiology of CRC.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

This study finds a weak association between PID and CRC. Further studies are needed to confirm this association and examine plausible mechanisms, including the influence of specific microorganisms.

Disclosure of interests

The authors have no conflict of interest to disclose.

Contribution to authorship

MIH explained the results of data analysis and wrote the paper; HWL designed the study, analysed the data and wrote the paper. Both authors critically revised the manuscript for important intellectual content and approved the final version.

Details of ethics approval

The present study used anonymised secondary data from the NHI dataset, and therefore, it was exempt from full institutional review board approval.

Funding

The study received no external funding.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

This study is partially based on data from the National Health Insurance Research Database, which was provided by the Bureau of National Health Insurance, Department of Health. This study was conducted by the National Health Research Institutes, and it was financially supported by the National Science Council of Taiwan under grant number 101-2118-M-031-001-MY2. The interpretation and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health or National Health Research Institutes.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Commentary on ‘Risk of colorectal cancer in women with pelvic inflammatory disease: a matched cohort study’

Numerous studies have documented an association between inflammatory bowel disease and colorectal cancer. The paper by Hsu and Lin in the current issue of BJOG is the first to purport an association between pelvic inflammatory disease (PID) and colorectal cancer risk.

How can we explain this association? Chlamydia is known to inhibit apoptosis (Fan et al. J Exp Med 1998;187:487–96). An obligate intracellular pathogen, Chlamydia has evolved strategies to protect its host cell against death-inducing stimuli to allow it to complete its lifecycle. Recent data suggest that Chlamydia-infected cells are able to resist apoptosis through stabilisation of the cellular protein hypoxia-inducible factor-1α (Sharma et al. Cell Microbiol 2011;13:1573–85). A cell protected from apoptosis has more opportunity to harbour genetic anomalies that subsequently predispose to tumorigenesis.

Beyond a possible role for Chlamydia, numerous biological mechanisms support the link between inflammation and cancer. In their seminal paper Hallmarks of cancer: the next generation, Hanahan and Weinberg postulated that inflammation facilitates tumorigenesis by delivering several bioactive molecules to the tumour microenvironment, including reactive oxygen species, growth factors, pro-angiogenic factors and extracellular matrix-modifying enzymes that assist tumour growth, angiogenesis, invasion and metastasis (Hanahan et al. Cell 2011;144:646–74).

It is therefore tempting to accept the association between PID and colorectal cancer documented by Hsu and Lin as fact. But before we start recommending colonoscopy for all women with a history of PID, we should scrutinise the data and be open minded to all possibilities.

The authors' decision to include women aged 13–45 years is unusual, especially as the incidence of colorectal cancer in women in their teens is extremely low. Some women were only followed up for a total period of 6 months, which is of course inadequate, particularly for the younger members of the cohort. The authors were also limited by the information held in the database. There was no information on potential confounding factors including body mass index, history of inflammatory bowel disease or genetic predisposition. In addition to flagging up an association between PID and colorectal cancer risk, the analysis also identified a number of somewhat surprising associations with PID, including statistically significant associations with congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, connective tissue disease, rheumatic disease, peptic ulcer disease, mild liver disease, diabetes without complications, paraplegia and hemiplegia, and renal disease. It is difficult to come up with a plausible biological mechanism that links all of these disparate diagnoses. Furthermore, if PID does predispose to cancer, one might intuitively expect that it would predispose to gynaecological cancer, as this is the primary site of infection, rather than colorectal cancer. The most likely explanation is that some other, unmeasured factor links PID and colorectal cancer in this cohort of women. We hope this paper stimulates further debate in this area.

Disclosure of interests

The authors have nothing to disclose.

  • EJ Crosbiea, E Buescherb, M Mahmoudb & F Nezhat

  • aUniversity of Manchester, Manchester, UK

  • bSt Luke's and Roosevelt Hospitals and Columbia University, New York, NY, USA