Impact of carbon monoxide poisoning on the risk of breast cancer

Carbon monoxide (CO) is a toxic gas and an endogenous signaling molecule. Some studies involving cell lines have revealed the potential antibreast cancer effects of CO. Data on such effects in humans, however, are limited. Thus, we conducted a study on patients with CO poisoning (COP) to evaluate the effects of CO on the risk of breast cancer. We identified female patients who were diagnosed with COP over the period of 2002 and 2009 from the Nationwide Poisoning Database of Taiwan. For comparison, we selected females without COP from the National Health Insurance Research Database. Participants in the COP and comparison cohorts were matched on the index year, age, monthly income, and geographic region of residence at a 1:6 ratio. We followed up the two cohorts until the end of 2014 and compared their risks of developing breast cancer. We included 7053 participants with COP and 42,318 participants without COP. Participants with COP were at a lower risk of developing breast cancer than those without COP (0.7% vs. 1.0%, p < 0.001). Cox proportional hazard regression analyses revealed that COP was associated with a hazard ratio of 0.67 (95% confidence interval [95% CI] 0.50–0.90) for breast cancer after we adjusted for age, monthly income, geographic region, and comorbidities of hypertension, diabetes, and hyperlipidemia. Our result provides evidence for the potential protective effects of CO against breast cancer in humans. Further studies that directly evaluate the potential effects are warranted.


Variable definitions.
We defined a patient with COP as a participant who was assigned diagnosis codes 986, E868, E952, or E982 in accordance with the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) during hospitalization or a visit to the emergency department. A patient with breast cancer was defined as a participant who had been assigned ICD-9-CM diagnosis codes 174 or 175 during at least one hospitalization or at least three visits for ambulatory care. Those who were diagnosed with breast cancer before the index date were excluded from the study.
We categorized the participants into five groups on the basis of age: < 20, 20-34, 35-49, 50-64, and ≥ 65 years 3 . Common underlying comorbidities, including hypertension (ICD-9-CM 401-405), diabetes (ICD-9-CM 250), and hyperlipidemia (ICD-9-CM 272), were included for analyses. A patient with these diseases was defined as a participant who had been assigned the relevant codes during at least one hospitalization or at least three visits for ambulatory care before the index date 3 . We categorized the participants into three groups in accordance with monthly income: < 20,000, 20,000-40,000, and > 40,000 New Taiwan Dollars (NTD) 3 . Statistical methods. We followed up the two cohorts until 2014 to compare their breast cancer risks.
We applied independent t-tests to evaluate differences in continuous variables and χ 2 tests to evaluate those in categorical variables. Cox proportional hazard regression with competing risk analysis was used to identify the independent predictors of breast cancer and evaluate their effects. We also performed multivariate regression to adjust for potential confounding effects. In addition, we applied the Kaplan-Meier method and the log-rank test to compare the breast cancer risks of the two cohorts during the follow-up period. Given that the risk of COPassociated breast cancer might change over time, we conducted further analyses by using a cutoff of 1 year for follow-up duration. All analyses were performed by using SAS 9.4 for Windows (SAS Institute, Cary, NC, USA) at a two-tailed significance level of 0.05.

Results
We included 7053 female patients with COP and 42,318 females without COP in this study (Fig. 1, Table 1). In the COP cohort, the mean age was 34.1 years (standard deviation = 14.4 years), and the 20-34-year-old group had the highest number of participants (3650, 51.8%), followed by the 35-49-year-old group (2552, 35.8%). No differences in the distributions of age, monthly income, geographic region or the prevalence of hypertension, diabetes, or hyperlipidemia existed between the two cohorts. The COP cohort had a lower risk of breast cancer than the comparison cohort (0.7% vs. 1.0%, p < 0.001). The average age at diagnosis of breast cancer was similar between COP and comparison cohorts (50.0 vs. 49.2 years old, p = 0.569). Cox proportional hazard regression with competing risk analysis showed that COP was associated with a hazard ratio (HR) of 0.67 (95% confidence interval [CI] 0.50-0.91; p = 0.009) ( Table 2). The decrease in the risk associated with COP persisted after we adjusted for age, monthly income; geographic region; and hypertension, diabetes, and hyperlipidemia comorbidities (adjusted HR [AHR]: 0.67; 95% CI 0.50-0.90; p = 0.009). The Kaplan-Meier's method and log-rank test also showed that the COP cohort had a lower breast cancer risk than the comparison cohort (Fig. 2).
In further analyses stratified by follow-up duration, we found that COP was associated with a HR of 0.52 (95% CI 0.12-2.21; p = 0.378) in the first year of follow-up (Table 3). Moreover, the AHR remained the same as the HR after we adjusted for age; monthly income; geographic region; and hypertension, diabetes, and hyperlipidemia comorbidities (0.52, 95% CI 0.12-2.21; p = 0.377). The reduction in risk was larger than that observed over the whole follow-up period but did not reach statistical significance. COP was associated with a HR of 0.70 (95% CI 0.52-0.96; p = 0.024) ( Table 4) after 1 year of follow-up, and the AHR was close to the HR after adjustment for other variables (0.71, 95% CI 0.52-0.96; p = 0.026). We divided COP cohort into those with one and multiple episodes COPs and compared them with the comparison cohort. The result showed that both those with one and multiple episodes of COP had a lower risk for breast cancer than the comparison cohort (0.70% and 0.30%, respectively). However, the difference between those with one and multiple episodes of COP did not reach statistical significance (Supplementary Table S1). The Kaplan-Meier's method and log-rank test also showed that the COP cohort was at a lower risk for breast cancer than the comparison cohort in the first year of follow-up (see Supplementary Fig. S1) and after 1 year of follow-up (see Supplementary Fig. S2).

Discussion
This study revealed that female patients with COP had a lower risk for breast cancer than those without COP. This epidemiologic study provided an interesting finding and indirect evidence for the potential role of CO in breast cancer treatment. The reduction in breast cancer risk after CO exposure may be attributed to the possible direct toxic effect of CO and to the death of cancer cells due to severe hypoxia. Vítek et al. used CO gas (500 ppm 1 h/day) to treat mice that had been xeno transplanted subcutaneously with pancreatic cancer cells 15 . They found that CO exposure significantly inhibits the proliferation of human pancreatic cancer cells and doubles the survival rates Table 2. Independent predictors for breast cancer in all patients of the two cohorts during the overall follow-up period. The independent predictors were identified through competing risk regression analysis. COP carbon monoxide poisoning, HR hazard ratio, AHR adjusted hazard ratio, HR SD adjusted competing risks hazard ratio, CI confidence interval, NTD New Taiwan Dollars. *Adjusted for age, hypertension, diabetes, hyperlipidemia, monthly income, and geographic region.  16 . The induction of apoptosis in lung tumors is associated with the increased expression of CD86 and the activation of mitogen-activated protein kinase/extracellular signal-regulated kinases 1/2 pathway 16 . Grau et al. found that CO inhalation could increase tumor hypoxia, which may affect tumor control 17 . Whereas hypoxia may play a role in tumor progression 18 , an acute episode of severe hypoxia may lead to the death of cancer cells before the resistance to hypoxic environment being developed. Therefore, we may reasonably speculate that CO exposure before the clinical diagnosis of breast cancer might affect precancerous cells or existing cancer cells in patients with COP. Studies on endogenous CO may cast some light on the anticancer effect of CO. Endogenous CO is a byproduct of the oxidative conversion of heme 1 . The conversion of heme to biliverdin, ferrous iron, and CO is catalyzed by HO-1 and HO-2 1 . Biliverdin is further reduced to bilirubin by biliverdin reductase 1 . The role of HO-3 is not fully understood. HO-3 is suspected to be a pseudogene that is derived from HO-2 transcripts 19 . HO-1 is found in the spleen, liver, vascular endothelial cells, and smooth muscle tissues 1 . It is the only inducible HO isoform, and its increase is stimulated by cellular stress 1 . HO-2 is responsible for neurotransmission and vascular tone regulation 9,20 . HO-2 and HO-3 are ubiquitously expressed in the brain, liver, and testes 9,20 . CO maintains cell and tissue homeostasis via its antiapoptotic, anti-inflammatory, and antioxidant effects 1 . CO also has antiproliferative and vasodilative effects and may participate in tissue regeneration and in strengthening the innate immune system 1,10 .
Although CO may have therapeutic applications, its possible toxicity stemming from its effect on oxygen transport and toxic dose control from systemic CO gas administration are major concerns 1 . COP may contribute to hypoxia and inflammation and subsequently to neurologic and cardiac dysfunction, injury to other organs, and even death [2][3][4][5][6][7][8] . In recent years, CORMs, a group of transition metal carbonyls or boranocarbonates that can release CO upon transformation, have provided another avenue for CO application 1,9,10,20 . High CO concentrations exert cytotoxic effects via the inhibition of the mitochondrial respiratory system, the induction of oxidative stress, and the production of reactive oxygen species 1,21 . CORMs may enable the localized release of high amounts of CO for specific cytotoxicity against targeted tumors 1 .
A growing number of studies have revealed that CO administration is an emerging hope for cancer treatment. Lee et al. found that treatment with RuCO, a type of CORM, reduced the growth of human MCF7 and MDA-MB-231 breast cancer cells 11 . RuCO down-regulated the expression of growth-related proteins, including cyclinD1, CDK4, and hTERT 11 . Given that HSP90 stabilizes several proteins required for tumor growth, the feasibility of using HSP90 inhibitors as anticancer drugs has been investigated 22 . The contradictory effects of RuCO treatment on wild-type and mutant p53 proteins are similar to those of cells treated with geldanamycin, a HSP90 Table 3. Independent predictors for breast cancer in all patients of the two cohorts in the first year of follow-up. The independent predictors were identified through competing risk regression analysis. COP carbon monoxide poisoning, HR hazard ratio, AHR adjusted hazard ratio, HR SD adjusted competing risks hazard ratio, CI confidence interval, NTD New Taiwan Dollars. *Adjusted for age, hypertension, diabetes, hyperlipidemia, monthly income, and geographic region.  26 . In addition to CDRMs, CO gas also showed anticancer potential against pancreatic cancer 15 and lung adenocarcinoma 27 . The major strength of the present study is its nationwide population-based design with a large sample. While the novel finding of a decreased risk for breast cancer in COP patients has implicated the potential use of CO as a therapeutic agent, this study has several limitations. First, information on several risk factors, including family history, reproductive history, physical activity, and body mass index, is not available in the NHIRD. Consequently, we were unable to adjust for the effects of these potential confounders. Second, we did not recruit male participants given the rare incidence of breast cancer in the male population. Therefore, the results of this study might not be applied to the male population. Third, the participants were relatively young (about 34 years old on average at the beginning of follow-up). However, nearly half of patients (48.2%) were aged ≥ 35 years initially, and the peak age at the diagnosis of breast cancer in Taiwanese women was 45 to 50 years 28 . Therefore, we believe following up the participants for 12 years in the present study is sufficient to cover the age at the highest risk for a substantial portion of them. Fourth, we did not evaluate the survival advantage associated with COP or compare the distributions of breast cancer stage and ER/PR/HER2 status between patients with and without COP because they are out of the scope of the present study and the databases do not contain the some of the information. Separate studies are needed to clarify these issues. Fifth, the number of breast cancers was relatively small due to the relatively young age at the beginning of follow-up. Nonetheless, the size of patients was large enough to provide sufficient statistical power to detect the effect of COP on breast cancer. As to the potential antitumor effect, recruiting more patients and further animal and laboratory studies are needed to support its clinical application.

Conclusions
This study demonstrated that female patients with COP had a lower risk of breast cancer than those without COP. This result may be attributed to the direct and indirect inhibitory effects of CO on tumor growth. Further studies involving collection of complete variables for possible confounders in the patients with breast cancer as well as animal and laboratory trial experiments on the mechanisms are warranted. Table 4. Independent predictors for breast cancer in all patients of the two cohorts after 1 year of follow-up. The independent predictors were identified through competing risk regression analysis. COP carbon monoxide poisoning, HR hazard ratio, AHR adjusted hazard ratio, HR SD adjusted competing risks hazard ratio, CI confidence interval, NTD New Taiwan Dollars. *Adjusted for age, hypertension, diabetes, hyperlipidemia, monthly income, and geographic region.