Gram-negative bacteremia as a clinical marker of occult malignancy

Accepted Manuscript Gram-negative bacteremia as a clinical marker of occult malignancy Kirstine K. Søgaard, Dóra K. Farkas, Mette Søgaard, Henrik C. Schønheyder, Reimar W. Thomsen, Henrik T. Sørensen PII:

S0163-4453(16)30282-1

DOI:

10.1016/j.jinf.2016.09.011

Reference:

YJINF 3836

To appear in:

Journal of Infection

Received Date: 11 April 2016 Revised Date:

19 September 2016

Accepted Date: 28 September 2016

Please cite this article as: Søgaard KK, Farkas DK, Søgaard M, Schønheyder HC, Thomsen RW, Sørensen HT, Gram-negative bacteremia as a clinical marker of occult malignancy, Journal of Infection (2016), doi: 10.1016/j.jinf.2016.09.011. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Gram-negative bacteremia as a clinical marker of occult malignancy

RI PT

Running title: Gram-negative bacteremia and cancer

Kirstine K. Søgaard1, Dóra K. Farkas1, Mette Søgaard1, Henrik C. Schønheyder2,3, Reimar W. Thomsen1, and Henrik T. Sørensen1

Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Allé 43-45, 8200

SC

1

2

M AN U

Aarhus N, Denmark

Department of Clinical Microbiology, Aalborg University Hospital, Mølleparkvej 10, 9000 Aalborg,

Denmark

Department of Clinical Medicine, Aalborg University, Søndre Skovvej 15, 9000 Aalborg, Denmark

TE D

3

Key words: bacteremia; neoplasm; risk.

EP

Corresponding author: Kirstine K. Søgaard. Department of Clinical Epidemiology, Aarhus

AC C

University Hospital, Olof Palmes Allé 43-45, 8200 Aarhus N, Denmark. E-mail: [email protected]

ACCEPTED MANUSCRIPT Summary Objectives: Gram-negative bacteremia may be a harbinger of occult cancer. We examined the risk of cancer following hospitalization with bacteremia.

RI PT

Methods: Using medical databases, we conducted a nationwide population-based cohort study of all Danes with a discharge diagnosis of Gram-negative bacteremia during 1994– 2013. We calculated absolute risks and standardized incidence ratios (SIRs) of cancer,

SC

comparing the observed risk to that expected in the general population.

M AN U

Results: We observed 1,379 cancers vs. 988 expected among 11,753 patients with Gramnegative bacteremia, corresponding to an overall SIR of 1.40 (95% confidence interval (CI): 1.32-1.47). During the first 6 months following the bacteremia diagnosis, the SIR for cancer was 3.33-fold (95% CI: 2.99-3.69) increased, corresponding to an absolute risk of 3.05%. The

TE D

increased risk stemmed mainly from higher than expected occurrence of gastrointestinal cancer (3- to 13-fold higher), genitourinary cancer (4- to 10-fold higher), non-Hodgkin lymphoma (5-fold higher), non-specified metastatic cancer (5-fold higher), and breast and

EP

lung cancer (2-fold higher). The 6-12 months SIR for any cancer was 1.46 (95% CI: 1.22-1.72),

AC C

and beyond 1 year of follow-up, the SIR declined to 1.13 (95% CI: 1.05-1.20). Conclusions: Gram-negative bacteremia is a clinical marker of occult cancer.

1

ACCEPTED MANUSCRIPT Introduction Gram-negative bacteria are a common cause of bloodstream infections in the western world1,2. The most frequent source of Gram-negative bacteremia is urinary tract infection,

RI PT

followed by gastrointestinal or pulmonary infection1,2, but the source remains unidentified in around 20%-30% of cases 3,4. Importantly, bacteremia may be the first sign of an

unrecognized cancer. Disruption of the normal mucosal barrier caused by tumor growth, or

SC

obliteration of bile ducts or collecting ducts, can facilitate hematogenous spread of coexisting bacteria5,6.

M AN U

Although evidence is limited, some bacteria are thought to have carcinogenic properties7, and colonization or infection with these bacteria possibly predict long-term increased cancer risk. Case-reports and small cohort studies have suggested a link between specific Gram-positive organisms and colorectal cancer8-11. The pathomechanism could be bacterial

TE D

ability to attach to pre-neoplastic lesions in the colonic mucosa, promoting malignant progression of epithelial cells6. Lower urinary tract infection may be associated with an increased risk of bladder cancer12. The potential mechanism is inflammation-related

EP

production of nitric oxide, which promotes tumor growth and proliferation13. The finding of

AC C

reduced cancer risk associated with antibiotic treatments after cystitis supports such a direct effect 12.

To date, no studies have examined whether an episode of Gram-negative bacteremia is a clinical marker of occult cancer using a comparison cohort. We therefore conducted a population-based cohort study to examine absolute and relative cancer risk among patients with Gram-negative bacteremia compared with that in the general population.

2

ACCEPTED MANUSCRIPT Methods Data Sources and Study Population This cohort study was based on the cumulative population of approximately 7 million

RI PT

persons in Denmark between 1994 and 2013. The Danish healthcare system provides taxfunded medical care to all Danish residents and guarantees free access to hospitals and outpatient clinics14. We used data from the Danish National Patient Registry (DNPR)15, coded

SC

according to the International Classification of Diseases (ICD), Eighth Revision (during 1977-

M AN U

1993) and Tenth Revision (since 1994). In the DNPR, the main condition prompting a hospital admission is recorded as the primary diagnosis, whereas other conditions or diseases are recorded as secondary diagnoses 15.

In the current study, we identified patients with a first episode of Gram-negative bacteremia

TE D

between January 1, 1994 and November 30, 2013, using the ICD-10 diagnosis code “septicemia/sepsis due to other Gram-negative organism”. This code has an overall positive predictive value for Gram-negative bacteremia of 86%4. We then restricted the cohort to

EP

patients with a primary discharge diagnosis of Gram-negative bacteremia, and excluded patients with an earlier diagnosis of this condition during the 1977-1993 period to avoid

AC C

including recurrent cases.

Using ICD-8 and ICD-10 codes, we retrieved information from the DNPR starting in 1977 on comorbidities characterizing the study cohort. These included diabetes, chronic kidney disease, and chronic liver disease. We also identified patients with an inpatient or outpatient hospital contact for urologic or gastrointestinal infection within the 30 days prior to the hospitalization for Gram-negative bacteremia. As well, we extracted information on surgical or endoscopic gastrointestinal or urologic interventions, urinary catheterizations,

3

ACCEPTED MANUSCRIPT and imaging examinations performed within 30 days before or during the hospitalization for Gram-negative bacteremia. Our motivation for documenting patients with interventions and imaging examinations was the possibility that they had other symptoms suspicious of cancer, or that the interventions themselves provoked the bacteremia. We retrieved

RI PT

information on secondary diagnoses (grouped as defined in the ICD-system) recorded

during the same admission, to explore the prevalence of concurrent diseases or conditions

M AN U

SC

indicating an underlying focus or risk factors for the bacteremic episode.

Cancer

Since 1943 all new cancer cases in Denmark have been recorded in the Danish Cancer Registry (DCR)16, currently classified according to ICD-10 codes. We identified all incident

TE D

cancers among members of the Gram-negative bacteremia cohort by linking data between the DNPR and the DCR. Patients with a previous cancer diagnosis (other than nonmelanoma skin cancer) were excluded. We grouped cancers as gastrointestinal (cancers of

EP

the esophagus, stomach, small intestine, colon (including rectosigmoid colon), rectum, anus, liver, gallbladder, and pancreas), genitourinary (cancers of the kidney, renal pelvis, ureter,

AC C

urinary bladder, and prostate), and other cancers. All ICD and procedure codes used in the study are provided in the Appendix (see supplementary material).

Statistical Analysis Patients were followed for cancer occurrence starting from the date of their first-time inpatient hospitalization with a primary discharge diagnosis of Gram-negative bacteremia 4

ACCEPTED MANUSCRIPT until the date of death or November 30, 2013, whichever came first. We described the characteristics of Gram-negative bacteremia patients, including distributions and frequencies of gender, age categories (<40, 40-59, 60-79, and 80+ years), calendar period of bacteremia diagnosis (1994-1998, 1999-2003, 2004-2008, and 2009-2013), and the

inclusion [with interquartile range (IQR)].

RI PT

covariates discussed above. We computed median follow-up time and median age at

SC

Standardized incidence ratios (SIRs) were used as a measure of relative risk, comparing cancer incidence observed among patients with Gram-negative bacteremia with that

M AN U

expected in the entire Danish population. Expected numbers of cancer cases were estimated based on national cancer incidence rates by age (5-year age groups), sex, and calendar year (5-year periods). We computed confidence intervals (CIs) for SIRs using Byar’s approximation17. Exact 95% CIs were used when the observed number was less than ten. We

TE D

used the following categories of follow-up time: 0-<6 months, 6-<12 months, 1+ years, and the entire follow-up period. We computed SIRs for all cancers combined, for individual gastrointestinal and genitourinary cancers, and for other selected cancers. We also stratified

EP

our analyses on patient characteristics and covariates. We calculated the cumulative

AC C

incidence (or absolute risk) of cancer in patients with Gram-negative bacteremia after 6 months of follow-up, considering death as a competing risk18. We depicted the cancer incidence overall, and for gastrointestinal cancers and genitourinary cancers during the complete follow-up period of 20 years. We treated death as competing risk for cancer incidence, and specifically for gastrointestinal cancers and genitourinary cancers we also treated risk of other cancers as competing risk. In a sensitivity analysis, we excluded patients with a recent gastrointestinal or urologic intervention. This allowed us to explore whether the association persisted even after 5

ACCEPTED MANUSCRIPT exclusion of patients who already may have started to be worked up for cancer, and for whom the procedures themselves may have caused the bacteremia. All statistical analyses were conducted using the SAS statistical software package, v. 9.2 (SAS Institute, Cary, NC). The study was approved by the Danish Data Protection agency, record

RI PT

number 1-16-02-1-08. Danish registry data are generally available to researchers, and in

SC

accordance with Danish law, use of the data does not require informed consent.

Results

M AN U

Study Cohort

We identified 11,753 patients with a first-time hospitalization for Gram-negative bacteremia and no previous cancer diagnosis (Table 1). The distribution of men and women was equal (49% vs. 51%), and the median age was 75 years (IQR: 63-83 years). The total number of

years (IQR: 0.4-5.9 years).

TE D

person-years of follow-up was 45,529, corresponding to a median follow-up time of 2.5

EP

For patients diagnosed with cancer subsequently to Gram-negative bacteremia, the most frequent diseases registered as secondary diagnoses during the admission with the

AC C

bacteremic episode were: diseases of the genitourinary system (30.5%), diseases of the circulatory system (28.4%), endocrine, nutritional and metabolic diseases (19.9%), followed by diseases of the digestive system (10.9%), diseases of the respiratory system (10.4%), neoplasms (7.7%), diseases of the musculoskeletal system (7.3%), and other infectious and parasitic diseases (7.2%). A fifth of patients (20.9%) had no other disease or condition registered during the same admission for Gram-negative bacteremia (Table 2).

6

ACCEPTED MANUSCRIPT Overall Cancer Risk In total, 1,379 patients were diagnosed with cancer during follow-up, corresponding to an overall SIR of 1.40 (95% CI: 1.32-1.47). Men had a slightly higher SIR than women (Table 1). The majority of cancers were diagnosed in patients aged 60-79 years (n=766, absolute

RI PT

risk=21.13%) and in patients older than 80 years (n=406, absolute risk=11.02%). The

observed number of cancers and the absolute risk were notably high among patients aged

SC

40-59 years (n=185, absolute risk=21.77%), whereas only a few cancers were diagnosed in the youngest age group, <40 years (n=22, absolute risk=7.78%). In contrast, the relative risk

M AN U

of cancer was highest among the youngest patients and less elevated among the elderly. The overall cancer incidence treating death as competing risk is depicted in Figure 1. The risk of a cancer diagnosis was substantially increased during the first 6 months following the bacteremic episode. In total, 356 were diagnosed with cancer during this period,

TE D

corresponding to an absolute cancer risk of 3.05%, and a SIR of 3.33 (95% CI: 2.99-3.69). Among the patients diagnosed with cancer during the first 6 months, 109 were diagnosed within 30 days after discharge for bacteremia [SIR 5.51 (95% CI: 4.52-6.64)]. During 6-12

EP

months of follow-up the SIR was 1.46 (95% CI: 1.22-1.72), and after one or more years of

AC C

follow-up the SIR was 1.13 (95% CI: 1.05-1.20). The cancer SIR was higher among patients who underwent an imaging examination within 30 days before or during the hospitalization for bacteremia [SIR = 1.82 (95% CI: 1.66-2.00)], than for the overall cohort. The relative risk increased over calendar time: among patients diagnosed with bacteremia during 1994-1998, the SIR was 1.20 (95% CI: 1.07-1.35) and for patients diagnosed during 2009-2013, the SIR was 1.86 (95% CI: 1.62-2.11).

7

ACCEPTED MANUSCRIPT Gastrointestinal Cancers Among patients diagnosed with gastrointestinal cancer (n=289), the prevalence of concurrent diagnoses of diseases of the digestive system (16.6%) and neoplasms (11.4%)

bacteremia and subsequent cancer (Table 2).

RI PT

was higher than for the overall subgroup of patients diagnosed with Gram-negative

Patients in our study cohort were diagnosed with gastrointestinal cancer 3- to 13-times

SC

more frequently than expected during the first 6 months following their Gram-negative bacteremic episode. This increased risk arose from associations with esophageal, stomach,

M AN U

colon, rectal, liver, pancreatic, and gallbladder and biliary tract cancers (Table 3). However, precision for individual risk estimates was low. Although relative risks for gastrointestinal cancers were substantially elevated, the combined 6-month absolute risk of any gastrointestinal cancer was only 0.97%, and that for each cancer individually was low (Table

TE D

3). During the 6-12 month follow-up period, an excess risk was present most clearly for pancreatic and gallbladder and biliary tract cancer. Thereafter, the risk of any gastrointestinal cancer was similar to that in the general population (Table 3).

EP

When we further stratified individual cancers by selected underlying diseases, we found

AC C

that the increased risk of liver cancer associated with bacteremia was present mainly among patients with chronic liver disease [SIR= 26.99 (95% CI: 11.63-53.16)] and/or diabetes [SIR=6.51 (95% CI: 2.98-12.37)] (not presented). Compared to the general population, the overall increased risk of gastrointestinal cancer in bacteremia patients was higher among patients who underwent a surgical or endoscopic gastrointestinal intervention in the 30 days preceding their admission for bacteremia [SIR=3.99 (95% CI: 2.43-6.16)] than among patients with no intervention registered 30 days before their admission [SIR=1.57 (95% CI: 1.39-1.77)]. 8

ACCEPTED MANUSCRIPT The gastrointestinal cancers observed among patients with recent intervention were localized to the esophagus (n=1), colon cancer (n=6), rectum (n=1), liver (n=4), gallbladder and biliary tract (n=2), and pancreas (n=6). Among 192 patients (2%) with a recent gastrointestinal infection, only four patients were

RI PT

subsequently diagnosed with a gastrointestinal cancer (colon n=1, liver n= 1, pancreas n=2).

SC

Genitourinary Cancers

Among patients diagnosed with genitourinary cancer (n=255), the prevalence of concurrent

M AN U

diagnoses of diseases of the circulatory system (36.5%) and diseases of the genitourinary system (37.3%) was higher than for the overall subgroup of patients diagnosed with Gramnegative bacteremia and subsequent cancer (Table 2).

The risk of a diagnosis of genitourinary cancer was 4- to 10-fold increased within 6 months

TE D

following a hospitalization for Gram-negative bacteremia, arising from increased risks of kidney cancer [SIR=9.83 (95% CI: 5.37-16.49)], bladder cancer [SIR=4.03 (95% CI: 2.58-5.99)], and prostate cancer [SIR=5.41 (95% CI: 4.16-7.90)]. The associated 6-month absolute risks of

EP

these cancers were 0.12%, 0.21%, and 0.56%, respectively.

AC C

During 6 to 12 months of follow-up, the risk of bladder cancer remained 3-fold increased. We also observed more cases of kidney cancer than expected, but the numbers were small (3 observed vs. 1 expected). After the first year of follow-up, the excess risk of kidney cancer persisted, while urinary bladder cancer was diagnosed as frequently as expected and prostate cancer occurred with 20% fewer cases than expected, based on national cancer incidence (Table 3). When we stratified individual genitourinary cancers by underlying diseases, we found that the risk of kidney cancer was higher among patients with chronic kidney disease [SIR=8.55 9

ACCEPTED MANUSCRIPT (95% CI: 4.09-15.72)] and diabetes [SIR=4.61 (95% CI: 2.30-8.25)] than among patients without these diseases (2- to 2.5-fold increased), compared to the general population. The increased risk of genitourinary cancer overall was higher among patients who underwent a urologic intervention in the 30 days preceding their admission for Gram-negative

RI PT

bacteremia [SIR=3.36 (95% CI: 2.08-5.13)], than for patients who did not have a recent

urologic intervention [SIR=1.49 (95% CI: 1.31-1.69)] (not presented). The genitourinary

SC

cancers diagnosed among patients with a recent urologic intervention were localized to the prostate (n=7), kidney (n=1), ureter (n=1), and bladder (n=12). Patients that received any

M AN U

type of urinary catheter in the 30 days preceding their admission had a cancer risk that was similar to the overall cohort, but the result was based on small numbers.

Other Cancers

TE D

Within the first 6 months following diagnosis of Gram-negative bacteremia, occurrence of non-Hodgkin lymphoma was higher than expected [SIR=5.17 (95% CI: 3.01-8.27)], as were breast cancer [SIR=1.98 (95% CI: 1.13-3.21)], lung cancer [SIR=2.06 (95% CI: 1.20-3.13)], and

EP

non-specified metastatic cancer [SIR=5.00 (95% CI: 2.66-8.55)]. After this initial period, the

AC C

excess risk dropped for most cancers. However, for non-specified metastatic cancer it remained 3-fold increased after 6 to 12 months of follow-up and almost 2-fold increased after more than 1 year of follow-up. Gram-negative bacteremia was not associated with a higher occurrence of basal cell carcinoma.

10

ACCEPTED MANUSCRIPT Sensitivity Analysis In our sensitivity analysis, exclusion of patients with a recent gastrointestinal or urologic intervention (n=628) did not alter overall cancer risk. The SIR was 1.36 (95% CI: 1.29-1.44), similar to that in the main analysis [1.40 (95% CI: 1.32-1.47)]. However, exclusion of these

RI PT

patients attenuated the 0-<6 month SIRs of cancers of the liver, pancreas, gallbladder and biliary tract (Table 4, this also contains information on the number and types of excluded

SC

cancers).

M AN U

Discussion

In this population-based cohort study, we found that Gram-negative bacteremia was a marker of occult cancer. In particular, we observed a higher than expected risk of gastrointestinal cancer, genitourinary cancer, and non-Hodgkin lymphoma during the first 6

TE D

months after hospitalization for Gram-negative bacteremia. We also observed an increased risk of breast and lung cancer. While the excess cancer occurrence decreased within 6 months following the hospitalization, an increased risk of pancreatic and bladder cancer

EP

remained for up to 12 months, and kidney cancer and non-specified metastatic cancer

AC C

remained increased beyond 1 year of follow-up.

Several mechanisms may be at play behind the observed associations. Cancer itself may cause bacteremia, resulting in the apparent short-term risk increase of cancer among the patients. A tumor with coexisting bacteria may allow their hematogenous spread and thereby be a direct cause of bacteremia. In addition, cancer-related suppression of the immune system may promote the infection.

11

ACCEPTED MANUSCRIPT Another mechanism behind the association may be that carcinogenic properties of bacteria facilitate cancer development. Such properties have been suggested as the link between some Gram-positive organisms [Streptococcus gallolyticus (previously Streptococcus bovis), Clostridium septicum, and Bacteroides species] and colorectal cancer5,8-11,19. Bacteria may

RI PT

attach to colonic polyps and stimulate proliferation or transformation of the neoplastic

tissue 6. Certain bacteria, such as E.coli and Klebsiella species, are thought to play a causal role in development of bladder cancer by enhancing endogenous formation of N-nitroso

SC

compounds13,20.

M AN U

We restricted our study to primary diagnoses of Gram-negative bacteremia, to ensure that bacteremia was in fact the main condition leading to hospitalization. Our finding of a greatly increased short-term risk of a cancer diagnosis within 6 months implies that the cancer preceded the bacteremia, and bacteremia may even have been the first presenting

TE D

symptom leading to the cancer diagnosis. However, the overall higher cancer SIR among patients with a recent imaging examination or a gastrointestinal or urologic intervention indicates that in some cases, the work-up for cancer had already been initiated, and

EP

bacteremia was more likely a consequence of this work-up. Still, patients with these recent

AC C

interventions constituted only a small subset of all patients and in a sensitivity analysis excluding these patients the overall results and conclusions concerning early cancer risk remained unchanged.

We found that the association with some cancers apparently was driven by chronic underlying diseases, such as prevalent liver disease or diabetes that increase the risk of both Gram-negative bacteremia and cancer21-23. In addition to cancers of the gastrointestinal or urogenital tract, we found associations between Gram-negative bacteremia and other types 12

ACCEPTED MANUSCRIPT of cancer. The increased short-term risk of lung cancer is likely explained by pneumonia caused by Gram-negative bacteria such as Klebsiella spp., Pseudomonas aeruginosa or other aerobic rods. The higher than expected occurrence of non-Hodgkin lymphoma may stem from disease with extra-nodal involvement (e.g., gastrointestinal or genitourinary cancer).

RI PT

Another plausible explanation is that clinical examination revealed hepatomegaly or

splenomegaly, diagnostic tests indicated cytopenia, or chest x-ray indicated malignancy (i.e.,

SC

diagnostic bias). We speculate that the increased risk of breast cancer may stem from

immune system suppression. Importantly, we found that the risk of basal cell carcinoma,

M AN U

which is not logically associated with Gram-negative bacteremia, was around unity. This may indicate that our observed associations are not explained by surveillance bias alone (as we would then have expected to see an increased risk of any type of cancer).

TE D

Our study has a number of strengths. All Danish residents have free tax-funded access to medical care, including hospital admission and treatment, which minimizes the risk of selection bias. Data in the DNPR are recorded by the treating physician and collected mainly

EP

for administrative use, unrelated to research purposes. The registry diagnoses used in our

AC C

study generally have high validity 15,24, including for inpatient Gram-negative septicemia/sepsis diagnoses 4. While the completeness of these diagnoses may only be around 35%25, their high specificity (reflected by the high positive predictive value) likely means that the relative measures of risk are unbiased 26. A study limitation is that DNPR lack detailed information on bacterial species for patients diagnosed with Gram-negative bacteremia. The prevalence of the different pathogens for the entire cohort of Gram-negative bacteremia was unknown to us, but the majority of isolates were most likely E. coli, Klebsiella spp., and P. aeruginosa, based on findings of a 13

ACCEPTED MANUSCRIPT small subsample from the cohort4. We find it unlikely that an association with a single infrequent pathogen other than these prevalent organisms can fully explain the observed association between Gram-negative bacteremia and cancer.

RI PT

In conclusion, we found that Gram-negative bacteremia is a marker of occult

gastrointestinal and genitourinary cancer, non-Hodgkin lymphoma, and non-specified

SC

metastatic cancer. This may, together with associated chronic conditions, explain part of the increased long-term mortality observed in patients diagnosed with Gram-negative

M AN U

bacteremia. Although the absolute risk of cancer is low, our findings suggest that a thorough diagnostic work-up in selected patients hospitalized with Gram-negative bacteremia may be

AC C

EP

TE D

appropriate.

14

ACCEPTED MANUSCRIPT Funding This work was supported by the Program for Clinical Research Infrastructure (PROCRIN) established by the Lundbeck Foundation and the Novo Nordisk Foundation; the Aarhus University Research Foundation; Manufacturer Einar Willumsen Foundation; Inge and

RI PT

Jørgen Larsen Foundation; Helga and Peter Kornings Foundation; and Director Jacob

Madsen and Wife Olga Madsen Foundation. The sponsors did not influence the design or

SC

conduct of the study; collection, management, analysis, and interpretation of the data;

preparation, review, approval of the manuscript; or the decision to submit the manuscript

M AN U

for publication.

Author Contributions

TE D

KKS conceived the study idea. KKS reviewed the literature. HTS collected the data. KKS, MS, HCS, RWT and HTS designed the study and directed the analyses, which were carried out by DKF. All authors participated in the interpretation of the results. KKS organized the writing

EP

and wrote the initial drafts. All authors critically revised the manuscript for intellectual

AC C

content and approved the final version.

Disclosure of Conflicts of Interest None of the authors report conflicts of interests, financial interests, activities, relationships, or affiliations relevant to this study.

15

ACCEPTED MANUSCRIPT References

1. Laupland KB, Church DL. Population-based epidemiology and microbiology of communityonset bloodstream infections. Clin Microbiol Rev., 2014; 27: 647-664.

RI PT

2. Reimer LG, Wilson ML, Weinstein MP. Update on detection of bacteremia and fungemia. Clin Microbiol Rev. 1997; 10: 444-465.

SC

3. Sogaard M, Norgaard M, Dethlefsen C, Schonheyder HC. Temporal changes in the

incidence and 30-day mortality associated with bacteremia in hospitalized patients from

M AN U

1992 through 2006: A population-based cohort study. Clin Infect Dis, 2011; 52: 61-69.

4. Sogaard KK, Thomsen RW, Schonheyder HC, Sogaard M. Positive predictive values of the international classification of diseases, 10th revision diagnoses of Gram-negative

2015; 7: 195-199.

TE D

septicemia/sepsis and urosepsis for presence of Gram-negative bacteremia. Clin Epidemiol,

EP

5. Ellmerich S, Scholler M, Duranton B, et al. Promotion of intestinal carcinogenesis by Streptococcus bovis. Carcinogenesis, 2000; 21: 753-756.

AC C

6. Tjalsma H, Scholler-Guinard M, Lasonder E, Ruers TJ, Willems HL, Swinkels DW. Profiling the humoral immune response in colon cancer patients: Diagnostic antigens from Streptococcus bovis. Int J Cancer, 2006; 119: 2127-2135.

7. Abreu MT, Peek RM,Jr. Gastrointestinal malignancy and the microbiome. Gastroenterology, 2014; 146: 1534-1546.e3.

16

ACCEPTED MANUSCRIPT 8. McKenna AJ, O'Donnell ME, McMullan R, Irwin T. Long-term gastrointestinal outcomes after Streptococcus bovis bacteraemia. Int J Clin Pract, 2011; 65: 1203-1205.

9. Boleij A, van Gelder MM, Swinkels DW, Tjalsma H. Clinical importance of Streptococcus

RI PT

gallolyticus infection among colorectal cancer patients: Systematic review and metaanalysis. Clin Infect Dis. 2011; 53: 870-878.

4 cases. Case Rep Med, 2011; 2011: 248453..

SC

10. Mao E, Clements A, Feller E. Clostridium septicum sepsis and colon carcinoma: Report of

M AN U

11. Yoshino Y, Kitazawa T, Ikeda M, et al. Clinical features of Bacteroides bacteremia and their association with colorectal carcinoma. Infection, 2012; 40: 63-67.

12. Vermeulen SH, Hanum N, Grotenhuis AJ, et al. Recurrent urinary tract infection and risk

TE D

of bladder cancer in the nijmegen bladder cancer study. Br J Cancer, 2015; 112: 594-600.

13. Xu W, Liu LZ, Loizidou M, Ahmed M, Charles IG. The role of nitric oxide in cancer. Cell

EP

Res, 2002; 12: 311-320.

14. Ministry of health and prevention. eHealth in denmark. 2012. available at:

AC C

http://Www.sum.dk/~/media/Filer%20-%20Publikationer_i_pdf/2012/sundhedsIT/Sundheds_IT_juni_web.ashx. 2012. Accessed June 25, 2015.

15. Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sørensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol, 2015; 7 :449-90.

16. Gjerstorff ML. The Danish Cancer Registry. Scand J Public Health, 2011; 39: 42-45. 17

ACCEPTED MANUSCRIPT 17. Breslow NE, Day NE. Statistical methods in cancer research. volume II--the design and analysis of cohort studies. IARC Sci Publ, 1987; 82: 1-406.

18. Kim HT. Cumulative incidence in competing risks data and competing risks regression

RI PT

analysis. Clin Cancer Res, 2007; 13: 559-565.

19. Sears CL, Geis AL, Housseau F. Bacteroides fragilis subverts mucosal biology: From

SC

symbiont to colon carcinogenesis. J Clin Invest, 2014; 124: 4166-4172.

20. Mostafa MH, Sheweita SA, O'Connor PJ. Relationship between schistosomiasis and

M AN U

bladder cancer. Clin Microbiol Rev. 1999; 12:97-111.

21. Thomsen RW, Hundborg HH, Lervang HH, Johnsen SP, Schonheyder HC, Sorensen HT. Diabetes mellitus as a risk and prognostic factor for community-acquired bacteremia due to

628-631.

TE D

Enterobacteria: A 10-year, population-based study among adults. Clin Infect Dis. 2005; 40:

474.

EP

22. Botwin GJ, Morgan TR. Bacterial infections in cirrhosis. Hepatol Int, 2014; Suppl 2: 467-

AC C

23. Sorensen HT, Friis S, Olsen JH, et al. Risk of liver and other types of cancer in patients with cirrhosis: A nationwide cohort study in Denmark. Hepatology, 1998; 28: 921-925.

24. Storm HH, Michelsen EV, Clemmensen IH, Pihl J. The Danish Cancer Registry--history, content, quality and use. Dan Med Bull, 1997; 44: 535-539.

18

ACCEPTED MANUSCRIPT 25. G Storm HH, Michelsen EV, Clemmensen IH, Pihl J. The Danish Cancer Registry--history, content, quality and use. Dan Med Bull, 1997; 44: 535-539.

26. Schneeweiss S, Avorn J. A review of uses of health care utilization databases for

AC C

EP

TE D

M AN U

SC

RI PT

epidemiologic research on therapeutics. J Clin Epidemiol, 2005; 58: 323-337.

19

ACCEPTED MANUSCRIPT Table 1. Characteristics of 11,753 patients with a first-time inpatient Gram-negative bacteremia diagnosis, the observed and expected number of cancers, and standardized cancer incidence ratios (SIRs) versus the general population Patients (N, %) 11,753 (100)

Observed/Expected* cancers (N) 1379/988

Cancer SIR* (95% CI) 1.40 (1.32-1.47)

AC C

EP

TE D

M AN U

SC

RI PT

All patients Gender Women 6036 (51) 577/462 1.25 (1.15-1.36) Men 5717 (49) 802/526 1.52 (1.42-1.63) Age group, years <40 767 (7) 22/9 2.35 (1.47-3.55) 40-59 1704 (14) 185/107 1.73 (1.49-1.99) 60-79 5138 (44) 766/567 1.35 (1.26-1.45) 80+ 4144 (35) 406/304 1.33 (1.21-1.47) Calendar period 1994-1998 2121 (18) 281/234 1.20 (1.07-1.35) 1999-2003 3043 (26) 418/326 1.28 (1.16-1.41) 2004-2008 3485 (30) 449/304 1.48 (1.34-1.62) 2009-2013 3104 (26) 231/124 1.86 (1.62-2.11) Chronic diseases Chronic kidney disease 1482 (13) 165/84 1.96 (1.67-2.28) Chronic liver disease 800 (7) 84/38 2.22 (1.77-2.75) Diabetes 2328 (20) 252/177 1.42 (1.25-1.61) Acute infection within 30 days Gastrointestinal 192 (2) 21/11 1.91 (1.18-2.91) Urologic 732 (6) 73/58 1.26 (0.99-1.58) GastrointesVnal intervenVon† 30 days before 449 (4) 62/28 2.20 (1.68-2.82) During current admission 753 (6) 124/63 1.96 (1.63-2.34) Urologic intervention† 30 days before 194 (2) 44/23 1.90 (1.38-2.56) During current admission 185 (2) 29/18 1.63 (1.09-2.34) Urinary catheterization$ 30 days before 167 (2) 18/12 1.44 (0.85-2.28) 153 (2) 11/7 1.47 (0.73-2.64) During current admission $‡ Imaging examination 3360 (51) 448/246 1.82 (1.66-2.00) Abbreviations: O/E, observed/expected; CI, confidence interval * The expected number is calculated based on national cancer statistics. The SIR is thus a relative risk measure estimating risk in the patient cohort versus the general population. † Endoscopy or surgery. These analyses were motivated by the possibility that the patients likely had other symptoms that occurred before bacteremia and led to the cancer diagnosis, or that the interventions themselves may have provoked bacteremia. $ Data available only since 2002. ‡ Performed within 30 days before or during the hospitalization for Gram-negative bacteremia.

20

ACCEPTED MANUSCRIPT Table 2. Secondary diagnoses registered among Gram-negative bacteremia patients subsequently diagnosed with cancer, n (%) Gastrointestinal cancer (n=289)

Genitourinary cancer (n=255)

Infectious and parasitic diseases

99 (7.2)

16 (5.5)

20 (7.8)

Neoplasms

106 (7.7)

Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism

61 (4.4)

275 (19.9)

33 (11.4)

20 (7.8)

20 (6.9)

10 (3.9)

65 (22.5)

40 (15.7)

SC

Endocrine, nutritional and metabolic diseases

RI PT

Overall cancer (n=1379)

78 (5.7)

9 (3.1)

10 (3.9)

Diseases of the nervous system

43 (3.2)

9 (3.1)

8 (3.1)

15 (1.1)

5 (1.7)

1 (0.4)

18 (1.3)

5 (1.7)

7 (2.3)

391 (28.4)

66 (22.8)

93 (36.5)

144 (10.4)

37 (12.8)

22 (8.6)

150 (10.9)

48 (16.6)

24 (9.4)

Diseases of the skin and subcutaneous tissue

14 (1.0)

3 (1.0)

5 (2.0)

Diseases of the musculoskeletal system and connective tissue

100 (7.3)

21 (7.3)

15 (5.9)

Diseases of the genitourinary system

420 (30.5)

63 (21.8)

95 (37.3)

Congenital malformations, deformations and chromosomal abnormalities

15 (1.1)

2 (0.7)

1 (0.4)

Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified

77 (5.6)

16 (5.5)

20 (7.8)

Injury, poisoning and certain other consequences of external causes

35 (2.5)

8 (2.8)

6 (2.4)

Factors influencing health status and contact with health services

125 (9.1)

26 (9.0)

23 (9.0)

No secondary diagnoses registered

288 (20.9)

61 (21.1)

57 (22.4)

Diseases of the eye and adnexa Diseases of the ear and mastoid process Diseases of the circulatory system Diseases of the respiratory system

AC C

EP

TE D

Diseases of the digestive system

M AN U

Mental and behavioural disorders

21

ACCEPTED MANUSCRIPT

Table 3. Standardized incidence ratios (SIRs) of cancer after a first-time inpatient diagnosis of Gram-negative bacteremia 0-<6 months

6-<12 months

SIR (95% CI)

Absolute risk (%)

Cancer O/E

SIR (95% CI)

356/107

3.33 (2.99-3.69)

3.05

133/91

1.46 (1.22-1.72)

Esophagus

6/1

5.10 (1.87-11.13)

0.05

2/1

2.01 (0.24-7.26)

Stomach

7/2

4.10 (1.65-8.45)

0.06

2/1

Colon

42/9

4.77 (3.44-6.45)

0.36

Rectum

14/4

3.43 (1.87-5.76)

Liver

11/1

Pancreas Gallbladder/biliary tract

Overall

Cancer O/E

SIR (95% CI)

Cancer O/E

SIR (95% CI)

890/790

1.13 (1.05-1.20)

1379/988

1.40 (1.32-1.47)

14/8

1.71 (0.93-2.86)

22/10

2.12 (1.33-3.21)

1.39 (0.17-5.03)

9/11

0.79 (0.36-1.50)

18/15

1.24 (0.73-1.96)

10/7

1.33 (0.64-2.45)

60/63

0.95 (0.73-1.22)

112/79

1.41 (1.16-1.70)

0.12

3/3

0.87 (0.18-2.53)

30/29

1.05 (0.71-1.50)

47/36

1.30 (0.95-1.73)

12.91 (6.44-23.11)

0.09

2/1

2.77 (0.32-9.99)

8/6

1.32 (0.57-2.61)

21/8

2.76 (1.71-4.21)

24/3

9.46 (6.06-14.07)

0.21

11/2

5.06 (2.52-9.06)

23/19

1.23 (0.78-1.85)

58/23

2.48 (1.89-3.21)

5/1

7.94 (2.57-18.50)

0.04

3/1

5.59 (1.15-16.31)

3/4

0.67 (0.14-1.97)

11/6

1.96 (0.98-3.50)

Kidney

14/1

9.83 (5.37-16.49)

0.12

3/1

2.47 (0.51-7.22)

20/10

1.97 (1.20-3.04)

37/13

2.89 (2.03-3.98)

Bladder

24/6

4.03 (2.58-5.99)

0.21

14/5

2.80 (1.53-4.70)

42/40

1.06 (0.77-1.44)

80/50

1.58 (1.26-1.97)

Prostate

65/12

5.41 (4.18-6.90)

0.56

7/10

0.70 (0.28-1.43)

66/82

0.80 (0.62-1.02)

138/104

1.32 (1.11-1.56)

Non-Hodgkin lymphoma

17/3

5.17 (3.01-8.27)

2/3

0.71 (0.09-2.57)

28/24

1.16 (0.77-1.67)

47/30

1.55 (1.14-2.06)

Breast

16/8

1.98 (1.13-3.21)

0.14

4/7

0.56 (0.15-1.43)

54/66

0.81 (0.61-1.06)

74/82

0.91 (0.71-1.14)

Lung

22/11

2.06 (1.29-3.13)

0.19

11/9

1.21 (0.60-2.17)

83/76

1.09 (0.87-1.35)

116/96

1.21 (1.00-1.45)

Basal cell carcinoma

16/21

0.77 (0.44-1.25)

0.14

21/18

1.17 (0.72-1.79)

183/165

1.11 (0.95-1.28)

220/204

1.08 (0.94-1.23)

4/2

2.65 (0.72-6.77)

0.03

0/1

-

11/11

0.97 (0.48-1.74)

15/14

1.06 (0.59-1.75)

6/2

2.73 (1.00-5.94)

29/18

1.62 (1.08-2.32)

48/23

2.11 (1.56-2.80)

All Cancer

Brain

M AN U

TE D

0.15

AC C

Other cancers

EP

Genitourinary cancer

Non-specified metastatic 13/3 5.00 (2.66-8.55) 0.11 cancer Abbreviations: O/E, observed/expected; CI, confidence interval

SC

Gastrointestinal cancer

RI PT

Cancer O/E

1+ years

22

ACCEPTED MANUSCRIPT

Table 4. Standardized incidence ratios (SIRs) of cancer after a first-time inpatient diagnosis of Gram-negative bacteremia and no gastrointestinal or urologic intervention in the 30 days preceding the admission 6-<12 months

SIR (95% CI)

Absolute risk (%)

Cancer O/E

SIR (95% CI)

320/102

3.15 (2.81-3.53)

2.90

124/87

1.43(1.19-1.70)

Esophagus

6/1

5.41 (1.98-11.78)

0.05

2/1

2.13 (0.26-7.69)

Stomach

7/2

4.32 (1.73-8.90)

0.06

2/1

Colon

38/8

4.53 (3.20-6.21)

0.34

Rectum

13/4

3.36 (1.79-5.74)

Liver

9/1

Pancreas Gallbladder/biliary tract

Overall

Cancer O/E

SIR (95% CI)

Cancer O/E

SIR (95% CI)

832/750

1.11 (1.04-1.19)

1276/939

1.36 (1.29-1.44)

13/8

1.68 (0.89-2.88)

21/10

2.15 (1.33-3.28)

1.47 (0.18-5.29)

9/11

0.84 (0.38-1.59)

18/14

1.31 (0.78-2.07)

10/7

1.40 (0.67-2.57)

57/60

0.95 (0.72-1.23)

105/76

1.39 (1.13-1.68)

0.12

3/3

0.91 (0.19-2.66)

30/27

1.11 (0.75-1.58)

46/34

1.34 (0.98-1.79)

11.16 (5.11-21.21)

0.08

2/1

2.92 (0.35-10.55)

6/6

1.05 (0.39-2.29)

17/7

2.36 (1.38-3.78)

18/2

7.45 (4.41-11.77)

0.16

11/2

5.31 (2.65-9.50)

21/18

1.18 (0.73-1.81)

50/22

2.25 (1.67-2.96)

4/1

6.65 (1.81-17.02)

0.04

2/1

3.89 (0.47-14.06)

3/4

0.71 (0.15-2.06)

9/5

1.68 (0.77-3.18)

Kidney

13/1

9.65 (5.13-16.50)

0.12

3/1

2.61 (0.54-7.63)

18/10

1.88 (1.11-2.96)

34/12

2.81 (1.95-3.93)

Bladder

20/6

3.55 (2.17-5.48)

0.18

11/5

2.33 (1.16-4.17)

35/37

0.94 (0.65-1.31)

66/48

1.39 (1.07-1.76)

Prostate

58/11

5.15 (3.91-6.66)

0.53

6/9

0.64 (0.23-1.39)

62/76

0.81 (0.62-1.04)

126/97

1.30 (1.08-1.54)

All cancer

TE D

EP

Genitourinary cancer

M AN U

Gastrointestinal cancer

SC

Cancer O/E

1+ years

RI PT

0-<6 months

AC C

Abbreviations: O/E, observed/expected; CI, confidence interval This analysis excluded 628 patients who underwent a gastrointestinal or urologic intervention (endoscopy or surgery) within 30 days before hospitalization for Gramnegative bacteremia. In total, 103 cancers occurred among the excluded patients [including cancers of the esophagus (n=1), colon (n=7), rectum (n=1), liver (n=4), pancreas (n=8), gallbladder and biliary tract (n=2), kidney (n=3), bladder (n=14), prostate (n=12), non-Hodgkin lymphoma (n=5), breast (n=2), lung (n=8), basal cell carcinoma (n=16), non-specified metastatic cancer (n=4), and others (n=16)].

23

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 1. Cumulative incidence (%) of cancer treating death as competing risk.

24

ACCEPTED MANUSCRIPT Highlights Gram-negative bacteremia is a clinical marker of occult cancer

-

The absolute risk of cancer within the first 6 months after bacteremia was 3%

-

The overall risk of cancer subsequent to bacteremia was 40% higher than expected

AC C

EP

TE D

M AN U

SC

RI PT

-