Smoking and body mass index as risk factors for subtypes of cancer of unknown primary

Cancer of unknown primary site (CUP) accounts for 3–5% of all cancer but because it is a fatal disease it ranks high among causes of cancer deaths. It has been called “an orphan disease” because it is diagnosed through metastatic tissue and the primary tumor often remains undetected. Based on autopsy data, the most common primaries were lung (27%), pancreas (24%), liver/bile (8%) and kidney/adrenals (8%).1 CUP can be called “orphan disease” also because, as a metastatic disease, it is disregarded as a cancer entity when only primary cancers are considered. Thus, hardly anything is known about the underlying causes or risk factors. A recent European study published in this Journal showed that heavy smoking causes a risk of 3.66 for CUP patients of diverse and unspecified histologies; alcohol drinking and wide waist circumference were borderline risk factors.2 The pathophysiology of CUP has remained an enigma, exemplified by the question presented in a review on CUP: “unknown primary or unknown biology?”3 The inability to find a primary tumor may not merely depend on its small size but may involve newly described cell biological processes related to stem cell characteristics, microenvironment, immunoediting and senescence-apoptosis which may halt tumor growth and induce phenotypic changes.3-5 In essence, an increased understanding of the interplay of host and tumor during the protracted process of carcinogenesis has made CUP less of a conundrum and more as a learning opportunity, in a 21st century parallel to the cancer biology gains provided by unravelling of several hereditary cancer syndromes in the previous century. Our recent studies support these data and suggest that the lung is a more common primary tissue than what the earlier data show.6-8 In the present letter, we extend the study of Kaaks et al. addressing the question of tobacco smoking and body mass index (BMI) as risk factors of site-specific CUP, that is, by location of the diagnosed metastasis.2 We hypothesize that smoking would have an effect on sites where smoking risks are highest among primary cancers, that is, lung. Patients and cancer-free controls were recruited from three Swedish prospective biobanks, the Umea Medical Biobank with 203 CUP patients and 1015 controls and the Malmo Diet and Cancer Study and the Prevention Study, jointly with 272 CUP patients and 1360 controls;9 the controls were frequency matched on sex and age at interview. In Umea, the mean age at sampling was 57 years and at diagnosis it was 66 years; in Malmö, the respective ages were 51 and 71 years. Cancers were identified up to year 2012 through the regional Oncology Centers. Diagnoses were based on International Classification of Diseases (ICD) codes 7, 9 and 10, the latter two were particularly useful because they included the anatomic site where metastases were found. These included “unspecified CUP” (ICD-9 code 1990–1991, CUP often spread to multiple organs), “liver CUP” with liver metastasis, “lung CUP” with lung involvement (including thorax and lymph nodes and organs, brain and bone, for which lung cancer is usually given as the cause of death in CUP patients7, 8), “abdominal CUP” with abdominal metastases (including ovary) and “other CUP” with other metastatic locations (any other specified site). Histology was adenocarcinoma in 82.1%, squamous cell carcinoma in 5.0% and melanoma in 8.8% of the cases: 4.1% had miscellaneous or missing histology. Logistic regression analysis was used in the analysis of data. Detailed data on BMI were available but the dichotomous cutoff was at 20 kg/m2, best discriminated the risk groups. The study was approved by the ethics committee at Umea University. Linkage with cancer data up to year 2012 identified 475 CUP patients on whom smoking data were available on 463 CUP patients and BMI data on 447 patients. Unspecified CUP was the largest disease subgroup (43.4% of all cases), followed by lung (16.2%), abdomen (14.5%), liver (13.8%) and other CUP (12.1%). Smoking was a risk factor for all CUP cases taken together with an OR of 1.82 but the effect was particularly strong on lung CUP with an OR of 4.90 (Table 1). The commonest subgroup of unspecified CUP was also influenced by smoking (OR 1.89) as was liver CUP (OR 2.03). In univariable analysis (not shown), it was observed that the only BMI class that affected CUP was <20 kg/m2 and thus in the logistic regression analysis we included this and 20+ as the two distinct groups. BMI 20+ showed a weak, nonsignificantly decrease for all CUP with an OR of 0.77. The only subtype for which BMI 20+ showed a significant decrease was liver CUP with an OR 0.23. In prospective analysis, it is of concern if the interviews have taken place shortly before diagnosis when the growing cancer might have already influence the physiology of the future patients. We thus restricted the population to those who were interviewed >1 or >5y before diagnosis. The results showed some minor changes for smoking but not for BMI. The OR of smoking was 5.98 among those interviewer >5y before lung CUP diagnosis. The present series of 475 cases is the largest uniform collection of patients with CUP yet studied, and the only one where risk factors could be assessed by site-specific CUP. The recent European multicenter study on 273 adenocarcinoma patients from nine countries, including 44 patients from Sweden, which were most likely part of the present study as well.2 Smoking was a risk factor for CUP with an OR of 1.82 but the effect was particularly strong on lung CUP with an OR of 4.90. The OR was 5.93 in individuals sampled 5 years before CUP diagnosis. BMI 20+ showed decreased risk for liver CUP with an OR 0.23. The results highlight the etiological heterogeneity of CUP, smoking being specific to lung-related CUP and BMI influencing only liver CUP. The data suggest that the effect of smoking on the lung and the assumed effect of alcohol on the liver were due to local primary tumors not recognized by the tissue-of-origin determinations. Yours sincerely, Kari Hemminki Bowang Chen Olle Melander Jonas Manjer Göran Hallmans Akseli Hemminki