Alcohol consumption, cigarette smoking and cancer of unknown primary risk: Results from the Netherlands Cohort Study

Abstract Cancer of unknown primary (CUP) is a metastasised malignancy with no identifiable primary tumour origin. Despite the frequent occurrence and bleak prognosis of CUP, research into its aetiology is scarce. Our study investigates alcohol consumption, tobacco smoking and CUP risk. We used data from the Netherlands Cohort Study, a cohort that includes 120 852 participants aged 55 to 69 years, who completed a self‐administered questionnaire on cancer risk factors at baseline. Cancer follow‐up was established through record linkage to the Netherlands Cancer Registry and Dutch Pathology Registry. After 20.3 years of follow‐up, 963 CUP cases and 4288 subcohort members were available for case‐cohort analyses. Multivariable‐adjusted hazard ratios (HRs) were calculated using proportional hazard models. In general, CUP risk increased with higher levels of alcohol intake (P trend = .02). The association was more pronounced in participants who drank ≥30 g of ethanol per day (HR: 1.57, 95% confidence interval [CI]: 1.20‐2.05) compared to abstainers. Current smokers were at an increased CUP risk (HR: 1.59, 95% CI: 1.29‐1.97) compared to never smokers. We observed that the more the cigarettes or the longer a participant smoked, the higher the CUP risk was (P trend = .003 and P trend = .02, respectively). Interaction on additive scale was found for participants with the highest exposure categories of alcohol consumption and cigarette smoking frequency and CUP risk. Our findings demonstrate that alcohol consumption and cigarette smoking are associated with increased CUP risk. Lifestyle recommendations for cancer prevention regarding not drinking alcohol and avoiding exposure to smoking are therefore also valid for CUP.


| INTRODUCTION
Cancer of unknown primary (CUP) is a heterogeneous group of metastasised malignancies with no identifiable primary tumour origin. 1,2 Cancer treatment, if any, is generally based on the primary tumour origin, which makes treating CUP challenging. Another complexity is the absence of consensus on a CUP definition. Due to the use of different definitions globally, it is difficult to compare this entity. 3 In the Netherlands, the cancer clinical practice guidelines advise to use the definition "CUP" if the patient has a metastasis of an unknown primary tumour origin, based on a cytological and/or histological proven metastasis of a cancer. 4 In 2018, CUP accounted for approximately 1300 incident cases in the country, this contributed to almost 2% of all cancers as registered by the Netherlands Cancer Registry (NCR). 5,6 CUP occurrence is equal in both sexes. On average, patients are aged 74 years at diagnosis. 5 The disease primarily concerns adenocarcinoma (ca. 60%) and undifferentiated carcinoma (ca. 20%), with the most common metastatic sites of presentation being the liver (ca. 40%) and lymph nodes (ca. 20%). 2,7 In the Netherlands, the overall median survival for patients with a CUP diagnosis between 2010 and 2012 was 1.7 months. 2 Despite the frequent occurrence and bleak prognosis of CUP, research into its aetiology is particularly scarce.
Potential CUP risk factors that have previously been identified include diabetes, family history of cancer, waist circumference and smoking. [8][9][10][11] Lifestyle recommendations for cancer prevention have described both alcohol consumption and tobacco smoking as modifiable cancerrisk factors. These recommendations advise against drinking alcohol and avoiding exposure to smoking. [12][13][14][15] To date, few studies have investigated the association between alcohol consumption, tobacco smoking and CUP risk. 10,11,16 Two prospective cohort studies investigated alcohol consumption and did not observe an association with CUP risk. 10,11 Three studies demonstrated a strong association between smoking and CUP risk. 10,11,16 None of these studies, however, assessed the association between smoking duration and CUP risk, and one did not investigate smoking frequency. 16 Therefore, we aimed to investigate the association between alcohol consumption, tobacco smoking and the development of CUP in greater depth. We hypothesise that (a) CUP risk is higher in participants with a high intake of alcoholic drinks, (b) CUP risk is higher in participants who smoke and (c) there is a multiplicative or additive interaction effect between alcohol consumption, tobacco smoking and CUP risk.

| Design and study population
The Netherlands Cohort Study (NLCS) was started on 17 September 1986 and includes 120 852 participants aged 55 to 69 years at baseline from 204 Dutch municipalities. Data processing and analyses were based on the nested case-cohort design. Cases were derived from the full cohort while the number of person-years at risk for the full cohort was estimated from a subcohort of 5000 participants who were randomly sampled from the full cohort at baseline. 17

| Outcome measure
In our study, CUP is defined as a metastasised epithelial malignancy with no identifiable primary tumour origin after cytological and/or histological verification during a patient's lifetime. This CUP definition only includes epithelial malignancies (ICD-O-3: M-8000-M-8570), which excludes, for example, sarcoma, lymphoma, mesothelioma and melanoma.

| Follow-up
Cancer follow-up was established through annual record linkage of the full cohort with the NCR and the Dutch Pathology Registry (PALGA) to identify CUP cases within the NLCS. 18 Information regarding the site of metastasis was obtained from the NCR, but data were only partially available and, therefore, information was requested and retrieved from PALGA pathology excerpts. These pathology excerpts were also used to determine whether cytological and/or histological confirmed cases had been correctly categorised in the data received from the NCR. according to number of metastases (multiple metastases of the same type were counted as one metastatic site, for example, bone metastases in hip and vertebra were counted as one); according to localisation

What's new?
Little is known about factors that raise the risk of cancer of unknown primary (CUP), in which metastases have no identifiable primary tumor origin. This prospective study examined possible associations between CUP and alcohol consumption and cigarette smoking, using sex-stratified analyses and assessing variables such as smoking frequency and duration.
Analyses show that higher alcohol consumption, current cigarette smoking, more cigarettes smoked, and longer smoking duration are associated with increased CUP risk. The data highlight the significance of lifestyle factors in CUP risk and suggest that lifestyle recommendations regarding alcohol intake and smoking are relevant to CUP prevention. of metastasis (up to four localisations); and according to survival duration (≤1 and >1 year after diagnosis). The subcohort consisted of 4774 participants after excluding members who reported a history of cancer (except for skin cancer) at baseline. Participants were also excluded when there were missing values on alcohol consumption or cigarette smoking. As a result, 963 CUP cases and 4288 subcohort members were available for investigation (see Figure 1).

| Questionnaire data
All cohort members completed a self-administered questionnaire, which included detailed questions on alcohol consumption, tobacco smoking and other cancer risk factors. Alcohol consumption was measured over the year preceding baseline and was addressed by questions on beer, red wine, white wine, sherry, other fortified wines, liqueurs and liquor. Frequency of alcohol consumption ranged from "never" to "6 to 7 times per week" and information on the number of glasses consumed per day. Participants who reported "never" or consumed "less than once per month" were considered abstainers. Four items from the questionnaire (red wine, white wine, sherry and liqueurs) were combined into one single wine variable since these items were substantially correlated, and separate analysis would have resulted in small numbers of subjects within each stratum. Mean daily alcohol consumption was calculated by using the computerised Dutch food composition table. 19 Based on pilot study data, standard glass sizes were defined as 200 mL for beer, 105 mL for wine, 80 mL for sherry and 45 mL for both liqueurs and liquor, corresponding to 8, 10, 11, 7 and 13 g of ethanol, respectively. The food frequency questionnaire was validated against a 9-day diet record. The Spearman correlation coefficient between alcohol consumption as assessed by the questionnaire and that estimated by the diet record was 0.89 for all subjects and 0.85 for alcohol consumers. 20,21 Tobacco smoking was addressed through questions on baseline smoking status and the ages at first exposure and last (if stopped) exposure to smoking. Questions were also asked about smoking frequency and smoking duration (excluding stopping periods) for cigarette, cigar and pipe smokers. As the vast majority of smoking subcohort members were cigarette smokers, analyses were restricted to that particular group. Based on the questionnaire data, the following cigarette smoking variables were constructed: cigarette smoking status (never, ex, current); frequency (cigarettes per day); duration (years); and time since smoking cessation (years). Participants who indicated that they had never smoked cigarettes were considered never smokers. Time since smoking cessation was calculated as age at baseline minus age at smoking cessation. To avoid collinearity problems, smoking frequency and smoking duration were centred as proposed by Leffondré et al. 22  Cox proportional hazards models were used to evaluate associations of alcohol consumption, cigarette smoking and CUP risk. Associations were estimated using age-and sex-adjusted, and multivariable adjusted hazard ratios (HRs) with 95% confidence intervals (CIs).

| Statistical analysis
Attributable risks were calculated for alcohol consumption and cigarette smoking. SEs were calculated using the robust Huber-White sandwich estimator to account for additional variance introduced by sampling from the full cohort. 23 The proportional hazards assumption was tested using the scaled Schoenfeld residuals, 24

| RESULTS
Data analysis was based on 963 CUP cases and 4288 subcohort members for whom the data on alcohol consumption and cigarette smoking were complete. Overall, CUP patients were on average aged 73 years at diagnosis, the majority of whom were men (62.6%) and most cases were histologically verified (71.3%) (see Table 1). The most common histological subtype was adenocarcinoma (64.8%). The majority of patients were registered with a single organ metastasis (80.3%), and the most frequent metastatic site of presentation was the liver (37.9%). Most patients had died within a year after CUP diagnosis (73.4%).
Overall, CUP cases were more often alcohol consumers with a substantially higher ethanol intake (≥30 g of ethanol) in comparison to subcohort members (16.6% vs 9.0%) (see Table 2). This higher intake was especially evident in male CUP patients of whom 23 In general, a higher ethanol intake was associated with an increased CUP risk (P trend = .02; see Table 3). Participants who reported consuming ≥30 g of ethanol per day were compared to abstinence, at the highest risk of developing CUP (multivariable adjusted HR: 1.57, 95% CI: 1.20-2.05). The attributable risk for alcohol consumption on CUP risk was 4% (95% CI: 2%-6%). No multiplicative interaction was observed between alcohol consumption categories, sex, and CUP risk (P interaction = .86).
Current smokers were at an increased risk of developing CUP (multivariable-adjusted HR: 1.59, 95% CI: 1.29-1.97) compared to never smokers (see Table 4). For cigarette smoking status, the attributable risk for CUP risk was 6% (95% CI: 4%-8%). After stratification for sex, we observed that CUP risk was the highest for current smokers compared to never smokers, in both men and women (HR: We observed that the more cigarettes a participant smoked, the higher the CUP risk (P trend = .003). This trend was also observed in men (P trend = .004). Overall, participants who smoked 10 to <20 or ≥20 cigarettes per day had a higher CUP risk (HR: 1.27, 95% CI: We observed no multiplicative interaction between alcohol consumption, cigarette smoking frequency and CUP risk (P interaction = .12) (see Table 5). However, we did find increased risks for most exposure combinations of alcohol consumption and cigarette smoking categories, for participants who smoked 10 to <20 or ≥20 cigarettes per day compared to abstainers and never smokers as the reference group. In  Table 6).
Results from the sensitivity analysis with restriction to histologically verified CUP cases, and results after excluding the first 2 years and the first 5 years of follow-up did not differ substantially from the findings of the complete multivariable analysis (data not shown).

| DISCUSSION
In this prospective cohort study, alcohol consumption and cigarette smoking were found to be associated with CUP risk. Associations were increased for participants who drank ≥30 g of ethanol per day.  10 The Australian study is a prospective cohort study that compared 327 incident cancer registrynotified CUP cases to two sets of controls that were randomly selected (3:1) using incidence density sampling with replacement.
They observed no associations between alcohol consumption and CUP risk compared to the metastatic cancer controls and compared to the general cohort population controls. 11 In the NLCS, we have found a positive association between alcohol consumption and CUP risk. The association was more pronounced in participants who drank ≥30 g of ethanol per day compared to abstainers. Additionally, our stratified analysis indicates that the CUP risk was especially increased However, the exposure category of no smoking was not described in-depth and possibly included ex-smokers. In the NLCS, we also found current cigarette smokers to be at a greater risk of developing CUP (HR: 1.59) compared to never smokers. Although this association between smoking and CUP is weaker compared to those findings in the abovementioned studies, it should be noted that our study used different categories for measuring cigarette smoking. However, in accordance with the European study, we observed an association between smoking and CUP risk, which was elevated in the highest category of smoking frequency. 10 In contrast, the Australian study observed no difference in risk associated with <20 or ≥20 cigarettes per day. 11 The European study also reported that participants who had quit smoking ≤10 years ago were at a higher risk of CUP (HR: 1.34) than participants who had never smoked. 10 In the NLCS, we found that participants who had stopped smoking <10 years were at a higher risk of developing CUP (HR: 1.26) compared to never smokers. Accordingly, our results are similar to those of the European study, which means our results are again in line with those of the European study.
Our study provides novel information on the association between cigarette smoking frequency, cigarette smoking duration and CUP risk. We found CUP risk to be more pronounced in the highest exposure categories of both cigarette smoking frequency and cigarette smoking duration, suggesting that the more the cigarettes (N/d) or the longer (years) the participants smoked, the greater their risk of developing CUP.
We found no multiplicative interaction effect between alcohol consumption, cigarette smoking frequency and CUP risk. However, we did find that participants who consumed the highest intake level of alcohol and smoked the highest number of cigarettes had a greater risk of CUP than either abstainers or never smokers. In addition, we found a significant additive interaction between the highest exposure categories of alcohol consumption and cigarette smoking frequency and CUP risk. This means that the combined effect of alcohol consumption and cigarette smoking frequency is larger than the sum of the individual effects of both alcohol consumption and cigarette smoking frequency. 25 It should however be acknowledged that for assessing the interaction on additive scale, exposure categories were generated in a dichotomous manner.

| Strengths and limitations
An important strength of our study is its prospective cohort design. A further strength is that the NLCS consists of a large cohort of 120 852 participants who were followed up for cancer incidence by the cancer registry in the Netherlands. Cases were registered by trained registry clerks who had access to the medical files and entered data by applying uniform coding rules. Furthermore, we were able to analyse 963 incident CUP cases, which is a much higher number of cases than previous studies have used to investigate CUP aetiology. It should, however, be acknowledged that the CUP definition used here might differ from that used in other countries because the criteria for defining "CUP" are heterogeneous. CUP cases within our study were consistently registered by NCR registry clerks, for which data were retrieved from pathology and clinical reports. 26 Within the NLCS, information on alcohol consumption and exposure to smoking were collected before the outcome, minimising the effect of information bias. A potential limitation of the current study is that data on all exposure variables are self-reported, which may have resulted in bias due to misclassification. However, we expect this misclassification to be nondifferential. Another potential limitation is that the questions regarding smoking behaviour are not validated. Even so, the questions included detailed categories that the participant could answer. Unfortunately, we do not have data to check which diagnostic methods were used for our CUP patients. Nevertheless, if we restrict our analysis purely to histologically verified CUP cases, for whom extended diagnostic methods are more likely, we find that the results do not differ greatly from the complete multivariable analyses. Accordingly, the findings from the complete multivariable analyses are representative of CUP cases with or without an extensive diagnostic work-up. Note: Measure of interaction on additive scale: RERI = 1.14 (95% CI: 0.33-1.96); P = .006. Abbreviations: CI, confidence interval; RERI, relative excess risk due to interaction.

| CONCLUSIONS
In our study, alcohol consumption and cigarette smoking were found to be associated with an increased CUP risk. These findings suggest that lifestyle recommendations on cancer prevention regarding not drinking alcohol and avoiding exposure to smoking are also valid for CUP.