Familial aggregation of urothelial cell carcinoma
Urothelial cell carcinoma (UCC) is not considered to be a familial disease. Familial clustering of UCC was described in several case reports, however, some with an extremely early age at onset suggesting a genetic component. Epidemiological studies yielded inconsistent evidence of familial UCC, possibly because of low power and the inability to adjust for strong confounding. In our study the existence of a familial subtype of UCC was evaluated, as well as familial clustering of UCC with other types of cancer. A population-based family case-control study was performed including patients newly diagnosed with UCC of the bladder, ureter, renal pelvis or urethra, between January 1995 and December 1997, in the southeastern part of the Netherlands. Information on the patients' first-degree relatives was collected by postal questionnaire and subsequent telephone calls. The patients' partners filled out a similar questionnaire on their relatives. All reported occurrences of UCC were verified using medical records. Disease occurrence among case-relatives and control-relatives was compared to obtain the familial risk. Random effect proportional hazards regression analyses were used to calculate this familial risk while adjusting for age, gender and smoking behavior. In 95 families of the 1,193 patients and in 36 families of the 853 partners at least 1 relative was diagnosed with UCC. This yielded an adjusted hazard ratio (HR) of 1.8 (95% CI: 1.3–2.7). An increased risk was also found for cancer of the hematolymphopoietic system (hazard ration = 1.9, 95% CI: 1.2–3.1) among case-relatives. These results indicate that UCC has a familial component with an almost 2-fold increased risk among first-degree relatives of patients with UCC, which cannot be explained by smoking. Future segregation analyses may indicate whether this clustering can be attributed to genetic susceptibility. © 2001 Wiley-Liss, Inc.
Bladder cancer is the fourth most common malignancy among Western men and it accounts for 5–10% of all malignancies.1 More than 95% of all bladder cancers are urothelial cell carcinomas. The majority of urothelial cell carcinomas are bladder tumors, whereas upper urinary tract tumors and tumors of the urethra contribute less than 10%.2 Numerous epidemiological studies have shown that smoking and some occupational exposures (e.g., aromatic amines, polycyclic aromatic hydrocarbons) are important environmental risk factors for the development of UCC.3 Most UCC patients do not have a family history of UCC. Several case reports, however, have been published on families in which several relatives developed UCC, often at an extremely early age.4 Some case-control studies have been performed suggesting a positive family history of UCC to be a risk factor.4 None of these studies, however, collected information on family size, age, gender and smoking behavior of the relatives and consequently could not adjust for these factors. A study from Utah was able to adjust for demographic characteristics of the relatives, although not for smoking.5 In this study first-degree relatives of bladder cancer probands were found to have an increased risk of bladder cancer of 1.5 (95% CI: 1.0–2.2). When only young probands (age <60 years) were considered, the familial risk was 5.1 (95% CI: 1.0–12.5). Only 2 epidemiological studies have specifically addressed the issue of familial clustering of UCC. In a study from the state of New York, Kramer et al.6 reported an almost 2-fold increase in the risk of bladder cancer among first-degree relatives of UCC patients compared to relatives of controls (hazard ratio [HR] = 1.9, 90% CI: 0.9–4.1). A study from Iceland reported only a slightly increased risk of UCC among relatives of UCC cases (observed-to-expected ratio = 1.2, 95% CI 0.9–1.2).7
In 1996, the only constitutional chromosomal aberration ever demonstrated in a UCC family was reported by Schoenberg et al.8 They performed a small study on young bladder cancer patients (under 40 years of age). One male patient (27 years old) had a mother who died from bladder cancer at the age of 65 years. Triggered by a history of several miscarriages in his wife and his mother, karyotype analysis was performed. A constitutional, balanced translocation t(5;20) (p15;q11) was found. Detailed study of the breakpoints of these chromosomes has not yet revealed a bladder cancer susceptibility gene.
Previous epidemiological studies have shown limitations and inconsistent evidence of familial UCC. It is important to evaluate the existence of an inherited subtype of UCC, because inherited and non-inherited cancers may share genetic lesions, which are more easily identified in inherited cancers. Therefore, we started a large-scale population-based family case-control study9 on familial UCC in 1996.
MATERIAL AND METHODS
A family case-control design was used in our study.9 The primary outcome in this design is disease occurrence among relatives, known as the familial risk. Familial aggregation of the disease exists if the risk of disease among case-relatives is higher than that among control relatives. An important difference exists between the conventional case-control design and this family case-control design. In the former a comparison is made between cases' characteristics and controls' characteristics, e.g., family history. In the latter design, however, the comparison takes place between the relatives of cases and the relatives of controls. The statistical comparison of these 2 cohorts of relatives can take place with cohort-type analysis while taking the familial dependency of relatives into account.
Subjects and data collection
Case-families were ascertained through UCC patients (probands) newly diagnosed between January 1, 1995 and December 31, 1997, identified from 2 population-based cancer registries held by the Comprehensive Cancer Center East and South in the Netherlands. A total of 2,051 probands with histologically confirmed UCC of the bladder, renal pelvis, ureter and urethra were identified (topography codes 188.1-9 and 189.1-3,10 morphology codes 8120 and 813011). Inclusion criteria were: diagnosed under the age of 75 years, Caucasian, living in the catchment area of the Comprehensive Cancer Center East and Comprehensive Cancer Center South on date of diagnosis and they had to be capable of reading and understanding the Dutch language. Information on demographic factors, smoking behavior, occupation, history of cancer, history of urinary tract diseases, history of chronic diseases and drug history was collected by postal questionnaire. In the same questionnaire similar information was requested on first-degree relatives (demographic factors, smoking behavior, longest held occupation, history of cancer and history of urinary tract diseases).
Control families consisted of the family of the probands' partner. Information was provided by the probands' partners who were invited to fill out a similar questionnaire. If the proband was deceased or unable to provide information, the partner was requested to fill out both questionnaires. If both the proband and the partner were unable to provide information, the first-degree relatives were contacted.
Before any of the subjects were contacted, the treating urologist was asked for permission for his patients to participate in the study. In 47 cases, this permission was denied. The remaining 2,004 probands and their partners received an invitation letter, additional information and a reply card. They were asked to fill out and return the reply card indicating whether or not they were willing to participate. Eighty-three probands were excluded after sending the invitations because they appeared to be unable to read and understand the Dutch language, or they were deceased and no partner or family was available, or because the probands' addresses were unknown. When no reply was received within 3 weeks, a reminder was sent. A questionnaire was sent to all subjects who were willing to participate. All other subjects (non-responders or subjects who were not willing to participate), were contacted by telephone to ask the reason for not responding or not participating. Another reminder was sent when the questionnaires were not returned within 3 weeks. Subjects who did not return the questionnaires within approximately 6 weeks were contacted by telephone. A random selection of non-responders (both non-responders to the reply card and to the questionnaire) were contacted by telephone to obtain limited information on the probands and their families. When a returned questionnaire was incomplete, subsequent telephone calls were made for additional information.
All data on cancer of the urinary tract among relatives of the study participants were verified with medical records, after obtaining written consent from these patients themselves or, in the case of death, from their partners or next-of-kin. Verification was possible in 60% of all UCC affected relatives (56% of all affected case-relatives and 63% of all affected control-relatives). Except for only one relative, the questionnaire data appeared to be correct. Verification was not possible in all cases because the diagnosis had been made too long ago (more than 10–15 years) and the medical records were no longer available.
To assess the quality of the data concerning the occurrence of other types of cancer among relatives, verification was performed by linking questionnaire data to the population-based cancer registry. Only cancer occurrence among relatives who had been living in the catchment area of the Comprehensive Cancer Centre South (CCC South) since 1975 was checked by matching the data to the cancer registry to quantify under- and overreporting of cancer. Because the Comprehensive Cancer Center East is only complete since 1989 whereas the CCC South is considered to be complete since 1975, only the CCC South was used for this verification. A total of 339 relatives (case- and control-relatives), were reported to have had at least 1 malignancy diagnosed since 1975. Data linkage was successful in 301 of these relatives. In 77% of the tumors, the site recorded by the CCC South and the site mentioned in the questionnaire were identical. For the 38 relatives who could not be linked, additional information was obtained (through the proband or family); 15 relatives had been diagnosed/treated at hospitals outside the catchment area of the CCC South. Definitive positive linkage was therefore achieved in 93%. The remaining 23 relatives could not be linked for various reasons, such as inaccurate date of birth on the questionnaire, patient death within a few days after clinical diagnosis (no pathology report), or diagnosis made at a nursing home (also no pathology report).
Verification of negative family history was also performed. The probands and partners reported that 3,472 relatives (1,940 case-relatives and 1,532 control-relatives) who were born before 1955 and were living in the catchment area of the CCC South did not have any malignancies. The cut-off point of 1955 was chosen arbitrarily to include relatives who were old enough to have developed cancer. These relatives were also linked to the cancer registry, which yielded a linkage percentage of 2.6% (case-relatives 2.8% and control relatives 2.5%). These 2.6% included different cancer sites (tongue, colon, rectum, pancreas, nasal cavity, lung, stomach, bone, skin, prostate, mamma, cervix, ovarium, kidney, bladder, leukemia, bone marrow, primary site unknown). It is possible that some of these patients had not been informed about the fact they had cancer, or that the relative had not mentioned his/her diagnosis to the family.
To evaluate the existence of a familial subtype of UCC, the risk of UCC among case-relatives was compared to the risk of UCC among control-relatives by using a Cox proportional hazards model to which a random effect for family relatedness was added. The random effect is defined as a categorical variable that describes excess risk, or frailty for a family. The frailty is usually viewed as an unobserved covariate. In this way the inherent dependency in family data is taken into account. All presented HR (crude and adjusted) are adjusted for this family dependency.
Only parents and siblings were included in the analyses because the children are shared by cases and controls and are therefore non-informative.
The HR were adjusted for age, gender and smoking; 95% CI were calculated. The dependent variable was defined as age at diagnosis of UCC, age at time of interview, or age at time of death, whichever came first. The independent variable was defined as relative of a case or relative of a control. Smoking behavior and gender were included as covariables. Smoking was taken into account as a dichotomous variable (ever smoked vs. never smoked).
Subsequent analyses were carried out after stratification by tumor site in the probands (lower vs. upper urinary tract), after excluding case-relatives of probands with a superficial, papillary (pTa) tumor and after excluding case-relatives of probands older than 60 years. To detect possible genetic heterogeneity, clustering of UCC with other types of cancer was also studied. Analyses were performed in SAS version 6.12 for Windows12 and S-plus version 2000.13
The study was approved by the Institutional Review Boards of the University Medical Centre Nijmegen and the Comprehensive Cancer Center East and Comprehensive Cancer Center South.
The response to the invitation was 78% (1,507 UCC patients). In total 1,176 partners were also willing to participate in the study (the total number of partners is not known because information from non-responding patients is unavailable). Subsequently, 1,193 patients (79%) and 853 partners (73%) returned a questionnaire with sufficient information for the analyses. For only 1.3% of the patients, the information in the questionnaire was given by the partner or next-of-kin of the proband. Reasons for not returning or not completing the questionnaire were: “no contact with relatives,” “too much effort,” “too emotional,” “too complicated to retrieve all data on relatives” and “no permission from relatives to give information.” There were only slight differences between responders and non-responders concerning age at diagnosis, gender, tumor site, tumor morphology, stage and grade of the tumor (Table I). Also, the non-responders did not differ from the responders with respect to their family history of UCC.
Table I. Characteristics of Responders and Non-Responders1
|Number of patients||728||1193|
|Mean age at diagnosis UCC2 (years) (SD)3 (Range)||63.8 (9.1) (28–74)||61.8 (9.5) (16–74)|
| Upper urinary tract||6.2||8.5|
| Invasive (pT1 included)||49.0||43.5|
|Grade of differentiation (%)|
| Good (grade I)||24.7||27.3|
| Moderate (grade II)||32.6||35.0|
| Poor (grade III)||31.0||27.8|
| Undifferentiated (grade IV)||0.8||0.3|
| Grade not determined||10.9||9.6|
|Positive family history of UCC4|
Familial clustering of UCC
Information was obtained on 8,014 first-degree case-relatives and 5,673 first-degree control-relatives. Distributions of the type of relative, gender, smoking behavior and age were similar in the two groups (Table II).
Table II. Characteristics of Case-Relatives and Control Relatives
|Type of relative (%)|
| Father||1193 (14.9)||853 (15.0)|
| Mother||1193 (14.9)||853 (15.0)|
| Brother||2774 (34.6)||2032 (35.8)|
| Sister||2854 (35.8)||1935 (34.1)|
| Male||3967 (49.5)||2885 (50.9)|
| Female||4047 (50.5)||2788 (49.1)|
|Ever smoked (%)|
| Yes||4156 (51.9)||3063 (54.0)|
| No||3104 (38.7)||2193 (38.7)|
| Unknown||754 (9.4)||417 (7.4)|
|Mean age (SD1)||63.6 (17.4)||63.1 (16.9)|
|Number of UCC2 affecteds||101||38|
| Mean age at diagnosis (SD)1 (range)||64 (10.5) (39–87)||66 (12.6) (19–84)|
Among the case-relatives, 101 individuals were diagnosed with cancer of the bladder (n = 97), ureter (n = 3) and renal pelvis (n = 1), compared to 38 individuals among the control-relatives (bladder n = 36, ureter n = 1 and urethra n = 1). In 6 case-families and 2 control-families, 2 affected first-degree relatives were found. Overall, 8% of the patients had a positive family history of UCC compared to 4% of the controls. The mean age at diagnosis of patients with a positive family history was similar to that of patients with a negative family history (62 years). The mean age at diagnosis of UCC among affected case-relatives was only slightly lower than that of affected control-relatives (64 vs. 66 years). The majority of affected relatives had smoked, but a larger proportion of the affected control-relatives had smoked compared to the case-relatives (90% vs. 73%).
The cumulative risk of UCC among case-relatives was 3.8% compared to 2.1% among control-relatives. The crude HR of UCC for case-relatives compared to control-relatives was found to be 1.8 (95% CI: 1.2–2.6). The proportional hazards model yielded an age, gender and smoking adjusted HR of 1.8 (95% CI: 1.3–2.7) (Table III).
Table III. Familial Clustering of UCC1
|All relatives||1.8 (1.3–2.7)|
| Parents||2.2 (1.3–3.6)|
| Siblings||1.6 (0.9–2.7)|
| Females||3.7 (1.3–10.6)|
| Males||1.6 (1.1–2.5)|
| Non-smokers||4.2 (1.4–12.7)|
| Smokers||1.6 (1.0–2.4)|
|Relatives of probands with bladder UCC||1.9 (1.3–2.8)|
|Relatives of probands with upper urinary tract UCC||1.8 (0.7–4.3)|
|Relatives of probands with a ≥ pT1 tumor||2.0 (1.3–3.1)|
|Relatives of probands ≤ 60 years||2.5 (2.0–4.0)|
This risk appeared to be higher among women and among non-smokers. When only parents were included in the analyses the HR increased, whereas it decreased when only siblings were included. After stratification by tumor site in the probands (upper vs. lower urinary tract), the adjusted HR was 1.8 (95% CI: 0.7–4.3) among relatives of probands with upper urinary tract UCC and 1.9 (95% CI: 1.3–2.8) among relatives of probands with bladder UCC.
When all the relatives of probands with a pTa tumor were excluded from the analyses, the HR increased slightly. The same was found when the relatives of probands older than 60 years were excluded from the analyses.
Familial clustering of UCC with other types of cancer
All the tumors among relatives were classified into 14 different sites. The majority of these cancers did not appear to cluster with UCC. An increased risk was found for tumors of the hematolymphopoietic system among case-relatives compared to control-relatives whereas a decreased risk was found for unknown and other tumors (Table IV). Familial clustering of UCC with other tumors was also evaluated when only relatives of probands with upper urinary tract UCC were analyzed and when only relatives of probands with a positive family history of UCC were analyzed (Table IV).
Table IV. Familial Clustering of Urothelial Cell Carcinoma With Other Types of Cancer. Case-Relatives vs Control-Relatives
|Mouth and pharynx||1.0 (0.6–1.8)||0.8 (0.2–3.2)||1.4 (0.5–4.1)|
|Digestive tract||1.0 (0.9–1.2)||1.1 (0.8–1.7)||1.0 (0.7–1.5)|
|Respiratory tract||1.1 (0.9–1.4)||0.5 (0.3–1.0)||1.0 (0.6–1.6)|
|Bone and soft tissue||1.0 (0.7–1.7)||0.8 (0.2–2.7)||1.0 (0.4–3.0)|
|Skin3||0.7 (0.4–1.0)||0.6 (0.2–1.9)||0.8 (0.3–2.3)|
|Male genital organs4||—||—||—|
|Prostate||1.1 (0.8–1.7)||0.7 (0.3–2.0)||1.0 (0.4–2.4)|
|Breast||1.2 (0.9–1.6)||1.5 (0.9–2.4)||1.5 (0.9–2.4)|
|Breast (females)||1.2 (0.9–1.5)||1.3 (0.8–2.2)||1.3 (0.8–2.3)|
|Female genital organs||1.0 (0.7–1.5)||1.4 (0.7–2.8)||1.8 (1.0–3.4)|
|Urinary tract (other than bladder, ureter, renal pelvis)||1.2 (0.6–2.3)||1.7 (0.5–6.0)||1.5 (0.4–5.4)|
|Eye and CNS||1.2 (0.7–1.9)||0.7 (0.2–2.9)||0.9 (0.3–3.1)|
|Endocrine glands||1.6 (0.4–6.1)||—4||—4|
|Haematolymphopoietic system||1.9 (1.2–3.1)||2.2 (0.9–5.4)||1.7 (0.6–4.7)|
|Unknown and other||0.5 (0.3–0.8)||0.3 (0.1–1.3)||0.6 (0.2–1.7)|
|All non-UCC1||1.0 (0.9–1.1)||1.0 (0.8–1.3)||1.2 (0.9–1.5)|
Among relatives of probands with UCC of the upper urinary tract, the risk of tumors of the respiratory tract appeared to be decreased, whereas the risk of tumors of the hematolymphopoietic system appeared to be increased (similar to the results among all case-relatives). When relatives of probands with a positive family history of UCC were analyzed an increased risk was found for tumors of the female genital organs.
Familial clustering of UCC
Our study indicates that UCC has a familial component, which results in a 1.8-fold increase in the risk for first-degree relatives of patients with UCC. This familial clustering could not be attributed to age, gender or smoking. The highest risk was found among women and non-smokers. Among parents, the risk was higher than that in siblings. This familial clustering cannot be explained by any of the known cancer family syndromes. Therefore, familial UCC seems to be a separate entity. Clustering of UCC appears to be present in both pTa and ≥pT1 tumors. Furthermore, when pathology reviews were performed on a selection of the probands with a positive family history (n = 35), none of the UCC tumors showed histological characteristics different to UCC tumors of unselected patients.
Familial clustering of UCC with other types of cancer
Tumors of the hematolymphopoietic system seem to cluster with UCC. A similar observation was made in a study by Goldgar et al.,5 who reported a 3.7-fold increase (95% CI: 1.1–7.4) in the risk for lymphocytic leukemia in relatives of early bladder cancer cases. Nevertheless, this finding may be the result of chance because of the large number of comparisons. The opposite was found for tumors of unknown or other origin. In the literature, several case reports14–17 and some epidemiological studies18–20 describe clustering of UCC with other tumors such as retinoblastoma and colorectal cancer. The higher risk of bladder cancer among relatives of retinoblastoma patients appeared to be confined to known carriers of the mutated retinoblastoma gene.18 Furthermore, the risk of upper urinary tract tumors is increased in families with the hereditary non-polyposis colorectal carcinoma syndrome (HNPCC)19 whereas in families with the Muir-Torre syndrome the risk of bladder and upper urinary tract cancer appears to be increased.20 Our findings did not confirm familial clustering of UCC with these types of tumors. When relatives of probands with UCC of the upper urinary tract were compared to all control-relatives, no familial clustering was found with tumors of the digestive tract.
Our study included all incident cases with UCC registered in 1995–98 by the population-based cancer registry, suggesting that selection bias is not a problem in our study. The response, however, was not 100%. Fortunately, a non-response study showed that the percentage of probands with a positive family history was similar in responders and non-responders.
Another problem may be residual confounding by smoking. Although adjustment was made for smoking in the analyses, inaccuracies concerning the reporting of smoking among relatives may have resulted in residual confounding. Other studies, however, reported high concordance between the smoking status reported by the proband and the smoking status reported by the relative(s) themselves.21, 22 We did not collect detailed data on the amount of smoking by relatives. We do not believe this to be a major drawback of the study however, because it is not expected that the amount of smoking clusters within families.
Furthermore, in theory, evidence of familial aggregation or clustering may be due to a shared environment (instead of genetic factors). Because smoking is the most important risk factor for UCC, it was taken into account. Of course, some other risk factors have been described for UCC (occupational exposures, quality of drinking water). Information on occupational history was collected in our study and it was checked whether affected relatives had high risk-occupations fairly frequently but this was not the case. For reason of simplicity, this was not elaborated in the analyses. Other factors such as drinking water quality (arsene, chlorination) do cluster in families but the variation between families is practically negligible in the Netherlands. Such factors can therefore not confound the analyses. Khoury et al.23 showed that it is very unlikely that even strong environmental risk factors cause a 2-fold risk of disease for relatives of patients.
Finally, because the occurrence of cancer among relatives was reported by the patient or partner, inaccuracies may lead to information bias. As described in the Material and Methods section, however, the presence or absence of malignancies among relatives was verified and it appeared that the data were quite valid.
Overall, our findings confirm the existence of a familial subtype of UCC that appears to be site-specific and not part of any known hereditary cancer syndrome. The familial clustering observed in our study may be the result of variation in low-penetrance genetic susceptibility factors, such as N-acetyltransferase or glutathion-S-transferase genotype or high-penetrance Mendelian inheritance. The next step will be to define the nature of this familial aggregation and to assess whether Mendelian inheritance can explain the familial clustering of UCC.
We would like to thank Dr. A.P.M. van der Meyden, L.M.H. Schreinemachers, Dr. J.W. Hoekstra (Bosch Medicentrum, Den Bosch), Dr. J.J. Bade, A.J. Smans (St. Anna Ziekenhuis, Oss), Prof. Dr. F.M.J. Debruyne, Dr. G.O.N. Oosterhof, Dr. J.A. Witjes (Academisch Medisch Centrum St. Radboud, Nijmegen), Dr. H.F.M. Karthaus, R.L.F.M. Corten, Dr. H. Vergunst (Canisius-Wilhelmina Ziekenhuis, Nijmegen), Dr. E.J. Barten, Dr. J.W. Hoefakker, Dr. G.A.H.J. Smits, J.G. Idema, Dr. P.C. Weijerman, E.A. Rodrigues Pereira, Dr. R.P. Heybroek (Ziekenhuis Rijnstate, Arnhem/Het ziekenhuis Velp, Velp), C.A.H.M. Kerckhoffs, V.W.J. de Kort (Carolus-Liduina Ziekenhuis, Den Bosch/Boxtel), H. Jansen, P.J. Posthumus (Interconfessioneel Ziekenhuis De Baronie, Breda), Dr. G.A. Dijkman, P.J. van den Broeke, E.H.G.M. Oomens (Ignatius Ziekenhuis, Breda/Pasteur Ziekenhuis, Oosterhout), Dr. A.D.H. Geboers, T.I. Tan (Slingeland Ziekenhuis, Doetinchem), A.C. Viddeleer, Jhr. O.B. van Vierssen Trip (Ziekenhuis Gelderse vallei, Ede), C.L.A.H. Bruijnen, Dr. A.A.J.M. Daenekindt, Dr. A.J.M. Hendrikx, F.H.A. Marechal, H.J.E.J. Vrijhof (Catharina Ziekenhuis, Eindhoven), F.P.P.M. Pernet, R.B. de la Fuente (Diaconessenhuis, Eindhoven), E.W. Stapper, C. van der Sluis (Elkerliek Ziekenhuis, Helmond), N.A.V. Fievez, T.T.H. Jansen (St. Annaziekenhuis, Geldrop), E.P. ten Donkelaar, (Ziekenhuis St. Jansdal, Harderwijk), J.J.M. Schroeder, Dr. P.J.M. Kil, Dr. G.J. Montagne (St. Elisabeth Ziekenhuis, Tilburg/Tweesteden Ziekenhuis, Waalwijk), J.G. Lardenoye, R.J.A.M. Davits (Tweesteden Ziekenhuis, Tilburg), J.P.J. Mackenbach, A.Q.H.J. Niemer (St. Joseph Ziekenhuis, Veghel), P.F.C.M. Braam, Dr. W.J. Levens, Mr. W. Zwiers (St. Joseph Ziekenhuis, Veldhoven), Dr. R.P.E. Pauwels, J.C. Scheepers, A.H.P. Meier (St. Maartens Gasthuis, Venlo), all the participants and the Comprehensive Cancer Centers East (Dr. J.A.A.M. van Dijck, H.A. Ament, M. de Kok) and South (Dr. J.W.W. Coebergh, L.H. van der Heijden, E. Masseling). Furthermore, we thank Mrs. A. Pellegrino for collecting the data.