If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Individuals who are younger, have a high socioeconomic background and/or have a healthy lifestyle are more inclined to participate in screening trials. This form of bias may affect the generalizability of study results to the target population. This study aimed to investigate the generalizability of the NELSON lung cancer screening trial to the Dutch population.
People at high risk for developing lung cancer were identified by sending a health questionnaire to 606,409 persons aged 50–74 years, based on population registries. Eligible subjects received an invitation to participate (n = 30,051). 15,822 subjects agreed to participate and were randomized, whereas 15,137 did not respond (so-called eligible nonresponders). Baseline characteristics and mortality profiles were compared between control group participants and eligible nonresponders.
Participants had better self-reported health (p = 0.02), were younger, more physically active, higher educated, and more often former smokers compared with eligible nonresponders (all p < 0.001). No differences were seen in self-reported outcomes of pulmonary tests, history of lung cancer, and smoked pack-years. Mortality due to all-causes (p < 0.001) and mortality classification separately was lower among participants. However, the proportion of subjects death due to cancer was higher among participants (62.4% vs. 54.9%).
Modest differences in baseline characteristics between participants and eligible nonresponders, led to minor differences in mortality profiles. However, group sizes were large and therefore it seems unlikely that these small differences will influence the generalizability of the NELSON trial. Results of the NELSON trial can roughly be used to predict the effect of population-based lung cancer screening.
In the United States, this finding has led the United States Prevention Service Task Force to recommend lung cancer screening for current and former smokers, if quit within the past 15 years, aged 55 through 80 years with a smoking history of at least 30 pack-years.
However, many issues remain regarding the technical and logistical aspects of screening, cost–effectiveness and generalizability. In Europe, no lung cancer screening trial has yet demonstrated a significant reduction in lung cancer mortality.
However, the largest European trial, the Dutch-Belgian lung cancer screening trial (NELSON), is still ongoing. The NELSON trial investigates whether screening using low-dose CT (LDCT) can reduce lung cancer mortality by at least 25% at 10 years of follow-up
. Major differences between the NELSON trial and the National Lung Screening Trial are that NELSON (1) offers no screening to control group participants, (2) has different intervals between screening rounds, and (3) uses different management protocols for nodules and abnormalities.
In interpreting the results of screening studies, it is important to know whether study participants were representative of the target population, as volunteers who are healthier and more concerned about their own health are more willing to participate in screening programs.
diseases, than nonparticipants. One pilot study of lung cancer screening even showed that participants had a lower mortality rate for all types of cancer besides lung cancer, cardiovascular diseases, and noncancerous diseases other than cardiovascular and respiratory diseases compared with nonparticipants.
However, lung cancer mortality was higher among participants. This mortality difference might be explained by selection bias; attendees of screening programs may have more awareness of being at risk of developing lung cancer, which may increase their interest in screening.
So far, previous research showed that the NELSON study population is younger, has a better general health, has a higher proportion of current heavy smokers and is slightly lower educated compared with the general Dutch population.
However, less is known about potential differences in physical activity, alcohol consumption, smoking-related symptoms, the effect on the mortality profile of participants and eligible nonparticipants (the so-called eligible nonresponders; Fig. 1).
The aim of this study was to investigate whether differences in characteristics and mortality profiles of participants of the NELSON study, and eligible nonresponders exist. The results of this study are relevant for the interpretation of the forthcoming mortality analyses of the NELSON trial.
In the NELSON trial, 15,822 high-risk volunteers were randomized (1:1) to screening (n = 7915) using LDCT at respectively baseline and 1, 3, and 5.5 years after baseline, or to no screening (n = 7909).
The NELSON trial was approved by the Dutch Minister of Health after positive advice from the Dutch health Council and by the Ethics Boards of the participating centers.
During the recruitment phase, which occurred in two waves (during the second half of 2003 and the second half of 2005), addresses of subjects aged between 50 and 74 years were obtained from the population registries of seven districts in the Netherlands and 14 municipalities around Leuven in Belgium.
These subjects received a questionnaire about their general health, medical check-ups and history, physical activity, body weight and length, smoking history, alcohol consumption, family history of cancer, education and their opinion on screening programs in general.
General health was determined by the subjects “ability to climb two flights of stairs” (yes, no, don't know) and how they would describe their health: excellent, very good, good, moderate, or severe. Questions regarding smoking-related symptoms of lung disease were: did you have symptoms of coughing/sputum/wheezing/dyspnea for at least 3 months this year? (yes, no).
Questions on medical history and check-ups were as follows: was one of the following diagnostic procedures performed last year, 1–5 year, or greater than or equal to 5 years ago: (1) chest x-ray, pulmonary function test, CT-scan of the chest or sputum test?, (2) did you undergo lung surgery (e.g., pneumonectomy or lobectomy)?, (3) were you diagnosed with cancer and if so, when (less than 5 years ago, greater than or equal to 5 years ago, or greater than or equal to 5 years ago and still under treatment)? and (4) what type of cancer were you diagnosed with (lung cancer, breast cancer, kidney cancer, melanoma, or other type)? Furthermore, physical activity was assessed as follows: how many times a week are you physically active for greater than or equal to 30 minutes (daily, 5–6 times, 2–4 times, 1 time, or less than 1 time a week). Alcohol consumption was assessed by asking how much alcohol was consumed at once (in pints) and at which base: daily, 5–6 times a week, 3–4 times a week, 1–2 days a week, 1–3 days a month, less than 1 glass a month or never. Willingness to participate in screening programs was assessed for prostate cancer, colon cancer, diabetes, cholesterol and cardiovascular diseases (yes, no, do not know) and their opinion on an acceptable number of persons to screen to detect one case of lung cancer at early stage (10, 100, 1000, 10,000, 100,000, or 1,000,000). The highest completed level of education was determined through a single question with seven options: primary education, lower technical or vocational education, general secondary education, secondary technical or vocational education, senior general secondary education or pre-university education, higher technical or vocational education and university. The questionnaire also assessed smoking in detail.
The eligibility criteria were as follows: age 50–75 years, smoking history of greater than or equal to 15 cigarettes per day for greater than or equal to 25 years or greater than or equal to 10 cigarettes for greater than or equal to 30 years, and were still smoking or had quit less than or equal to 10 years ago. Exclusion criteria were: a moderate or bad self-reported health and inability to climb two flight of stairs, a body weight greater than or equal to 140 kg, a history of renal, melanoma or breast cancer, lung cancer diagnosed less than 5 years ago or greater than 5 years ago but still receiving treatment, or a chest CT examination within the past year.
In addition, the second questionnaire assessed smoking habits and exposure to asbestos in more detail.
Eligible responders who provided informed consent and completed the second questionnaire (n = 15,822, response rate of 51.1%) were randomized (1:1) to either the screen group or the control group.
Inclusion in this substudy
For this substudy, subjects randomized to the control group (n = 7453) were compared with eligible subjects who did not participate (n = 13,661). Subjects randomized to the screen group were excluded because of the potential effect of screening on their mortality profiles and the embargo on mortality outcomes of this group. Furthermore, this substudy was limited to Dutch subjects, as only Dutch mortality data was available at the time of analyses.
Anonymised mortality data for both groups were obtained via Statistics Netherlands. January 2013 was chosen as end date of this substudy, at which point 99.1% of the subjects were traceable. To obtain mortality data, this study population was matched using four variables: sex, date of birth, zip code, and date of obtaining addresses. This led to an accuracy of almost 98% in matching.
Person-years were calculated as the time between obtaining the addresses of the subject and subject's date of death or the end date of this study, whichever came first.
To analyze mortality profiles, we classified the causes of death by disease groups, using the International Classification of Diseases, 10th edition: all-causes, all cancer causes, cardiovascular diseases, respiratory diseases, and noncancer diseases other than cardiovascular or respiratory diseases.
Baseline characteristics of control group subjects and eligible nonresponders were retrieved from the first questionnaire. Differences in baseline characteristics were assessed using the following tests: for continuous variables, normality was tested using the Kolmogorov–Smirov test and differences between the two groups were assessed by using the Mann–Whitney U test, as appropriate. For nominal variables, the χ2 test was used and the Mann–Whitney U test was used for categorical variables.
Classified mortality data were compared between the two groups by using the χ2 test. For all analyses, p values less than 0.05 were considered statistically significant. SPSS version 21 and STATA 13 special edition were used to perform the analyses.
A total of 7453 Dutch control group participants were compared with 13,661 Dutch eligible nonresponders (Table 1). Participants were younger (p < 0.001), more often male (p < 0.001), had better self-reported general health (p = 0.02), higher level of physical exercise (p < 0.001) and a higher level of education (p < 0.001). Participants also consumed more alcohol (p < 0.001) and consisted of higher proportion of former smokers (p < 0.001). However, most differences in proportions were small.
TABLE 1Baseline Characteristics of NELSON Control Group Participants and Eligible Nonresponders
Physical activity: high was defined as greater than or equal to 5 times active for greater than or equal to 30 minutes a week, moderate was defined as greater than or equal to 1 but less than 5 times active for greater than or equal to 30 minutes a week and low was defined as less than 1 time active for greater than or equal to 30 minutes a week.
Education level: lowest: only elementary; low education: Lower technical or vocational education and general secondary education; medium education level: secondary technical or vocational education and senior general secondary education; high education level: higher technical or vocational education and university.
Data were presented as %(n/N) unless stated otherwise.
BMI, body mass index.
a Physical activity: high was defined as greater than or equal to 5 times active for greater than or equal to 30 minutes a week, moderate was defined as greater than or equal to 1 but less than 5 times active for greater than or equal to 30 minutes a week and low was defined as less than 1 time active for greater than or equal to 30 minutes a week.
b Education level: lowest: only elementary; low education: Lower technical or vocational education and general secondary education; medium education level: secondary technical or vocational education and senior general secondary education; high education level: higher technical or vocational education and university.
Small differences were also seen in smoking-related characteristics (Table 2). Smoking duration was lower among participants (p < 0.001), whereas numbers of cigarettes smoked per day was higher among participants (p < 0.001). However, no differences were observed in the number of pack-years smoked between participants and eligible nonresponders. Participants started smoking at a younger age (p < 0.001) and were more willing to quit smoking than eligible nonresponders (p < 0.001). Among current smokers, participants were more often in an advanced stage- according to the stages of change- to quit smoking compared with eligible nonresponders (p < 0.001). Participants reported significantly more smoking-related symptoms (p = 0.04) and had undergone a pulmonary function test more often (p < 0.001). However, no differences were seen in the self-reported outcome of these pulmonary function tests (p = 0.28).
TABLE 2Smoking-Related Characteristics of NELSON Control Group Participants and Eligible Nonresponders
Stage of change: precontemplation phase: does not want to stop, wants to stop but not in the next 5 years, wants to stop but not in the next year, wants to stop but not in the next 6 months. Contemplation phase: wants to stop in the next 6 months. Preparation: wants to stop in the next 1 month. Action: stopped less than 6 months ago. Maintenance: stopped greater than 6 months ago.
Spirometry: comparison between subjects with spirometry only.
Data were presented as % (n/N) unless stated otherwise.
a Motivated to quit smoking: comparison between subjects who are current smokers only.
b Stage of change: precontemplation phase: does not want to stop, wants to stop but not in the next 5 years, wants to stop but not in the next year, wants to stop but not in the next 6 months. Contemplation phase: wants to stop in the next 6 months. Preparation: wants to stop in the next 1 month. Action: stopped less than 6 months ago. Maintenance: stopped greater than 6 months ago.
c Smoking-related symptoms: coughing, sputum, dyspnea, and wheezing.
d Spirometry: comparison between subjects with spirometry only.
During the study period, the all-cause mortality rate among eligible nonresponders was higher compared with the participants (p < 0.001; Table 3).The eligible nonresponders had a higher mortality rate due to all types of cancer (p = 0.002), cardiovascular diseases (p < 0.001), respiratory diseases (p = 0.018), and noncancerous diseases other than cardiovascular or respiratory (p < 0.001). However, the proportion of deaths due to cancer was higher among participants (62.4% vs. 54.9%). Higher educational achievement was significantly associated with higher mortality from all types of cancer (χ2 17.3; p < 0.001). Furthermore, a longer follow-up was seen for participants (10 years vs. 9 years).
TABLE 3Mortality Rates (per 1.000 Person-Years) by Causes of NELSON Control Group Participants and Eligible Nonresponders
Cause of Death
Mortality Rate Ratio
All cancer types
Noncancerous diseases other than CVD or respiratory diseases
Participants were significantly more likely to participate in any of the mentioned screening programs compared with the eligible nonresponders (all p < 0.001, data not shown). The median physical distance from home to one of the nearby participating screening centers was significantly less for eligible nonresponders than for participants (16.9 km versus 17.9 km; p = 0.003).
This study investigated differences in characteristics and mortality profiles of participants of the NELSON trial and eligible nonresponders. Results of this study are essential to determine whether mortality results of the NELSON trial are generalizable to the target Dutch population for lung cancer screening.
Participants of the NELSON trial were significantly younger, had better self-reported health, were more physically active, and higher educated compared with eligible nonresponders, although the differences in proportions were modest. These results are in line with previous studies in cancer screening trials
Different recruitments methods may explain the differences in study populations between NELSON trial and DLST: the NELSON trial was designed to recruit only men at first, because of fewer Dutch women met the smoking-related inclusion criteria of the NELSON study. However, in the second recruitment women were also invited to allow the NELSON study results to be generalizable to women. In contrast, the DLST recruited both sexes from the start of the study. Such overrepresentation of women participating in screening trials is also seen by others and may be because women are more used to screening from other cancer screening programs.
However, the DLST also reported that despite active smoking, participants were more willing to quit smoking than nonparticipants (a representative sample from the Danish population), suggesting that smokers who are motivated to quit smoking are more inclined to volunteer in a screening trial.
Eligible nonresponders had a higher all-cause mortality and mortality due to four other mortality classifications. However, the relative proportion of subjects that died due to all types of cancer was higher among participants. This might be explained by alcohol abuse, which is associated with higher socioeconomic status, e.g., higher educational achievement.
Another explanation might be that participants reported more smoking-related symptoms, which may have led to more general practitioner consults. This may have led to the higher proportion of former smokers among participants and could have facilitated the detection of cancer, cardiovascular, and respiratory diseases. This may have resulted in early treatment of smoking-related diseases among NELSON participants and may have led to lower mortality rates compared with eligible nonresponders. However, the slightly younger age, better self-reported health, and healthier lifestyle among participants may have had a bigger contribution to these differences in mortality profiles and resulted in a significantly longer follow-up among participants.
As mentioned, participants were more likely to participate in any of the mentioned screening programs compared with the eligible nonresponders. Higher education levels may have led to more awareness of their risk for lung cancer and influenced the decision to participate in the NELSON trial. In addition, there were more former smokers among participants. It has been previously reported that active smoking is a barrier to participate in screening for lung cancer.
Notable, living further from participating screening center, participants in the NELSON trial were more willing to participate than the eligible nonresponders. In contrast, the Lung-SEARCH screening trial reported that half of the responders found inability to travel the most significant reason not to participate.
The main strengths of this study are: (1) the large number of participants and eligible nonresponders, (2) access to all the completed first questionnaires of subjects, (3) the availability of mortality data from Statistics Netherlands, and (4) a long follow-up duration of 10 years. Finally, so far no large lung screening trial using LDCT has studied the differences in baseline characteristics and potential effect on mortality profiles between participants and eligible nonresponders.
This study was limited by the fact that Statistics Netherlands could only provide aggregated mortality data. Therefore, it was not possible to perform multivariate analyses. Furthermore, all questionnaire data were self-reported, as in other studies.
The questionnaires included few questions on socioeconomic class and no questions on ethnic background or psychosocial profile.
In conclusion, differences in age, health, lifestyle, and socioeconomic class can lead to a healthy participant effect, i.e., a different study outcome than would have been observed if the characteristics of participants were similar to that of the target population. As expected, the distribution of participant characteristics in the NELSON study suggest that the study population is somewhat younger, healthier (e.g., more physically active, less current smokers), higher educated and more willing to participate in a screening program. These differences have influenced the mortality outcome of participants and eligible nonresponders. But, these differences are modest and therefore it seems unlikely that these differences will influence the generalizability of the main results of the NELSON trial.
The authors thank the secretary M. Quak and the system controllers R. Faber and F.J.P. Santegoets (all from Department of Public Health, Erasmus University Medical Center) for their contribution and maintenance of the database. Further, we thank Statistics Netherlands for providing mortality data.
The NELSON trial is financially supported by “ZonMW: the Netherlands Organisation for Health Research and Development” (grant number 120610015), “KWF: Dutch Cancer Society” (grant number EMCR 2007–3857), “Erasmus MC trust fund”, “Health Insurance Innovation Foundation”, “G. Ph. Verhagen Stichting”, “Rotterdam Oncologic Thoracic Steering committee” (ROTS), Roche for providing a grant for the performance of proteomics-research, Siemens Germany for providing four digital workstations, “Stichting tegen Kanker” and “Vlaamse Liga tegen kanker”. The funding sources had only financial roles.