A Randomized Controlled Study of Short-term and Long-term Effects of Tailored Information versus General Information on Intention and Behavior Related to Early Detection of Cancer¹

It is widely accepted that the earlier cancer is detected, the better the chances of treatment and survival (1). Therefore, people should be encouraged to engage in early detection behaviors. Many programs have been developed to stimulate people to perform breast or testicular self-examination and to participate in screening programs for breast and cervical cancer (2, 3, 4). Far less has been done to enhance attentiveness to general cancer symptoms and intention to seek help for these symptoms.

A promising and relatively new approach in health education is computerized tailoring, which adapts health education messages to the characteristics, needs, and interests of the recipient (5). This leads to more personally relevant information, which is more likely to be thoughtfully considered (6) and is therefore thought to be more effective in changing determinants and behaviors than generic information. For a computer-tailored intervention, three elements are necessary: (a) a screening questionnaire, providing data on which the tailored feedback will be based; (b) a message source file, containing feedback messages for all possible answers included in the screening questionnaire; and (c) a computer program to connect the screening questionnaire and the message source file to facilitate the combining of messages and to print them as a personal letter. Computerized tailoring has been shown to change intentions and behaviors (for an overview, see Ref. 7).

Thus far, no tailored interventions have been developed or evaluated to motivate asymptomatic people to engage in early cancer detection behaviors, except for those stimulating women to have a mammography. Therefore, a randomized controlled study was conducted to test two different interventions, tailored information and a standard brochure currently provided by the Dutch Cancer Society (8), to encourage the Dutch adult population to participate in early detection behaviors. The effects of these interventions on passive detection and intention to seek help immediately after the intervention and after 6 months were studied and compared with the same parameters in a control group who received no information. It was hypothesized that the changes in intention and behavior would be significantly more positive for the recipients of the tailored information.

The study was conducted among 1855 Dutch adults, recruited via local and national newspapers. Subjects were randomly assigned to the three study groups. After registration, subjects received information about the study, an informed consent form, and the first questionnaire (T0). Participants were told that they participated in a study on the effects of different kinds of information about early detection of cancer. The tailored and general information were mailed to the subjects within 3 weeks after the first questionnaire had been returned. The content of the interventions and the process evaluation are described elsewhere (9). Members of the control group did not receive any information, but they were given the general information after completion of the study. Three weeks after the intervention (for the control group, 6 weeks after the first questionnaire), all subjects received a second questionnaire (T1). All subjects whose second questionnaire was received and who did not suffer from cancer at T0 were approached by telephone for a short interview on their behavior, and intentions regarding early detection 6 months after the intervention (T2). The study was approved by the university’s medical ethics committee.

Passive detection and help-seeking intention were measured at T0 (screening questionnaire and pretest), T1 (first posttest), and T2 (second posttest; Table 1).

Respondents who completed the three questionnaires and who did not suffer from cancer at T0 were included in the analysis. All analyses were done with SPSS.

Attrition was studied by means of logistic regression analysis with attrition as the dependent variable and the demographics and condition as predictors. χ² tests and F tests were performed to analyze whether significant differences were found in drop-outs of the different study groups. Baseline characteristics between study groups were compared by means of χ² tests and F tests.

To investigate the effects of the intervention on passive detection and help-seeking intention in the different study groups, differences between study groups at T1 and T2 were analyzed by means of repeated measure analyses of covariance, adjusting for the scores at T0. In the case of a group-time interaction effect, indicating that the mean difference between the two posttests depended on the study group, separate analyses of covariance were performed to study the effects on T1 and T2. Several covariates were included (intention and behavior at T0 and demographics). When a group effect was found, pairwise comparisons using Bonferroni (adjusted α = 0.0167) were performed to analyze differences between study groups, based on the estimated marginal means (i.e., corrected for the covariates). In the case of a group-time interaction effect, paired t tests were also performed to compare means between T1 and T2 within each study group.

Of the initial 1855 volunteers subjects, 1500 met the criteria for being approached for the telephone questionnaire at T2. A total of 1358 (73%) subjects completed the telephone questionnaire, equally distributed across the study groups. Attrition analysis revealed that drop-outs less often had cancer in the past (P < 0.01), less often had a partner (P < 0.001), and more often were younger respondents (P < 0.01), but no significant differences in the distribution of these variables were found between study groups.

The study population was predominantly female (80%), on average 47 years old (SD 12.93), and with a partner (79%). Of the respondents, 36% had completed primary school or basic vocational training, 35% had completed secondary vocational training or high school, and 29% had a higher vocational or university degree. Most respondents (92%) had been confronted with someone with cancer in their immediate environment, whereas 11% of the respondents had suffered from cancer themselves earlier in their lives. No differences in demographics, intention, and behavior between the study groups were found, with one exception (having had cancer in the past, P < 0.05).

Table 2 lists the mean scores of passive detection and help-seeking intention at T0, T1, and T2. Repeated measures analyses of covariance showed a significant group-time interaction effect for passive detection [F(2,1283) = 12.14, P < 0.001] and help-seeking intention [F(2,1294) = 8.51, P < 0.001], indicating that the three study groups differed significantly in the change in mean scores of passive detection and help-seeking intention between T1 and T2. The high mean scores of the tailored information group on passive detection and on help-seeking intention at T1 remained at the same level at T2 (ns), whereas the lower mean scores of the general information group and the control group continued to increase (Table 2).

At T1, a group effect was found on passive detection [F(2,1284) = 67.11, P < 0.001]. Pairwise comparisons showed that the tailored information group was more attentive to cancer symptoms than the general information group and the control group, whereas the general information group was more attentive than the control group (Table 2). There was also a significant group effect at T2 [F(2,1297) = 11.79, P < 0.001]. Pairwise comparisons showed that the tailored information group was more attentive than the general information group and the control group.

Similar analyses were performed for help-seeking intention. Analysis of covariance found a significant group effect at T1 [F(2,1294) = 36.98, P < 0.001]. Pairwise comparisons showed that the tailored information group reported more appropriate help-seeking intention than the general information group and the control group, whereas the general information group reported more appropriate help-seeking intention than the control group. A significant group effect was also found at T2 [F(2,1303) = 7.86, P = 0.001]. Pairwise comparisons showed a significantly higher score on help-seeking intention in the tailored group compared with both other groups.

The present study tested the impact of a computer-tailored intervention in encouraging people to engage in passive detection and an appropriately timed intention to seek help. Short-term effects showed that the recipients of the tailored information were more attentive to cancer symptoms and were more likely to seek appropriate help for cancer symptoms than the general information group and the control group after 3 weeks. After 6 months, significant differences in the dependent variables remained between the tailored information group and both of the other groups. Although differences seem rather small, on a population level these small differences could have an impact. This allows us to conclude that information tailored to the individual seems more effective than general information, which is in line with the findings of previous studies on tailored health information (7), and that tailored information is a viable alternative to existing methods.

Furthermore, we found a lack of change in the dependent variables between T1 and T2 in the tailored information group but additional effects in the general information group and the control group. Several explanations can be given, such as a social desirability effect due to comparisons of mean scores on the two posttests by using written questionnaires and telephone interviews, a ceiling effect that may have occurred in the tailored information group, or a testing effect that may have occurred because people are likely to score better on the third test than on the first two tests, even if their behavior has not actually changed. Moreover, it is likely that repetitive testing makes people aware that they should pay attention to the symptoms mentioned in the questionnaire.

It should be noted that we were not able to assess actual help-seeking behavior for cancer symptoms. People were asked what they would do if they experienced cancer symptoms, which assesses the appropriately timed intention to seek help. This does not necessarily mean that actual help is sought once a symptom is detected. However, because intention is generally the most significant predictor of behavior (10), actual help-seeking may be expected.

Although the short-term effects of the study were very much in favor of the tailored information, more research is needed to find out how the effects could be maintained in the long term. To prevent a decline in behavior and intentions toward early detection of cancer in the long term, the effects of multiple tailoring on relevant aspects of early detection should be studied in greater detail, for instance by ipsative feedback or nontailored reminders.