| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
Fox Chase Cancer Center (S.M.), Philadelphia, Pennsylvania; Research Analysis and Consultation, (M.G.), and Mt. Sinai School of Medicine (K.D.H.), New York, New York.
Address reprint requests to: Sharon L. Manne, PhD, Fox Chase Cancer Center, 510 Township Line Road, Philadelphia, PA 19012. Email: sl_manne{at}fccc.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: Patients (N = 189) undergoing treatment for cancer completed questionnaires at three time points, spaced 3 months apart. Intrusive thoughts, functional impairment, and psychological distress were assessed at Time 1, avoidance and functional impairment at Time 2, and psychological distress was assessed again at Time 3. The fit of the model was tested separately for patients with early-stage (stages 1 and 2) and late-stage (stages 3 and 4) disease.
RESULTS: The mediational role for avoidance was supported among patients with advanced stages of cancer but not for patients with early-stage disease. Results were inconsistent with predictions about the role of physical impairment. Among individuals with late-stage cancer, changes in functional impairment were not predictive of greater avoidance, and impairment had a significant but weak effect on the change in distress. Among patients with early-stage cancer, a deterioration in physical impairment was associated with increases in avoidance, and deterioration in physical impairment increased distress.
CONCLUSIONS: The results of this study were partially consistent with Creamer’s cognitive processing theory. A moderating effect was found for disease stage on associations between intrusions, avoidance, physical impairment, and distress.
Key Words: cognitive processing • cancer • avoidance • intrusive thoughts
Abbreviations: CARES = Cancer Rehabilitation and Evaluation System; OR = odds ratio; CI = confidence interval; SEM = structural equation modeling; CFI= Comparative Fit Index; GFI = Goodness of Fit Index; AGFI = Adjusted Goodness of Fit Index; NNFI = Nonnormed Fit Index; RMSEA = root mean square error of approximation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Adaptation to cancer may also be related to individuals’ cognitive processing of the disease (8). There has been a great deal of recent theoretical and empirical attention paid to the psychological effects of traumatic life experiences. Among the theories proposed to examine reactions to trauma, cognitive processing theory is perhaps the most fully developed. A common theme underlying most cognitive theories is that healthy adaptation is the result of repeated confrontation with the memories of the trauma (9–12). One cognitive processing theory, recently developed by Creamer and colleagues (13–14), is a synthesis and reconceptualization of existing formulations. Horowitz’s formulation of stress-response syndromes significantly influenced Creamer’s model. Horowitz (10, 15, 16) described two types of responses to trauma: an initial "outcry," which consists of repetitive intrusive thoughts about the trauma, and a subsequent suppression of affect by utilizing denial or deliberate avoidance of reminders of the trauma. Thus, Creamer’s theory suggests that avoidance mediates the association between intrusions and psychological adaptation. Avoidance increases distress because thoughts and memories are not confronted directly and thus are not processed sufficiently.
Several research studies have examined the relation between early intrusion and psychological outcomes. McIntosh et al. (17) prospectively studied parents of infants who had died of sudden infant death syndrome. The results indicated that voluntary attempts to think about the loss were associated with greater concurrent distress, but they predicted lower distress 18 months later. Likewise, Creamer et al. (14) studied office workers who were present in a building in which a mass shooting occurred. Higher levels of intrusion 4 and 8 months after the trauma predicted lower psychological distress at 8 and 14 months posttrauma.
However, the mediational role of avoidance in the relationship between intrusion and psychological adjustment has received inconsistent support in the two studies that have examined this relationship directly (14, 18). Creamer’s results were partially consistent with a mediational role of avoidance (14). At the first assessment, intrusion predicted avoidance, which, in turn, predicted psychological distress. However, avoidance did not predict distress at two follow-up assessments. Similarly, McFarlane (18) examined the mediational role for avoidance in the association between intrusion and psychological distress among firefighters. Cross-lag panel correlations indicated that intrusion at 4 and 11 months posttrauma predicted concurrent levels of avoidance but that avoidance was not correlated with later distress. However, intrusions predicted later distress at both time points. Because avoidance and intrusion were measured concurrently, conclusions about the temporal association between intrusions and avoidance cannot be made.
Relatively little empirical attention has been paid to examining the proposed mediating role of avoidance in the relation between intrusive thoughts and psychological adaptation among individuals dealing with cancer. Several recent studies have documented the association between intrusive thoughts about cancer and concurrent psychiatric morbidity (19–22). In contrast, the association between avoidance and cancer patients’ psychological distress has been inconsistent. Some studies have compared the contribution of intrusion and avoidance and found intrusion to be a stronger predictor of distress among patients with cancer (22). Cognitive avoidance has been associated with poor psychological adjustment in studies of patients with breast cancer (5, 23–26), lung cancer (27), and heterogeneous cancer (28–29). However, avoidance has also been associated with better psychological adjustment in other studies (30–32).
When the role of avoidance in processing a traumatic experience is examined, it may be important to take contextual factors into account. Suls and Fletcher (33) point out that the seriousness of the stressor may be an important contextual factor to consider when examining the relation between avoidance and later psychological distress. They propose that the more serious the implications of the event, the more likely that major intrapsychic readjustments are necessary and the less efficacious avoidance is as a coping strategy. In the oncologic context, the stage of disease would be a key index of severity of the stressor. Patients diagnosed with more advanced disease such as metastatic cancer are more likely to face a difficult disease course and negative outcomes (eg, death) than patients diagnosed with localized disease. Thus, one would predict that the use of avoidance would have more detrimental implications for patients diagnosed with advanced stages of cancer.
A second important contextual factor that influences the association among intrusions, avoidance, and psychological outcomes during cancer treatment is the degree of deterioration in physical impairment. Those individuals who experience a decline in physical functioning may be more likely to have increasing levels of psychological distress over the course of treatment. In addition, a decline in functional functioning may serve as its own cue for more avoidance and thereby influence cognitive processing.
Taken together, prior studies suggest that both intrusions and avoidance play a role in the adaptation to traumatic experiences. In the current study, we examined the longitudinal associations among intrusions, avoidance and psychological distress in a sample of patients cancer who were undergoing treatment for their disease. We tested Creamer’s model, proposing a mediating role of avoidance between intrusions and psychological distress. We also examined the role of two contextual factors, disease stage and functional impairment. A potential moderating role for disease stage was proposed. It was predicted that avoidance would have a significantly stronger association in predicting distress among patients with advanced-stage disease when compared with its association in predicting distress among patients with early-stage disease. Deterioration in physical functioning was hypothesized to be associated with more avoidance, as well as with an increase in patients’ distress.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Fifty-two percent of the participants were diagnosed with advanced-stage disease (stage 3 or 4). The types of cancer were heterogeneous, but all patients had solid tumors. The following were most common tumor sites: breast cancer (34%), colon or rectal cancer (24%), lung cancer (9%), and prostate cancer (20%). Thirty-seven percent of the sample had a metastatic site, with the most common metastatic sites being liver (36%), lung (18%), bone (10%), and lymph nodes (16%). Most participants had been diagnosed with a first occurrence of cancer within the preceding 6 months (80%). The remaining participants were either diagnosed with a disease recurrence or were diagnosed with a first occurrence of cancer >6 months before starting the study (20%). Among individuals diagnosed with recurrent disease, the median time from the original cancer diagnosis was 4 years, 8 months, with a range from 6 months to 13 years since the original diagnosis (all but one of the recurrences occurred within 6 years). All patients were receiving treatment for the cancer at the time of the first study assessment. Approximately 75% of participants were receiving chemotherapy, and 25% were receiving radiation treatments. At the time of the third assessment, 79% of patients were still receiving medical treatment for their disease. Approximately 59% of participants had received chemotherapy within the month before the third assessment, and 10% had received radiation in the month before the third assessment. Another 10% were receiving hormone suppression therapy (administered after radiation for individuals with locally advanced prostate cancer). The remaining 21% of patients were not receiving treatment at the 6-month assessment. At the Time 2 assessment, the response to treatment was categorized as follows: 42.9% of patients had disease that responded to chemotherapy or other interventions, 35.5% had stable disease, 7.9% had disease that was progressing or had progressed, 5.1% of patients had disease response that was not yet known, and 8.5% of patients had no evidence of disease. For patients diagnosed with non–sex-specific cancers, there were approximately equal percentages of men and women. However, cancer stages were not equally stratified among the cancer types: patients with breast cancer and prostate cancer were more likely to have an early-stage diagnosis (64% breast cancer and 70% prostate cancer), whereas patients diagnosed with other types of solid tumors were less likely to have an advanced-stage cancer (eg, 32% of colon cancer patients had an early-stage diagnosis).
Procedure
Four hundred seventy-two individuals were approached for study participation from the outpatient clinics of oncologists practicing in two large comprehensive cancer centers. Criteria for inclusion in the study were (1) participant 18 years of age or older; (2) participant having no known neurological impairment; (3) participant able to provide meaningful informed consent; and (4) participant receiving active medical treatment for the cancer (eg, chemotherapy and/or radiation). Eligible participants were identified by the study assistant and approached during an outpatient visit. Participants were given the first study questionnaire at this time (Time 1) and asked to mail the questionnaire back in a stamped envelope. Two follow-up questionnaires were mailed to participants who completed the first questionnaire at 3 months (Time 2) and 6 months after the initial questionnaire (Time 3). Participants who did not complete a study questionnaire were contacted several times by phone to encourage return of questionnaires.
Eighty individuals (16.9%) refused participation. Seventy-two patients (14.8%) provided informed consent but did not return the Time 1 questionnaire. One hundred forty-three participants (30.2%) completed the Time 1 questionnaire but did not complete one or both of the two follow-up questionnaires. The most common reasons given for study refusal were "don’t have time" and "too much stress right now." The final sample consisted of 189 participants. The participation rate across all assessments was 40%.
Measures
Intrusive thoughts. The 7-item Intrusive Thoughts scale of the Impact of Events Scale (34) was used to measure the frequency with which participants had intrusive thoughts about cancer. Participants were asked to endorse the frequency of thoughts about cancer during the prior week using a 4-point Likert scale (0 = "not at all" to 5 = "often"). Items included "I had dreams about it," "Pictures of it popped into my mind," and "Thought about it when I didn’t mean to." This measure was administered at Time 1. Cronbach’s alpha was 0.88.
Avoidance. The 8-item Avoidance scale of the Impact of Events Scale (34) was used to measure cognitive and behavioral avoidance. Participants were asked to rate how often they attempted to avoid thinking about cancer (eg, "I tried not to think about it" and "I tried to remove it from memory") and how often they attempted to avoid reminders of the cancer (eg, "I stayed away from reminders of it."). Participants were asked to endorse the frequency of these behaviors over the course of the prior week using a 4-point Likert scale (0 = "not at all" to 5 = "often"). This measure was administered at Time 2. Cronbach’s alpha was 0.80.
Psychological distress. The psychological distress subscale of the Mental Health Inventory (35) was used. This scale consists of 24 items that assess anxiety, depression, and lack of behavioral and emotional control. Participants used a 5- or 6-point Likert scale to rate their feelings over the past month. Higher scores indicate more distress. The scale was administered at Times 1 and 3. Cronbach’s alpha for this distress scale was 0.92 at Time 1 and 0.93 at Time 3.
Functional impairment. The physical functioning subscale of CARES (36) was used to assess the level of functional impairment induced by the cancer and its treatment. The scale contains 26 items that assess the degree to which the cancer treatment interferes with the performance of daily activities. An example is "I find have difficulty doing household chores." Ratings range from 0 (not at all) to 4 (very much). Patients were asked to provide ratings of their functioning during the prior month. Higher scores indicate higher levels of functional impairment. The scale was administered at Times 1 and 2. Cronbach’s alpha for this scale was 0.91 at Time 1 and 0.94 at Time 2.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Separate logistic regression equations were also conducted for disease stage, functional impairment at Time 1, Time 1 psychological distress, and intrusions. Results indicated that disease stage (OR = 1.45, CI = 1.16–1.80) and functional impairment (OR = 1.0, CI = 0.96–1.04) were significant predictors of completion status. Analyses indicated that the participant group had a significantly higher percentage of patients with Stage 1 (OR = 0.069, CI = 0.0066–0.727) and Stage 2 (OR = 0.074, CI = 0.0078–0.711) disease. Time 1 psychological distress and Time 1 intrusions were not significant predictors of completion status.
These analyses suggested that selection bias may be present and that education and disease- and treatment-related factors may discriminate between those subjects who completed the study and those subjects who did not complete the study (dropouts).
Estimation of the Structural Equation Model and the Moderating Effect of Disease Stage
Descriptive statistics. Descriptive statistics for the full sample and the subsamples of patients with early- and late-stage disease are presented in Table 1. A comparison of the distress scores with the norms derived from the general community sample (M = 47.45, SD = 15.39) (35) indicate that distress is slightly but not significantly higher in the present sample at both time points. Zero-order correlations between variables included in the model are presented in Table 2. Several interesting relationships are illustrated in the correlation matrix. First, Time 1 intrusions and psychological distress at Time 1 and 3 were moderately to strongly correlated. Second, results indicate that the correlation between avoidance at Time 2 and distress at Time 3 was of a moderate magnitude (r > .3).
|
|
Second, the moderational effect for disease stage was examined. To examine whether the predicted moderator effects were present, we formed two groups on the basis of their disease stage: an early disease stage group (N = 90), composed of participants with stage 1 or 2 cancer, and an advanced-stage group (N = 99), composed of participants with stage 3 or 4 cancer. Typically, testing for the moderating effect would entail two steps: (1) finding separate good-fitting model for the group of patients with early stage disease and for the group with late stage disease; and (2) testing for equality of the separate models. The test, named the test of equality of the covariance matrices, determines whether or not there were significant differences in the parameters of the two subgroup models. If this test is not significant, there are differences between the models for the two subgroups and the moderational hypothesis is supported (it was hypothesized that avoidance would have a stronger association with distress among patients with advanced stage disease than among patients with early stage disease). The SEM representing the hypothesized relationships were tested by use of LISREL 8.3 software (37). Maximum-likelihood estimation was used; when residuals are independent and sample sizes are <500, this approach is preferable (38). The fit of the model was assessed by use of multiple indices. Chi-square and degrees of freedom are reported. A nonsignificant chi-square indicates that the model fits the data, but this statistic is highly sensitive to the number of participants and the complexity of the model (39). Current general recommendations suggest the use of multiple indicators of model fit. Five measures of fit were used to supplement the chi-square: the CFI, the GFI, the AGFI, the NNFI, and the RMSEA. A model that met the criteria of a CFI, GFI, AGFI, and NNFI of >0.90 and a RMSEA value of 
0.05 was judged to be a good fit (40–41).
Separate group models. The model provided a good fit to the data in the late-stage subgroup (Satorra Bentler 
2 = 13.33, df = 5, p = .021, RMSEA = 0.14, GFI = 0.98, AGFI = 0.92, NNFI = 0.99, and CFI = 1.0) but not to the data for the early-stage subgroup (Satorra Bentler 2 = 13.33, df = 5, p = .021, RMSEA = 0.14, GFI = 0.95, AGFI = 0.79, NNFI = 0.93, and CFI = 0.95). This finding suggests that stage may be a moderator of the proposed models. Figure 1 presents the parameters for the late-stage subgroup, and Figure 2 presents the parameters for the early-stage subgroup.
|
|
With respect to functional impairment, neither Time 1 functional impairment nor Time 2 functional impairment was statistically significant predictors of Time 2 avoidance. The latter relationship may be conceptualized as the effect of a change in functional impairment from Time 1 to Time 2 on the level of Time 2 avoidance. When this conceptualization is used, these results suggest that changes in functional impairment have no effect on the participants’ levels of avoidance at Time 2. However, the level of Time 2 functional impairment does have a significant but weak effect on change in distress at Time 3. Individuals with higher levels of functional impairment at Time 2 have larger than expected increases in psychological distress at Time 3.
Early-stage model respecification. As was noted above, analyses of the early stage model suggested that the proposed model did not fit the data well (Satorra Bentler 2 = 13.33, df = 5, p = .021, RMSEA = 0.14, GFI = 0.95, AGFI = 0.79, NNFI = 0.93, and CFI = 0.95). In order to find the best-fitting model to these data, we respecified the model using modification indices computed by LISREL. Modifications that would likely result in a significant decrease in chi-square were implemented one at a time (37). The difference in chi-square between two comparison models has a chi-square distribution with degrees of freedom equal to the difference in degrees of freedom between the models being compared (42). Examination of the modification indices suggested two additional paths. The strongest modification index suggested a path from Time 1 functional impairment to Time 3 distress. However, once this path was added, the fit of the model was not improved to a significant degree (Satorra Bentler 2 = 11.30, df = 4, p = .02, RMSEA = 0.14, GFI = 0.96, AGFI = 0.78, NNFI = 0.86, and CFI = 0.96). On the basis of these results, this path was removed, and the second modification index was added, which consisted of a path from Time 1 intrusions to Time 3 distress. Once this path was added, the fit of the model was still not improved to a significant degree (Satorra Bentler 2 = 10.45, df = 4, p = .033, RMSEA = 0.13, GFI = 0.96, AGFI = 0.79, NNFI = 0.88, and CFI = 0.97). Examination of the modification indices suggested that the addition of the path from Time 1 functional impairment to Time 3 distress would improve the model. On the basis of these indicators, a path from Time 1 functional impairment was added in the model after the path from Time 1 intrusions to Time 3 distress. The resulting model with both additional paths fit the data much better than the model with two separate paths added separately (Satorra Bentler 2 = 4.87, df = 3, p = .18, RMSEA = 0.08, GFI = 0.98, AGFI = 0.88, NNFI = 0.96, and CFI = 0.99). Although the RMSEA value suggests that the fit is marginally acceptable, the other fit indices are generally acceptable. The final model for the early-stage subgroup is shown in Figure 2.
An inspection of Figure 2 suggests that the correlation between Time 1 intrusions and distress is identical to that found in the late-stage subgroup (r = .70, p < .05). However, the correlations between Time 1 functional impairment and each of the other exogenous variables are stronger than those observed in the early-stage subgroup. The direct effect of Time 2 intrusions on Time 2 avoidance is sizeable and significant (ß = 0.53, p < .05), as it was in the late-stage subgroup. However, the effect of Time 2 avoidance on the change in distress is virtually nonexistent (ß = 0.04, p > .05), indicating that the total effect of Time 1 intrusions on change in distress is almost completely direct (ie, not mediated by avoidance at Time 2 (ß =0.35, p < .05). Not surprisingly, the indirect effect of intrusions on change in distress (Time 3) via Time 2 avoidance is not significant (standardized indirect effect =0.02, t = 0.45, p > .05). This discrepancy between the early- and late-stage models is, perhaps, the most theoretically interesting finding. However, there are several other ways in which the two disease stage models differ.
First, unlike the late-stage model, there is a statistically significant path from Time 2 functional impairment to Time 2 avoidance (ß = 0.27, p < .05). On the basis of the alternative explanation of this particular path outlined above, increases (changes) in functional impairment between Time 1 and Time 2 lead to increases in the level of avoidance at Time 2. Because avoidance at Time 2 does not effect change in distress at Time 3, there is no significant indirect effect of change in functional impairment on change in distress at Time 3 (standardized indirect effect =0.01, p > .05). Second, there was a significant positive effect of change of functional impairment at Time 2 on change in distress at Time 3 (ß = 0.50, p < .05). This path is noticeably larger than its counterpart in the late-stage model. Because the path from baseline functional impairment to Time 3 distress was added to improve the model fit, the path from Time 2 functional impairment to Time 3 distress can be interpreted as indicating that increases in functional impairment from Time 1 to Time 2 are predictive of greater than average expected increases in distress at Time 3. The negative effect from Time 1 functional impairment to Time 3 distress suggests that individuals with higher levels of functional impairment at Time 1 changed less than expected with regard to distress at Time 3 (eg, they improved; ß = -0.30, p < .05).
This counterintuitive finding has been observed by other investigators who used linear panel analysis techniques in the context of two-wave, two-variable (panel) models (43, 44). Both of these investigators have pointed out that when the coefficient for the "change effect" (in this study, Time 2 functional impairment to distress at time 3; ß = 0.50) is large relative to the "static score effect" (in this study, Time 1 functional impairment at Time 2 to Time 3 distress; ß = 0.30), the sign of the static score effect will be negative, as it is here (44). Thus, the negative sign is a statistical artifact.
In two-wave, two-variable models, one can describe the effect of the static score variable (Time 1 functional impairment on residualized change in distress) by adding the coefficients for the static score and change effects by use of the unstandardized regression coefficients. Although the present model is not a two-wave, two-variable model, the sum was computed by use of the standardized coefficients (which are virtually identical to the unstandardized coefficients) and tested for statistical significance ([0.50+(-0.30)] = 0.20, p < .05). This significant result implies that individuals with higher levels of Time 1 functional impairment actually deteriorated with respect to Time 3 distress (ie, after controlling for other predictors of distress at Time 3, they reported higher distress scores at Time 3 than would be predicted from their Time 1 distress scores).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The principal finding of this study was that, among individuals with stage 3 and 4 cancer, intrusive thoughts were related to higher avoidance and had an indirect effect on the psychological distress through avoidance. To our knowledge, this finding is the first evidence indicating that the relation between intrusive thoughts and later psychological distress may be mediated by cancer patients’ attempts to avoid thinking about or dealing with reminders of the cancer experience. These findings are consistent with those of Creamer and colleagues (13, 14), who found that avoidance mediated the association between intrusive thoughts and distress.
A second key finding was that avoidance predicted later distress among individuals with late-stage cancer, after controlling for initial levels of distress, but did not predict distress among patients with early-stage cancer. The findings for patients with late-stage cancer are consistent with cognitive processing theories of trauma. As Foa and colleagues (45) note, "avoidance tactics are not adaptive since they do not allow emotional processing to take place." These findings also complement studies of individuals with breast cancer (eg, Refs. 5, 26) and individuals recovering from burn injury (46), which suggest that avoidance coping is a prospective predictor of distress.
In contrast, the results indicated that avoidance did not mediate the association between intrusions and distress among patients with early-stage cancer. Rather, intrusions were directly associated with increased distress. The insignificant role of avoidance for patients with early-stage cancer is consistent with some studies of patients with early-stage breast cancer (eg, Ref. 21) but is not consistent with other studies of patients with early stage breast cancer (eg, Ref. 26). One explanation for the apparent moderating effect of disease stage is that avoidance may not be detrimental when the cancer experience is short-term but may be detrimental when the cancer experience is long-term. An examination of the sample suggests that 36% of the late-stage patients had been diagnosed >6 months before beginning the study, and the median time since the original diagnosis for the recurrent cancer patients was >4 years. Meta-analytic studies have suggested that avoidant strategies are associated with more positive outcomes when used soon after the stress begins but are associated with poorer outcomes in the long run (33). Overall, our results suggest that avoidance may be maladaptive for more serious cancer threats with a less certain positive medical outcome and a longer timeline (eg, advanced cancer). Although early-stage cancer can be traumatic, the treatment is time-limited and the cure more certain. The treatments for late-stage cancer are less likely to be time-limited, and the potential for cure is less certain.
The association between functional impairment and distress is consistent with recent studies of patients with breast cancer (47) and of a mixed sample of patients with cancer (48). However, a closer examination revealed surprising findings with some inconsistency with the study’s hypotheses. We predicted that deterioration in physical abilities would be viewed as a reminder of a poor prognosis for late-stage patients and therefore result in avoidance and distress. However, our results indicated that functional impairment played a much stronger role in predicting avoidance and psychological distress among early-stage patients. This finding may be explained by alterations in expectations for physical functioning. Early-stage patients may have higher expectations for their level of physical functioning, compared with late-stage patients. Whereas late-stage patients may lower their expectations as a way of preparing for a decline in physical functioning and thus be less upset by a decline, early-stage patients may be less prepared and therefore be more upset by a decline. Late-stage patients may be more likely to reduce their expectations and goals for physical functioning, a process that has been labeled "response shift" in the literature (49, 50).
Limitations of the present study should be noted. First, a longer follow-up period would allow for assessment of long-term cognitive processing. Second, the sample was heterogenous with regard to cancer types. The sample was composed of a high percentage of individuals with newly diagnosed late-stage cancer and patients with recurrent disease. In addition, the early- and late-stage subgroups were not equally stratified in terms of cancer diagnosis (eg, more breast and prostate cancer patients in the early-stage group). In order to assess the moderating role of cancer stage, we formed two groups of patients with cancer. We did not examine the fit of the hypothesized model separately for different cancer sites, for each cancer stage, and for patients with recurrent vs. newly diagnosed disease. It is possible that cognitive adaptation processes differ for these patients, and future studies should focus on these subgroups. Third, the sample was composed of mostly white and educated patients, which may limit the ability to generalize from these findings. Fourth, the dropout rate for this study was high (60%). Our analyses indicated that individuals who dropped out were less educated, more functionally impaired at baseline, and diagnosed with a later stage of disease. The fifth limitation concerns the procedures used for finding the best-fitting model once the initial model for early-stage patients did not fit. The use of modification indices to determine which paths should remain in the model may have led to inflation of the fit of the model. This procedure, sometimes called a specification search (51), is data-driven and is thus susceptible to capitalization on chance (52). For this reason, a replication with another sample of patients with cancer should be conducted. Another limitation of the study is that other variables included in cognitive processing models such as elements of the fear network (ie, threat to life; Refs. 10, 53) were not measured. Finally, we proposed a sequence of events, but alternative sequences are possible, particularly transactional ones. For example, it is possible that distress influences intrusive thoughts.
Nonetheless, this study provides one of the first steps in examining the application of cognitive processing models to individuals coping with cancer. The study is also the first attempt in the cancer literature to examine a theoretical conceptualization of how two key components of cognitive processing, intrusion and avoidance, unfold. The present study also attempted to take into account contextual factors such as disease stage and changes in functional impairment when considering models of adaptation to cancer. As has been pointed out by recent reviews of cognitive processing, the predictive utility of these theories has not received sufficient attention (9). In particular, the application of this theory to ongoing rather than brief events requires additional examination. Future studies should expand on this work in at least three ways. First, studies should examine both intrusion and avoidance over the course of cancer treatment. Second, studies should include other components of cognitive processing, such as fear network variables (eg, life threat) (14) and accommodative and assimilative processes (54, 55). Third, future analyses should examine different types of avoidance. The Impact of Events Scale does not distinguish among suppressive efforts (eg, "I tried not to think about it"), physical avoidance (eg, "I stayed away from reminders of it"), and social avoidance ("I tried not to talk about it"). It is possible that certain types of avoidance are more detrimental than other types.
Refining existing models of cognitive adaptation to cancer has the potential to help clinicians who work with patients who have cancer by fostering the development of new intervention methodologies. The findings from this study suggest that different interventions might be appropriate for individuals with late- vs. early-stage cancers. Interventions to cope with intrusive thoughts (for both early- and late-stage patients) and avoidant coping with cancer-related intrusive thoughts (for early-stage patients) should be evaluated in randomized clinical trials. Intervention strategies that emphasize exposure to trauma-related experiences, similar to approaches used with individuals undergoing other traumatic experiences, might be future targets (55–58). However, because exposure therapies may initially raise distress levels, these approaches might be too upsetting for patients currently in treatment and might be best incorporated into preparatory interventions delivered before treatment onset or after treatment is completed, to prevent long-term distress.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication December 10, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. D. Cameron, R. J. Booth, M. Schlatter, D. Ziginskas, J. E. Harman, and S. R.C. Benson Cognitive and Affective Determinants of Decisions to Attend a Group Psychosocial Support Program for Women With Breast Cancer Psychosom Med, July 1, 2005; 67(4): 584 - 589. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. N. DuHamel, S. Manne, N. Nereo, J. Ostroff, R. Martini, S. Parsons, S. Williams, L. Mee, S. Sexson, L. Wu, et al. Cognitive Processing Among Mothers of Children Undergoing Bone Marrow/Stem Cell Transplantation Psychosom Med, January 1, 2004; 66(1): 92 - 103. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Manne, J. Ostroff, M. Sherman, M. Glassman, S. Ross, L. Goldstein, and K. Fox Buffering Effects of Family and Friend Support on Associations between Partner Unsupportive Behaviors and Coping among Women with Breast Cancer Journal of Social and Personal Relationships, December 1, 2003; 20(6): 771 - 792. [Abstract] [PDF]
|
|
|
|
 |
|
 M. Gurevich, G. M. Devins, and G. M. Rodin Stress Response Syndromes and Cancer: Conceptual and Assessment Issues Psychosomatics, August 1, 2002; 43(4): 259 - 281. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
