Introduction

The ability to express and regulate emotions is thought to play an important role in promoting and maintaining physical health [1]. Emotion dysregulation in the form of disattention to signals of distress has been hypothesized to result in physiological dysregulation due to cortical overriding of the autonomic nervous system and immunoregulatory homeodynamic mechanisms as well as increased risk-related behaviors [2]. Although the evidence is mixed [3], various measures of emotional avoidance and repressive coping have been found associated with both development [4, 5], and poor prognosis of breast cancer [6–8]. While acknowledging the methodological challenges in differentiating and measuring such concepts of impaired emotional processing [9], it remains relevant to explore possible effects of interventions aimed at improving emotion regulation on physical health outcomes in cancer patients and survivors. Expressive Writing Intervention (EWI)], is one type of intervention aimed at improving emotion regulation through emotional disclosure, and the present study extends our previously published findings concerning effects of EWI on psychological distress in our large population-based, randomized, controlled trial of EWI with women treated for primary breast cancer [11] by analyzing the effects on physical health outcomes.

According to the EWI paradigm, it may be possible to improve both mental and physical health through improving self-regulation of emotion-related experiences [12], and the results of early studies by Pennebaker and colleagues [13–15], did indeed suggest that EWI, instructing participants to write as little as 20 min for 3–4 days about emotions associated with a traumatic event, may be associated with psychological benefits as well as improvements in a number of physical health-related factors, including immune changes, reduced physical symptoms, and reduced health care utilization. Later studies have provided additional support for beneficial effects in both healthy samples [16, 17], and samples with physical illness [18–20]. The most comprehensive meta-analysis published so far [21] reviewed 146 randomized, controlled trials (RCTs) of various types of experimental disclosure in healthy and clinical samples and found statistically significant, but small, overall effect (r = 0.075≈Cohen’s d = 0.15). Specifically reviewing EWI with cancer patients and survivors, a recent meta-analysis of 16 RCTs suggests a negligible overall effect (Hedges’s g = 0.04;p = 0.38) [22]. Most pertinent to the present study are the 11 studies which investigated effects of EWI on various physical health outcomes in cancer patients, revealing a small and statistically non-significant effect (g = 0.08) [22].

Taken together, the research on physical health outcomes is heterogeneous and relatively sparse. However, results for both cancer and other populations suggest that written emotional disclosure may be more relevant for some participants than others. For instance, an emotionally under-regulated person tends to have little control over responses to provoking stimuli, and emotions are thus experienced more intensely with subsequent exaggerated physiological arousal and increased risk of, e.g. cardiovascular disorders [12]. An overregulated individual, in contrast, tends to avoid, inhibit, or suppress emotional responses, which may require physiological work that facilitates sympathetic dysregulation [23] with detrimental effects, e.g. on the immune function [20]. From an emotion regulation perspective, the following factors directly or indirectly associated with impaired emotion processing could thus be viewed as potential moderators. Repressive coping is defined as a tendency to under-report negative affect as a result of non-conscious protection of their self-esteem [24, 25], which may interfere with the effects of written disclosure [26]. Alexithymia is an insufficient ability to process emotions through cognitive mechanisms, and individuals who have difficulties describing their emotions, or lack the skills and ability to reflect upon and process their experiences, may benefit less from EWI [26, 27]. In contrast, rumination is viewed as conscious, spontaneous, and repetitive thoughts concerning past negative information that may influence affective responses to negative events [28]. Tendencies to ruminate have been associated with poorer self-reported physical health and increased health care utilization [29]. Although studies suggest that reduced rumination may explain the beneficial effect of EWI [30, 31], rumination has not yet been studied as a potential moderator of EWI in cancer patients. From a social perspective, social constraints, i.e., perceived inadequacy of social support resulting in a reluctance to express thoughts and feelings about a specific stressor [32], may hinder people from confronting and processing traumatic experiences, e.g., cancer [33], resulting in disturbed illness adaptation [34]. EWI could buffer against the negative consequences of social constraints by providing an opportunity to compensate for the lack of social support. This hypothesis has indeed found some support in EWI studies with cancer patients [22]. Finally, the writing topic in EWI studies with cancer patients has generally been constrained to their cancer, and it is unclear whether effects of EWI may differ according to the participants´ choice of writing topic. Hypothetically, some individuals may experience writing about their own cancer as threatening and therefore choose to write about more superficial topics or less specific traumas, thereby perhaps reducing the possibility of an effect. It might on the other hand also be beneficial for some patients not to dwell on concerns about their breast cancer, as they may be unable to cope efficiently with highly emotional thoughts and negative emotions within the short time-frame of EWI.

Taken together, there continues to be a need of larger, well-controlled trials evaluating: a) whether there are any relevant and meaningful main effects of EWI on physical health-related outcomes and, in particular, b) if any moderating effects can be found which could inform the selection of patients who could potentially benefit from the intervention. As a part of a large, nationwide EWI intervention (11), we therefore examined the effect of EWI on physical health-related outcomes in a nationwide sample of Danish women treated for early-stage breast cancer. First, we tested the hypotheses that women assigned to EWI would experience improvements in self-reported physical symptoms and reduced health care utilization (self-reported visits and telephone calls to the General Practitioner) compared to controls writing about a neutral topic. Secondly, we explored the possible moderating effects of the following participant characteristics: repressive coping, alexithymia, rumination, social constraints, and choice of writing topic.

Materials and methods

A nationwide, randomized controlled trial was conducted in collaboration with the Danish Breast Cancer Cooperative Group (DBCG) and 14 out of the 16 surgical departments responsible for treating breast cancer in Denmark. Detailed information regarding methods has been described in a previous report of the effects of EWI on psychological health outcomes in the present sample [11].

Participants

Participants considered for eligibility were 1156 women, able to read and write Danish, aged 18–70 years, who had been treated surgically (mastectomy or lumpectomy) over a 6-month period within 3 weeks of a diagnosis for invasive breast cancer stage I and II, and subsequently receiving adjuvant treatment according to Danish national guidelines [35, 36]. As shown in Fig 1, ineligibility and non-response resulted in the randomization of 507 women (age: 27–70 years) to either the EWI and or control (CTRL) group. The study was approved by The Central Denmark Region Committee on Health Research Ethics and the Danish Data Protection Agency.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CONSORT study flow diagram.

https://doi.org/10.1371/journal.pone.0192729.g001

Results

Baseline data and intervention characteristics

No baseline differences between EWI and CTRL were found for the demographic, disease, treatment, outcome or moderator variables analyzed in the present study (See Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics and baseline differences.

https://doi.org/10.1371/journal.pone.0192729.t001

All participants wrote for 3×20 min. Women allocated to the EWI group spent slightly longer time in total on the telephone with research assistants (Mean: 35.3 min/session) than the CTRL group (31.7 min.) (p<0.001). No further between-group differences were found for intervention characteristics, including number of minutes spent on the actual writings and number of days between baseline and intervention, between the first and last writing session, or between the last writing session and completion of follow-up questionnaires. No statistically significant differences were found in the frequency of EWI and CTRL sessions led by the 12 research assistants..

Physical symptoms and health care utilization

As shown in Tables 1 and 2, medium to large and statistically significant effects of time were found for number of GP visits and GP telephone calls, but not physical symptoms. No Time×Group interaction effects reached statistical significance. See descriptive data in Table 1, and results in Table 2A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results from MLMs (baseline, 3- and 9 month follow-up) (Time × Group) of physical symptoms (PHQ), number of GP visits, and number of GP telephone contacts (2a), and results of moderation analyses with repressive coping (2b), social constraints (SCS-C) (2bc), rumination (ECQ) (2d), alexithymia (TAS-20) (2e), and writing topic (own cancer vs. other topic vs. neutral writing) included as moderators.

https://doi.org/10.1371/journal.pone.0192729.t002

Alexithymia as moderator

When exploring total alexithymia scores (TAS-20) as moderator, a statistically significant interaction (Time×Group×TAS-20) of small magnitude (0.17) was found for GP telephone calls, when adjusting for depressive symptoms and age (p = 0.029). Without these control variables, the results only showed a trend (p = 0.052). While alexithymia was not significantly associated with the outcome in the CTRL group, low alexithymic women in the EWI group showed larger decreases in GP telephone calls over time than high alexithymic women and controls (Table 2E and Fig 2). This effect appeared to be driven mainly by difficulties describing feeling (TAS-DDF subscale; p = 0.043) and externally oriented thinking (TAS-EOT subscale, p = 0.085), but not difficulties identifying feelings (TAS-DIF subscale, p = 0.412).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Alexithymia as moderator: Mean number of telephone contacts with general practitioner (GP) (Error bars: 95%CI) from baseline to follow-up of high and low alexithymic women in the EWI and neutral writing control group.

https://doi.org/10.1371/journal.pone.0192729.g002

Writing topic as moderator

Examining writing topic, time and their product as predictors of GP visits revealed a small, but significant, interaction effect (time×writing topic) (p = 0.012). Women writing about their own cancer (z = 2.40;p = 0.017), but not women writing about other topics (z = -0.97;p = 0.333) showed a larger decrease than CTRLs. The same pattern was detected for number of telephone contacts with GP (p = 0.001), where women writing about their own cancer (z = 2.56;p = 0.011), but not women writing about other topics (z = -1.89;p = 0.059), although trending, showed a larger decrease than CTRLs (Table 2F and Figs 3 and 4). The results remained statistically significant when entering age and depressive symptoms at baseline as covariates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Writing topic as moderator: Mean number of visits to general practitioner (GP) (Error bars: 95%CI) from baseline to follow-up of EWI participants writing about their own cancer, EWI participants writing about other topics, and neutral writing controls.

https://doi.org/10.1371/journal.pone.0192729.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Writing topic as moderator: Mean number of telephone contacts with general practitioner (GP) (Error bars: 95%CI) from baseline to follow-up of EWI participants writing about their own cancer, EWI participants writing about other topics, and neutral writing controls.

https://doi.org/10.1371/journal.pone.0192729.g004

Discussion

Our main hypotheses that EWI would lead to reductions in physical symptoms (d = 0.02) and reduced health care utilization in the form of GP visits (d = 0.09) and telephone contacts (d = 0.06) were not supported. Our results are thus concordant with the overall results (g = 0.08) of a meta-analysis of 11 previous EWI studies on physical health-related outcomes, suggesting that EWI is ineffective for cancer patients and survivors [22].

Our results were also in line with the notion that EWI could possibly be relevant for subgroups of cancer patients [22]. A moderating effect was observed was for alexithymia, with low alexithymic EWI participants reporting larger reductions in number of telephone calls to their GP than low alexithymic women in the EWI group and controls. There are several reasons for suspecting a moderating role of alexithymia. One hypothesis states that individuals who have difficulties describing their emotions, or lack the skills and ability to reflect upon- and process their experiences, will benefit less from the emotional disclosure instructions of EWI [26, 27]. On the other hand, it could also be hypothesized that the structured and time-limited writing could be an opportunity for safe processing of traumatic material, giving alexithymic participants sufficient control over the disclosure process to allow themselves to be involved in the process. This latter hypothesis has found support in a study showing EWI to be more beneficial for healthy individuals with high scores on alexithymia [55], however, our findings appear to support the first.

The results also indicated that women who wrote about their cancer showed reduced healthcare utilization in the form of fewer GP contacts when compared to controls, while women who chose to write about other traumatic experiences did not. However, there was a baseline difference between groups both in terms of GP visits (p = 0.015) and phone calls (p = 0.036) with women writing about their own cancer scoring higher. One could speculate that those women who are more active health care users think more about their cancer experience and chose to write about it. For these women, EWI may be particularly helpful. Given the higher baseline level, it is, on the other hand, also possible that the moderation effect detected is an effect of regression toward the mean.

While previous EWI studies with cancer patients have constrained the participants to write about their cancer [22], participants in the majority of the non-cancer studies could select their own writing topics. The breast cancer could theoretically no longer be distressing or even be the most traumatic event for patients who have completed treatment, and when restricting patients to write about their cancer, some may thus be asked to write about a relatively non-traumatic event. To our knowledge, our study is the first to examine whether effects of EWI differ depending on the participants’ choice of writing topic, and we have no clear explanation for our results. It is possible that some individuals may have experienced writing about their own cancer as threatening and therefore chose to write about more superficial topics or less specific traumas, thereby perhaps reducing the possibility of an effect. Previously reported data from the present study [11] showed a similar pattern for psychological outcomes, with women writing about their cancer reporting fewer depressive symptoms and higher levels of positive mood post-intervention than controls, whereas no differences were found for women who wrote about other topics. It should be noted, however, that the interpretation of the results concerning GP visits and telephone contacts depends on the degree to which number of GP contacts can be interpreted as indicative of physical health. In some cases, reduced healthcare utilization could indicate insufficient and delayed response to symptoms, while in other cases, fewer contacts with the healthcare system could indicate a reduction in physiological arousal and reduced tendencies to notice and report physical symptoms.

We found no moderating effects for the remaining factors believed to be associated with emotion regulation: repressive coping style, rumination or social constraints. One comprehensive model of emotional non-expression describes five steps: 1) prereflective action, 2) conscious perception of the emotional response, 3) labeling and interpretation of the response, 4) evaluation the acceptability of the response, and 5) perceived social context for expression [56]. Repressive coping [24] can be viewed as failure at step 2, and repressive copers having difficulties acknowledging and identifying negative emotional experiences may benefit less from written emotional disclosure [26]. Research on EWI and repressive coping is, however, limited [26], and the validity of the available operational definitions of repressive coping style may be questionable [2, 25]. Social constraints, i.e., emotional inhibition due to inadequate opportunities to express emotions, are difficulties at step 5 [32], and are the most frequently researched moderator of EWI in cancer patients and survivors so far, with three studies finding benefits of EWI among participants reporting high social constraints [57–59]. Although the findings are inconsistent [60], social constraints remain a theoretically and empirically promising moderator to be studied further. In contrast to emotional non-expression constructs, rumination refers to excessive emotional expression which has been associated with intensified and prolonged negative emotions [28]. Although studies with healthy participants suggest that EWI buffers against maladaptive rumination [30, 31], rumination did not moderate the effect on physical health-related outcomes in women treated for breast cancer. One possibility is that rumination includes both adaptive and maladaptive factors [61].

The present population-based study is the largest RCT of EWI with cancer patients so far, enabling us to detect small effects. Second, we included a manipulation check revealing statistically significant and large effects (d = 0.84–1.04) in the expected direction. The effects were comparable to previously reported manipulation check outcomes [50, 57], and indicate that the intervention succeeded in influencing the women’s emotional responses. Some potential limitations could, however, have influenced the results. First, 44% of the eligible women actively declined or failed to respond to the invitation to participate. Second, 17% terminated their participation or failed to return follow-up questionnaires. Third, there are some issues concerning writing topic. One concern could be that we distinguished between women who wrote about their own cancer experience in any one of three sessions and women who wrote about another traumatic experience in all three sessions. Another concern is that we do not know the reasons for choosing not to write about cancer. For some women, the cancer might no longer be stressful, while for others the cancer might be experienced as too stressful to write about. As previously reported [11], participants writing about their cancer did not report greater cancer-related distress at baseline than those writing about other topics. Furthermore, lack of randomization may limit the generalizability of differences found between writing topic groups. A three-arm trial with a third group randomized to a group restricted to write about their cancer was considered but abandoned due to statistical power considerations. Finally, the women who agreed to participate could have been physically less affected by their cancer and cancer treatment than non-participants, leading to floor effects blurring effects of the intervention. It should be noted, however, that there were no disease, treatment, or sociodemographic differences between participants and dropouts and no baseline differences between EWI and CTRL participants.

Conclusions

The results from our large randomized trial are concordant with previous findings suggesting that EWI is unlikely to be a generally applicable intervention to improve health-related outcomes in cancer patients and survivors. However, although effects were small, written disclosure could have a beneficial impact for a subgroup of individuals low in alexithymia and who write about their own cancer. Future EWI studies with cancer patients could focus on specifically testing the efficacy of various intervention characteristics, i.e., writing topic, in groups with specific theory-driven characteristics.

Acknowledgments

We would like to thank all participants for spending valuable time on writing and completing questionnaires, the surgical departments for handing out the pre-information, the Danish Breast Cancer Cooperative Group (DBCG) and DBCG chief statistician Susanne Møller for recruitment assistance and database access, and the research assistants without whom this study would not have been possible.