A, Patients’ self-reported predicted and actual satisfaction with breasts (clothed and unclothed), stratified by treatment received; differences are significant between predicted vs actual satisfaction with breasts–unclothed in patients who had a mastectomy only (difference, −0.06; P = .04) and predicted vs actual satisfaction with breasts–unclothed (difference, 0.46; P = .01) in patients who had a mastectomy with reconstruction. B, Patients’ predicted and actual self-perceptions of sexual attractiveness (clothed and unclothed), stratified by treatment received; differences are significant between predicted vs actual sexual attractiveness–clothed (difference, 0.40; P = .03) and predicted vs actual sexual attractiveness–unclothed (difference, 0.95; P < .001) in patients who had a mastectomy with reconstruction.
Patients’ self-reported predicted and actual numbness and pain, stratified by treatment received; differences are significant in predicted vs actual levels of numbness in patients who had mastectomies only (difference, −0.73; P = .001) and in patients who had mastectomies with reconstruction (difference, −0.84; P = .01).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Lee CN, Pignone MP, Deal AM, et al. Accuracy of Predictions of Patients With Breast Cancer of Future Well-being After Immediate Breast Reconstruction. JAMA Surg. 2018;153(4):e176112. doi:10.1001/jamasurg.2017.6112
How well do patients undergoing mastectomy, with or without immediate breast reconstruction, predict their future well-being?
In this cohort study of 96 patients, those undergoing mastectomy without reconstruction underestimated their future quality of life and satisfaction with their clothed breasts, and those undergoing mastectomy with reconstruction overestimated their satisfaction with their unclothed breasts, unclothed sexual attractiveness, and clothed sexual attractiveness. Misprediction was associated with greater regret in both groups.
Patients make mispredictions about their future well-being after mastectomy with or without reconstruction and would benefit from decision support that specifically addresses expectations of future well-being.
Making a good decision about breast reconstruction requires predicting how one would feel after the procedure, but people tend to overestimate the impact of events on future well-being.
To assess how well patients predict future well-being after mastectomy, with or without immediate reconstruction, with the following a priori hypotheses: Patients will overestimate the negative impact of mastectomy and positive impact of reconstruction, and prediction accuracy will be associated with decision satisfaction and decision regret.
Design, Setting, and Participants
This prospective cohort survey study was conducted at a single, multidisciplinary academic oncology clinic from July 2012 to February 2014. Adult women undergoing mastectomy for stage 1, 2, or 3 invasive ductal or lobular breast cancer, ductal carcinoma in situ, or prophylaxis were invited to participate. Data analysis was conducted from September 2015 to October 2017.
Mastectomy only or mastectomy with immediate reconstruction.
Main Outcomes and Measures
Preoperative measures predicted were 12-month happiness (Cantril Ladder) and quality of life, predicted satisfaction with breasts, sexual attractiveness, breast numbness, and pain (measured with BreastQ single items). Measures at 12 months postoperative added the Decision Regret Scale and Satisfaction With Decisions Scale.
Of 214 eligible patients, 182 consecutive patients were approached, and 145 enrolled (80%). Of these 145 patients, 131 returned surveys (72%) and 111 of these remained at 12 months (88%). Fifteen who had delayed reconstruction were excluded from analysis, leaving a final cohort of 96 women; 54 had not had reconstruction and 42 had had reconstruction. The mean (SD) age of the cohort was 53.9 (12.1) years; 73 (76%) were white; 50 (52%) were college graduates; 54 (56%) were privately insured; 69 (72%) had disease at stages 0, 1, or 2; and 31 (32%) received adjuvant radiation. Patients having mastectomy without reconstruction underestimated future well-being in all domains. Differences were significant for quality of life scores (mean predicted, 68 vs mean actual, 74; t50, −2.47; P = .02) and satisfaction with breasts–clothed (mean predicted, 2.4 vs mean actual, 2.8; t49, −2.11; P = .04). Patients undergoing mastectomy with reconstruction overestimated future well-being in all but 1 domain. Differences were significant for satisfaction with breasts–unclothed (mean predicted, 3.1 vs mean actual, 2.6; t41, 2.70; P = .01); sexual attractiveness–clothed (mean predicted, 3.7 vs mean actual, 3.3; t39, 2.29; P = .03); sexual attractiveness–unclothed (mean predicted, 3.3 vs mean actual, 2.3; t40, 5.57; P < .001). Both groups experienced more numbness than predicted (mean predicted, 2.79 and 2.72 for mastectomy only and mastectomy with reconstruction groups, respectively; mean actual, 3.52 and 3.56, respectively; t47, −3.4 and t38, −2.9, respectively; P < .01). Patients who were less happy (β = 6.3; P = .02) or had greater pain (β = 8.7; P < .001) than predicted had greater regret.
Conclusions and Relevance
Patients underestimated future well-being after mastectomy and overestimated well-being after reconstruction. Misprediction was associated with regret. Decision support for breast reconstruction should address expectations about well-being.
Making a good decision requires predicting how one will feel in the future after the decision is implemented.1,2 In the case of surgery, patients needs to estimate how they would feel in the future with or without the surgical procedure. For relatively simple and familiar procedures, predicting how one will feel may be somewhat straightforward. For more complex and unfamiliar procedures, such as breast reconstruction after mastectomy, predicting future well-being may be much more challenging.
Patients considering breast reconstruction face numerous challenges in predicting how they will feel. The predictions involve complex aspects of their subjective and physical well-being because the procedure affects both appearance and physical functioning. Patients may not have heard about the procedure and may not be personally familiar with it. The decision about immediate reconstruction comes at a time of stress and fear related to a breast cancer diagnosis and in the setting of time pressure. These are judgments that approximately 100 000 women with breast cancer face each year.3-5 As more women undergo mastectomy,6-8 how well they make these forecasts becomes increasingly important.
People often mispredict how they will feel in the future. Specifically, they tend to overestimate the magnitude and duration of impact that life events have on their future well-being. Examples exist in our everyday judgments. For example, assistant professors believe that not achieving tenure will result in long-term unhappiness, when, in fact, a matter of months later, those who do not achieve tenure are as happy on average as those who do.9 People also make mispredictions about health care decisions. Patients with kidney failure awaiting organ transplant believe their quality of life will be much greater after transplant than it turns out to be.10 Patients with breast cancer awaiting surgery believe their quality of life will be poorer than what is reported by patients who have actually had the same surgery.11
Some women who have had mastectomy without reconstruction report barriers to access,12-14 while others who have had reconstruction report disappointment or even regret.15-18 Decision regret occurs in 9% to 20% of patients who have had reconstruction, and it is associated with dissatisfaction with preoperative information.17,18 The decision about reconstruction is highly preference-sensitive and therefore patients making this decision would benefit from shared decisionmaking in which they are informed about their options and make choices consistent with their preferences. Making such choices involves predicting how one would feel after surgery.
We sought to assess how well patients undergoing mastectomy forecast their future well-being after surgery. We hypothesized that patients would overestimate the negative impact of mastectomy and the positive impact of immediate reconstruction on well-being. We also hypothesized that prediction accuracy would be associated with satisfaction with decisions and decision regret.
We conducted a prospective cohort study of patients planning to have mastectomy at the Breast Clinic of the North Carolina Cancer Hospital. We have previously reported study findings about preoperative knowledge and preferences.19,20 The clinic is part of an National Cancer Institute–designated Comprehensive Cancer Center and treats approximately 500 patients who have breast cancer in stage 0, 1, 2, or 3 annually. Patients are not typically exposed to a decision aid or other formal decision support.
The University of North Carolina Chapel Hill institutional review board approved the study. The study is registered with clinicaltrials.gov (identifier: NCT01488357). All patients completed a written informed consent procedure before participating in the study.
Adult women (21 years or older) who were planning to have mastectomy were enrolled before surgery from June 2012 to February 2014. We included women undergoing mastectomy for treatment of early-stage (1-3) invasive ductal or lobular carcinoma, treatment of ductal carcinoma in situ, or prophylaxis (eg, patients with BRCA mutation). We excluded patients with stage 4 disease or other histologic diagnoses (eg, sarcoma), because their clinical courses and prognoses differ substantially from those of patients with stages 0 through 3 ductal or lobular carcinoma. We excluded patients with a psychiatric illness who could not make their own decisions and patients who could not read or speak English.
We identified eligible patients by screening clinic schedules and confirmed eligibility with healthcare professionals. The predominant approach to enrollment was in clinic after the surgical oncology visit, with some patients approached by mail. When a patient was undecided about mastectomy at the end of the surgical oncology visit, we followed her course until she had reached a decision, and then approached her for study enrollment. We attempted to approach every eligible patient during the study period.
All surveys were self-administered on paper. After providing informed consent, participants completed the baseline survey in clinic or at home before surgery. If a participant did not return the survey by 2 weeks, we reminded her with a telephone call. If she did not return the survey by 4 weeks and had not yet had surgery, we reminded her by mail. Patients who completed the survey after surgery became ineligible and were removed from the study.
We then mailed follow-up surveys to participants at 6, 12, and 18 months after surgery. Participants received a $50 gift card for completing the first survey and a $25 gift card after completing each follow-up survey.
The following clinical data were abstracted from the medical record on enrollment: age, comorbidities, body mass index (calculated as weight in kilograms divided by height in meters squared), smoking status, cancer stage, and insurance status (none/Medicaid only vs Medicare/Tricare/private). The following additional data were abstracted at 12 months: surgical treatments and dates, oncologic treatments (neoadjuvant chemotherapy, adjuvant chemotherapy, or adjuvant radiation therapy), and complications.
The baseline survey prediction exercise covered 6 domains: happiness, quality of life, satisfaction with breasts, sexual attractiveness, numbness, and pain. For each domain, we asked the participant her current response, her prediction of how she would feel in 12 months if she were to have mastectomy without reconstruction, and her prediction of how she would feel in 12 months if she were to have mastectomy with reconstruction. Measures were based on the Cantril Ladder for happiness,21 and single items from the BreastQ for satisfaction with breasts, sexual attractiveness, numbness, and pain. For quality of life, we used an ad hoc question with a numeric rating scale of 0 to 100, which we have used in previous studies.10 The baseline survey also included demographic questions, the Decision Quality Instrument,22,23 the BreastQ,24 and a preference elicitation exercise. Results for these other measures are being reported separately.20 The 12-month survey repeated the measures for the same 6 domains and included the Satisfaction With Decisions and Decision Regret Scales.25,26
We compared the demographic characteristics of the mastectomy without reconstruction group with the characteristics of the mastectomy with reconstruction group, using 2-tailed t tests for continuous variables and Fisher exact tests for categorical variables. We excluded from analyses those patients who underwent delayed reconstruction within 12 months, so that duration of recovery would be similar at the 12-month survey. We excluded those who did not complete the 12-month survey. An α of .05 was considered significant. Statistical analysis was completed from September 2015 to October 2017, with SAS version 9.4 (SAS Institute).
For each of the 6 prediction domains (happiness, overall quality of life, satisfaction with breasts, sexual attractiveness, numbness, and pain), we compared predicted scores with actual 12-month scores, using paired t tests. We did this within each treatment group (the mastectomy without reconstruction group vs the mastectomy with reconstruction group) separately. We then compared the 2 treatment groups, using a 2-group t test.
We then divided prediction accuracy into 3 categories: actual outcome was worse than predicted, actual outcome was as predicted, and actual outcome was better than predicted. The thresholds for being as predicted were defined by instrument, with answers within 1 point of 10 for happiness, 5 points of 100 for quality of life, and 0 points of 4 for satisfaction with breasts, sexual attractiveness, numbness, and pain considered as predicted. We conducted a sensitivity analysis of this threshold and found almost no change in the results with less stringent and more stringent thresholds.
To analyze the association between prediction accuracy and satisfaction with decisions, we fit a multivariable model with satisfaction with decisions as the outcome and 1 prediction accuracy domain as the independent variable of interest (using the dichotomous prediction accuracy variable), adjusting for treatment (mastectomy without reconstruction vs mastectomy with reconstruction), major complications, race/ethnicity, education level, income level, and insurance status. We included independent variables that were associated with the outcome on bivariable analysis at a level of P < .10. We repeated these models with decision regret as the outcome, fitting a separate model for each prediction domain.
During the 20-month study period, 214 patients were eligible. We approached 182 patients and missed the opportunity to approach 32 patients (including 5 who were not contacted at the requests of surgeons and 27 who left the clinic before we could ask them to participate). Of the 182 approached, 145 enrolled (79.7% enrollment), and 131 completed the survey (72.0% participation). Five participants became ineligible for various reasons (eg, had surgery at a different facility or completed surveys after surgery), leaving a baseline study population of 126 persons. At 12 months, 111 of the original 126 patients completed the 12-month survey (88.1% retention). Of the 111, 15 (13.5%) had delayed breast reconstruction and were excluded from analyses. The final 12-month study population consisted of 96 women, of whom 54 (56%) had had mastectomy without reconstruction and 42 (44%) had had mastectomy with reconstruction.
The mean (SD) age was 53.9 (12.1) years, and 73 participants (78%) were white, with a further 20 (24%) reporting black as their race/ethnicity (Table 1). About half (n = 49; 52%) of the participants were college graduates, 60 (63%) were partnered, 44 (47%) were not working, and 54 (56%) were privately insured. Most (n = 69; 72%) had stage 0, 1, or 2 disease, and 31 (32%) had received adjuvant radiation therapy. The immediate reconstruction rate was 40%, which was consistent with national norms.12,14,27 The 37 eligible patients who declined study enrollment were similar to participants in age (mean, 53 years) and insurance status (34; 92% insured), but more were nonwhite than the study-enrolled group (17/37, or 46%, vs 20/96, or 24%; P = .01), and fewer had reconstruction than the study-enrolled patients (8, or 22%, vs 42, or 44%; P = .05).
In general, patients undergoing mastectomy without reconstruction (n = 54) underestimated how well they would feel at 12 months (Table 2). For quality of life and satisfaction with breasts, these differences were statistically significant. Patients in this group predicted, on average, a quality of life mean score of 68 on a scale of 100 but experienced a mean score of 74, which was significantly higher (t50, −2.47; P = .02). For satisfaction with breasts–clothed, these patients predicted, on average, a mean score of 2.4 but experienced a mean score of 2.8 (t49, −2.11; P = .04). Of note, patients who had mastectomy without reconstruction predicted poorer future well-being compared with their preoperative state for each domain.
Patients undergoing mastectomy with reconstruction generally overestimated how well they would feel at 12 months, and these overestimates were statistically significant in 3 domains: satisfaction with breasts–unclothed, sexual attractiveness–clothed, and sexual attractiveness–unclothed (Table 2). For satisfaction with breasts–unclothed, patients predicted a mean score of 3.1, but experienced a mean score of 2.6 (t41, 2.70; P = .01). For sexual attractiveness–clothed, they predicted a mean score of 3.7, but experienced a mean score of 3.3 (t39, 2.29; P = .03). For sexual attractiveness–unclothed, they predicted a mean score of 3.3, but experienced a mean actual score of 2.3 (t40, 5.57; P < .001). In contrast with patients who had mastectomy without reconstruction, patients who had mastectomy with reconstruction predicted better or equivalent well-being compared with their preoperative state for every domain except numbness and pain.
We separately evaluated patients who did not have malignancy (n = 12/13 with reconstruction) and found similar overestimations of well-being. Because the results were similar, we are not reporting them separately.
Figure 1 and Figure 2 show these differences graphically and display the frequency of having outcomes different than predicted. Consistently, a higher percentage of patients who had mastectomy with reconstruction had more severe symptoms than they had predicted compared with patients who had mastectomy without reconstruction. Both treatment groups (mastectomy without reconstruction and mastectomy with reconstruction) predicted an increase in numbness (to mean [SD] scores of 2.8 [1.2] and 2.7 [1.4] on a scale of 5, an increase of 1.3 and 1.4, respectively), but both experienced even more numbness at 12 months than predicted (mean [SD] scores, 3.5 [1.2] and 3.6 [1.4], respectively; t47, −3.4 and t38,−2.9 respectively; P < .01 in both comparisons) (Figure 2). Both groups also predicted an increase in the frequency of pain, but those having mastectomy without reconstruction experienced even more pain than they had predicted (mean [SD], 2.1 [0.9] predicted vs 2.5 [1.15] actual; t49, 2.02; P = .05).
In general, satisfaction with decisions was relatively high, at a mean (SD) of 4.3 (0.81) on a scale from 0 to 5, and this did not differ based on prediction accuracy. Table 3 shows the unadjusted means for each prediction accuracy group. It also shows the adjusted mean difference between groups after controlling for treatment (mastectomy only vs mastectomy with reconstruction), complications, race/ethnicity, education, income, and insurance. In these multivariable models, prediction accuracy in any domain was not associated with satisfaction with decisions (Table 3).
In general, decision regret was moderate, with a mean (SD) of 41.9 (9.13) on a scale from 0 to 100. In multivariable models of decision regret, misprediction about happiness and misprediction about pain were associated with greater regret (Table 3). Specifically, patients who were less happy than predicted had greater regret (β = 6.3; P = .02), and patients who had greater pain than predicted had greater regret (β = 8.7; P < .0001).
Patients undergoing mastectomy with or without breast reconstruction made mispredictions about their future well-being after surgery. Those having mastectomy without reconstruction generally underestimated their future well-being, and those having mastectomy with immediate reconstruction generally overestimated it. Misprediction was associated with greater regret but not with satisfaction with decisions.
Our finding that patients overestimated the negative impact of mastectomy and positive impact of immediate reconstruction is consistent with prior psychological28 and clinical10,11 studies. People have a strong innate ability to adapt to adverse circumstances, but they tend to be unaware of that ability and to neglect it in their judgments.9,29,30 This tendency is particularly strong in the face of unfamiliar adverse events such as major surgery. Patients with breast cancer facing mastectomy may not realize that they will likely adapt somewhat to the loss of their breast.31-33 We were struck by the finding that participants undergoing immediate reconstruction believed that, on average, their well-being would improve beyond their preoperative state. We hypothesize that they may not have considered fully some negative aspects of reconstruction, such as scarring, pain, or multiple operations.20
Another reason people mispredict their future well-being is a tendency to focus on the issue at hand and not on other contributors to well-being.28,34 For example, when Midwesterners are asked to estimate how happy they would be in Southern California, they typically focus on climate and predict being much happier there. Because their attention is so narrowly focused on weather, they ignore other important lifestyle differences, such as traffic and cost of living.35 The decision-making process for patients with breast cancer may contribute to this tendency to focus. Decisions take place in a short time and often in the context of strong negative emotions. Surgical visits may emphasize changes that occur because of surgery and not emphasize other aspects of patients’ lives that will be stable (eg, their family situation) or that will change for unrelated reasons (eg, chemotherapy effects). We speculate that delayed reconstruction and bilateral prophylactic mastectomy decisions may be less prone to such hyperfocusing because they allow more time. Also, patients deciding about bilateral prophylactic mastectomy may experience surgical decisions differently because they do not need lymph node biopsy or adjuvant treatments, and they may have observed relatives having breast cancer. We did not specifically assess the role of delayed reconstruction or prophylactic mastectomy because of the limitations in our sample.
How clinically significant were the mispredictions that we found? Patients undergoing mastectomy without reconstruction underestimated their future quality of life by 6 points (on a scale of 100), and satisfaction with breasts clothed by 0.4 points (on a scale of 4). These are relatively small differences, but they may have been large enough to influence decisions. The mispredictions by patients choosing mastectomy with reconstruction were larger, including a 1.0 point overestimation for sexual attractiveness unclothed (on a scale of 5 points). Importantly, patients undergoing mastectomy without reconstruction anticipated a decline in well-being compared with their baseline scores in every domain, and patients undergoing mastectomy with reconstruction anticipated an improvement in nearly every domain. We were surprised by patients’ apparent belief that mastectomy with reconstruction would improve appearance over their baseline levels. We speculate that some patients may confuse aesthetic plastic surgery (ie, breast augmentation) with reconstructive plastic surgery.
Patients’ expectations of appearance and attractiveness unclothed may be a particularly important aspect of decisions about reconstruction. The unclothed domains had the largest prediction errors by patients who had mastectomy with reconstruction. These particular errors may reflect a failure to distinguish between clothed and unclothed outcomes. We believe that breast surgeons and plastic surgeons may discuss outcomes without specifically making such distinctions. We also note that media portrayals of women who have had reconstruction generally show positive, clothed photographs.36,37 Surgeons should explicitly address the distinction between clothed and unclothed outcomes when informing patients about the pros and cons of procedures and eliciting patients’ preferences.
Expectations about numbness appeared to be an issue for both groups of patients. Although patients generally predicted they would have numbness, they underestimated how much. Healthcare professionals should discuss breast sensation with patients, and decision aids should address the topic. Patients’ personal preferences about numbness could have important implications for decisions about lumpectomy vs mastectomy and about contralateral prophylactic mastectomy, since these procedures differ substantially in terms of numbness.
Patients who were less happy than they predicted, or who had greater pain than predicted, experienced greater regret. These findings point to the importance of helping patients have realistic expectations before surgery. It also suggests that pain management should be a priority for improving patient reported outcomes after mastectomy. We are uncertain why mispredictions were not associated with satisfaction with decisions. Satisfaction with decisions was generally high and did not vary much, consistent with prior studies,14,38 and our sample size was somewhat limited at detecting small differences. It is possible that breast cancer patients’ feelings about their treatment decisions 1 year later may be affected more by their oncologic outcome than the quality of the decision making process.
The finding of mispredictions has important implications for clinical care and decision support. Surgeons should discuss adaptation during the longer-term survivorship phase and other aspects of patients’ lives besides their breasts, including how those aspects will change or not. Decision support should also incorporate techniques that reduce prediction errors, such as patient narratives addressing adaptation,41 writing about other aspects of one’s life,28 or using virtual reality simulation or personal avatars.42 Patients may benefit from talking to survivors who have experienced adaptation and could describe their experience with it. They may also benefit from a decision-making process that allows for more time, which could discourage excessive focus on breasts and allow consideration of other aspects of well-being.
Our study has some methodological limitations. The sample was taken from 1 academic institution and did not include eligible patients who were not approached, chose not to participate, or did not complete 12-month surveys. We asked patients to predict well-being at 1 year, but we do not know which time frame is most important to decisions. Patients who have mastectomy without reconstruction probably continue to adapt after 1 year, so predictions about later points may have even larger errors. It is also possible that patients who had mastectomy with immediate reconstruction experienced changes in well-being beyond 1 year, since outcomes of reconstruction can change over time.39,40 Because of the small sample size, we did not assess how reconstruction type or eligibility affected prediction accuracy. Reconstruction type could be relevant, since implant outcomes are less stable over time than autologous outcomes.39 Reconstruction eligibility could also be relevant, because some patients who were not offered reconstruction may have predicted poorer well-being without reconstruction than if they had had a choice. We also did not account for individual clinician effects on prediction accuracy. Finally, most measures were single items adapted from the BreastQ and have not been validated for use in predictions.
Patients undergoing mastectomy, with or without reconstruction, mispredicted their future well-being after surgery. Those having mastectomy without reconstruction generally underestimated future well-being, and those having mastectomy with reconstruction generally overestimated it. Misprediction was associated with greater regret but not with satisfaction with decisions. Decision support for mastectomy and breast reconstruction should address expectations about future well-being.
Corresponding Author: Clara Nan-hi Lee, MD, MPP, Department of Plastic Surgery, The Ohio State University, 915 Olentangy River Rd, Ste 2100, Columbus, OH 43212 (email@example.com).
Accepted for Publication: November 18, 2017.
Published Online: February 7, 2018. doi:10.1001/jamasurg.2017.6112
Author Contributions: Dr Lee had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lee, Pignone, Deal, Ubel.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Lee, Deal, Blizard, Hunt, Huh, Liu.
Critical revision of the manuscript for important intellectual content: Lee, Pignone, Deal, Hunt, Ubel.
Statistical analysis: Deal, Huh, Liu.
Obtained funding: Lee.
Administrative, technical, or material support: Hunt, Liu.
Study supervision: Lee.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by National Institutes of Health/National Cancer Institute Cancer Prevention, Control, Behavioral, and Population Sciences Career Development Award (grant K07CA154850-01A1, Dr Lee); the Lineberger Comprehensive Cancer Center Population Sciences Cancer Research Award (Drs Lee, Pignone, and Ubel, and Ms Deal); and the North Carolina Translational and Clinical Sciences Institute Pilot Award (Drs Lee, Pignone, and Ubel, and Ms Deal).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.