Assessing Preparatory Grief in Advanced Cancer Patients as an Independent Predictor of Distress in an American Population

Background

There are many different concepts of grief that tend to overlap, such as acute grief, anticipatory grief, complicated grief, preparatory grief, and normal grief. Despite the presence of these various concepts, the term grief has generally come to be understood as an emotional response following any form of loss. ¹ A more comprehensive definition describes grief as the “psychological, behavioral, social, and physical reactions to the loss of someone or something that is closely tied to a person's identity.” ² Lindemann found five aspects that seem to be characteristic of the grief process, including (1) somatic distress, (2) a preoccupation with the image of the deceased or lost entity, (3) guilt, (4) hostility, and (5) a loss of previously consistent behaviors. ³

In addition to the discovery of these five common aspects of grief, Lindemann began to identify grief reactions due to separation as opposed to death. More specifically, Lindemann observed that when an individual was separated from family due to commitment to the armed forces, symptoms consistent with grief were often observed among the family members. Under these circumstances, the grief was not a response to death itself, but rather to the threat of death. Lindemann termed this reaction anticipatory grief. Observations revealed that the family or individual family members became so concerned with the possibility of having to manage life without the loved one that they ultimately experienced all of the identified symptoms of grief. ³

Since the term anticipatory grief usually refers to the grief process experienced by friends and family members before the death of a loved one, the term preparatory grief was created to refer to grief experienced by those individuals facing their own impending death. ⁴ Preparatory grief is the grief “that the terminally ill patient has to undergo to prepare himself for his final separation from this world” ⁵ and is a universal experience for patients living with a terminal cancer. ⁶ The challenge of working through preparatory grief is not just the eventual loss of life, but that the patient is also facing many losses along the way. The patient grieves for the loss of certain physical and cognitive abilities, in addition to the loss of one's overall independence. Other common grieving experiences include the loss of privacy, dignity, financial control, autonomy, and the loss of one's prior role in the family. ⁷ Finally, the patient is not just preparing for the loss of her/his own self, but is also preparing for the loss of everybody else she/he has ever loved. To be successful in this task, the patient must begin to separate oneself from those that she/he will be forced to leave behind in the near future. ⁸ It is this potentially devastating process of separating oneself from all that is known and loved and comfortable that makes the need for a specific measure of preparatory grief clinically relevant. With such a measure, these painful grief experiences can be identified and interventions can be implemented to help the patient comfortably journey through this final stage of life.

Only a few instruments are specifically designed to measure the experience of preparatory grief. The Inventory of Complicated Grief-Revised (ICG-R) was designed initially for measuring complicated grief of caregivers after the death of the patient, but was modified to measure preparatory grief in patients with a terminal cancer diagnosis. Preparatory grief and depression were found to be distinct clinical entities. In a factor analysis of 123 terminal cancer patients using the modified ICG-R to detect preparatory grief and the Structured Clinical Interview for DSM-IV (SCID) to detect major depressive disorder (MDD), two-thirds of patients with grief would have been missed using criteria for MDD alone. Preparatory grief was found to be associated with an increased risk for wish to die, mental health service use, and negative religious coping, but not with terminal illness acknowledgement. ⁹ The modified ICG-R for preparatory grief has not been validated in a population of advanced cancer patients.

The Terminally III Grief or Depression Scale (TIGDS) was developed with specific preparatory grief and depression subscales to distinguish between these two clinical entities. Using a small sample (n = 44), the TIGDS was successful in identifying and distinguishing preparatory grief in a population of advanced cancer patients with good internal consistency (intraclass correlation coefficient for test-retest reliability was 0.86 for the preparatory grief scores, 0.97 for the depression scores, and 0.92 for the grief versus depression scores). The depression subscale of TIGDS had acceptable convergent validity with the Hospital Anxiety and Depression Scale (HADS) total scores as well as significant discriminant validity between the HADS total scores and preparatory grief subscale of the TIGDS. ¹⁰ The small sample size and the homogenous population (single inpatient palliative care unit in a VA hospital) make generalizability of the TIGDS challenging.

The Preparatory Grief in Advanced Cancer (PGAC) instrument was designed to measure the experience of preparatory grief and has the most robust validation support to date for this purpose.^7,10 It contains 31 items and results in a 93-point measure of preparatory grief. The PGAC is meant to serve as an overall rating of preparatory grief, and, as such, the higher the score, the more preparatory grief the patient is presumed to be experiencing. ¹¹ Furthermore, this measurement does not aim to distinguish “normal” and “pathological” preparatory grief. Rather, it is meant to serve as a clinical tool that aids healthcare professionals in identifying those patients in need of more specialized and supportive treatment. ¹²

The instrument was developed and first administered to a sample of 200 advanced and terminally ill cancer patients in Greece. A factor analysis revealed seven factors: self-consciousness, disease adjustment, sadness, anger, religious comfort, somatic symptoms, and perceived social support. ⁷ Cronbach's alpha coefficient for reliability was 0.838 and the test-retest reliability was satisfactory for all subscales.

In a Greek population of advanced cancer patients, higher scores on the PGAC instrument were shown to have significant associations with hopelessness, anxiety, and depression.^7,13 Another study from the same institution examining demographic and clinical predictors of high PGAC scores found that patients with higher PGAC scores tended to be female, be younger, have undergone any surgery, be on strong opioids, and have a poorer performance status. ⁶ Despite the PGAC instrument being the most robustly validated tool for measuring preparatory grief, its use only in a Greek population to date has limited its generalizability.

Objective(s)

Our aim was to use the PGAC instrument in an American population of advanced cancer patients to explore demographic, clinical, and psychological factors associated with higher preparatory grief in this American population.

Design

Subjects were recruited by weekly screening of the inpatient oncology list and discussion with the inpatient oncology team regarding eligible patients as well as weekly discussions with outpatient oncologists regarding eligible patients being seen that day in the clinic. Subjects were approached after assent was obtained by their treating oncology team and a research assistant obtained consent. Finally, subjects completed a single cross-sectional time point evaluation after consent was obtained.

Descriptive statistics were used for characteristic variables, including sex, location, prognostic awareness, type of cancer, elevated anxiety, elevated depression, elevated grief, elevated distress, and elevated overall psychological morbidity (Table 1). Patients who self-reported they had incurable disease were considered to have prognostic awareness. In the statistical analysis, distress thermometer (DT), HADS-Anxiety subscale (HADS-A), HADS-Depression subscale (HADS-D), Edmonton Symptom Assessment Scale-Physical Subscale (ESAS-P), Edmonton Symptom Assessment Scale-Total (ESAS-T), current quality of life (QOL), and satisfaction with current QOL were continuous variables. Location, sex, and prognostic awareness were treated as dichotomous variables. Type of cancer was converted to a binary variable by lumping all cancer types together other than gastrointestinal cancers, since this was the most common type of cancer.

First, univariate analysis was conducted to assess which factors were associated with PGAC scores, including DT, ESAS-P, ESAS-T, HADS-A, HADS-D, current QOL, satisfaction with current QOL, location, sex, type of cancer, and prognostic awareness. Second, a multivariate logistic regression model was used to study the relationship between the PGAC score and other variables that met statistical significance in the univariate analysis.

Setting/Subjects

This study was approved by the IRB. Outpatients seen at the Robert H. Lurie Comprehensive Cancer Center or inpatients admitted to the Northwestern Memorial/Prentice Hospitals in Chicago, IL, were screened and enrolled between October 28, 2011 and March 9, 2012. Inclusion criteria included the following: incurable solid malignancy as judged by the treating oncologist, self-report of U.S. citizenship, ≥18 years old, and willing to sign consent. Subjects with a self-reported history of substance abuse or current psychiatric diagnosis were excluded.

Measurements

Directly after consent was obtained, the subject filled out demographic information, including sex, cancer type, location of care (inpatient vs. outpatient), and status of cancer as “incurable” [“Have you received a diagnosis of incurable cancer?”]. Finally, they completed the PGAC instrument,^7,13,14 the HADS,^15,16 the DT, ¹⁷ the ESAS, ¹⁸ and a QOL measure [rate current QOL on a 0–10 Likert scale as well as satisfaction with the current QOL on a similar scale]. ¹⁹ All patients received written resources for emotional support at the conclusion of the study, and any patient scoring >40 on the PGAC or >14 on the HADS total was flagged and his/her treating oncology team was informed immediately.^14,20

Results

Fifty-three patients were surveyed, of which 57% (30) of patients were outpatients and represented a broad assortment of patients with solid malignancies (Table 1). About 30% of patients included still felt their cancer was curable, despite their oncologist's opinion that it was not. Based on recommended cutoff scores for anxiety, depression, preparatory grief, and distress,^14,16,17 at least one clinically significant value was found in 57% of this sample (28% with one, 9% with two, 9% with three, and 11% with all four). The mean PGAC score was 26.9 out of 93 possible points (range 12.0–41.4 points).

In univariate analysis (Table 2), PGAC scores were significantly correlated with DT, HADS-A, HADS-D, ESAS-P, ESAS-T, and satisfaction with QOL. Multivariate analysis revealed only DT to be statistically correlated with the PGAC score (beta coefficient = 3.2, p = 0.001), meaning every one-unit increase in the DT score will result in a 3.2-unit increase in the PGAC score, while controlling for HADS-A, HADS-D, ESAS-P, ESAS-T, and Satisfaction with QOL in the model (Table 3). The results were consistent with this when clinically significant cutoffs (i.e., DT ≥4, HADS-A ≥7, HADS-D ≥7, or HADS-T ≥14) were used in data analysis (results not shown).

Discussion

Preparatory grief was a common experience in this sample of American advanced cancer patients with one in four reporting suffering from clinically significant levels of grief. Although univariate analysis revealed correlation with distress, anxiety, depression, physical symptom burden, overall symptom burden, and satisfaction with QOL, only distress was statistically correlated in the multivariate analysis.

It is of interest that preparatory grief, as measured by the PGAC instrument in our study, did not reveal multivariate associations with previously reported factors, including anxiety, depression, gender, and poor performance status.^6,13 Gender has mixed results in regard to PGAC, therefore, this study adds to the evidence for a lack of correlation. Although we did not directly measure performance status, we did quantify physical and total symptom burden, which follows an overall similar trajectory to performance status. ²¹ Physical and total symptom burden did show a correlation in univariate analysis, but this did not persist in the multivariate analysis in our population. Finally, it has been shown that the PGAC instrument has concurrent validity with HADS-A and HADS-D, ⁷ with both serving as independent predictors of higher PGAC scores in a separate multivariate analysis. ¹³ Our study confirmed a correlation with anxiety and depression, but this relationship did not hold in the multivariate analysis. In comparison with the Greek population the PGAC was originally studied with, our population seemed to have less grief (mean PGAC score 26.7, range 12–41.4 vs. mean 44.5, range 14–69), ¹⁴ similar depression (HADS-D ≥7; 33% vs. 32%), and less anxiety (HADS-A ≥7; 26% vs. 36%). ¹³ Although studies are hard to compare, it does not seem likely that these differences were driving the lack of correlation observed between HADS and the PGAC instrument.

Although our smaller sample size is a possible contributor to the lack of above correlations, it is reasonable to consider that the inclusion of the DT in our study may have altered the results of the multivariate analysis. Removing DT from the multivariate analysis caused HADS-D to become significantly associated with PGAC, partially supporting this argument (results not shown). Elevated levels of distress (i.e., ≥4), as measured by DT, are known to be associated with many factors, including female gender, poor performance status, as well as emotional and physical problems. ²² In addition, the DT is often used as a screening tool for detecting clinically significant anxiety and depression, but it also screens for psychological morbidity due to other sources such as poor symptom control and financial or family strain. ²³ Given no previous study has investigated the link with the DT and PGAC instrument and our study revealed a strong correlation, it may be that the DT tool is such a general measure of distress that it represents the common variance shared by anxiety, depression, symptom burden, and satisfaction with QOL, thereby leaving little unique variance for these factors in association with the PGAC instrument.

Increased levels of preparatory grief have been reported as a maladaptive response to loss,^13,24 and our study indicates that a commonly used screening tool for psychological distress (DT) may be a strong predictor for maladaptive grief in patients with advanced cancer. In addition, we know that a maladaptive emotional response to a terminal illness is a strong predictor of aggressive care received at the end of life. ²⁵ If this is the case, then grief-based interventions used by palliative care teams may start to explain some of the mechanism of action seen with early palliative care interventions. ²⁶ Clearly, further preparatory grief research is needed to better understand the role of maladaptive grief in end-of-life care.

Our findings are limited by the small sample size that can impact the ability to properly address confounding with multivariate analysis. It would benefit from replication in a larger cancer patient population. In addition, our cross-sectional design limits the interpretation of noted correlations and any directional nature to the correlations. We included a large and diverse cancer population and may not have been able to capture the unique experience that occurs with grief in each type of cancer. Finally, the generalizability of our data may be limited due to enrollment at a single institution, the urban setting of the institution, and the chance of nonresponse bias that was not captured before enrollment.

Conclusion

In our American population of advanced cancer patients, the PGAC instrument was able to quantify the amount of preparatory grief experienced by a patient. Despite measures of demographic, clinical, and psychologic morbidity, the only significant predictor of higher PGAC scores was higher DT scores. Further research is needed to understand how preparatory grief changes over the trajectory of an illness, how it relates to patient coping and resiliency, and its impact on death acceptance and healthcare delivery in the last year of life.