Introduction

Cancer is the leading cause of death by disease in the adolescent population (persons 10–21 years of age) [1, 2]. Although great strides have been made in the treatment of cancer, 25% of adolescents with cancer will ultimately die from their disease [2]. In 2004, there were 2,223 pediatric cancer deaths with 1,224 occurring in adolescents 10 to 19 years of age [3]. The primary goal of treatment when cancer is diagnosed is almost always cure. When cure is no longer a realistic possibility, health care providers (HCP) have a difficult time with the psychosocial, emotional, ethical, and existential issues related to an incurable disease that results in death.

Quality of life at end of life (EOL) though extremely important has been neglected in the research literature for adolescents with cancer [4–6]. Although adolescents and parents both indicate a preference for dying at home, most adolescent cancer deaths occur in a hospital setting [7–9]. End-of-life factors such as place of death, medications used to control EOL symptoms, and whether or not there was adequate time for communication between the adolescent and family or health care providers prior to death warrant further exploration.

Adolescents face unique developmental challenges that require special consideration particularly when death from disease is likely. Additionally, there are great developmental differences between younger and older adolescents facing death [10, 11]. The purpose of this study was to explore the experiences of adolescents dying from cancer, including place of death, medications used at EOL, and initiation of EOL discussions between health professionals and adolescents during early, middle, and late adolescence and their families. Additionally, we explored relationships among demographic and EOL variables to inform future studies.

The following research questions were used to explore experiences of adolescents at EOL including place of death, medications used, and EOL discussions:

  1. 1.

    What are the characteristics of adolescent cancer death, including location of death, symptom control measures used, and time for preparation for death?

  2. 2.

    Do gender, symptom control measures, EOL discussions, place of death, or cause of death differ by cancer diagnosis?

  3. 3.

    Do gender, symptom control measures, EOL discussions, or place of death differ by developmental level of the adolescent?

  4. 4.

    Is place of death influenced by the health professional who conducts the initial EOL discussion?

Methods

Data

This study was approved by the Institutional Review Board. A secondary analysis was completed of data from a larger pediatric EOL study. Quantitative data were collected through a manual chart review and entered into a Microsoft Excel spreadsheet. A predetermined list of variables was identified pertaining to adolescent cancer-related deaths. Data were obtained from each chart to obtain the birth date, date of diagnosis, and date of death. Exact dates for additional variables were collected as applicable (i.e., second neoplasm, 1st, 2nd, 3rd, or 4th relapse, stem cell transplant (SCT), initial EOL discussion, and do not resuscitate (DNR) order). Time lapse between data points was computed by subtracting one time point from another. For example, “time from diagnosis to death” was obtained by subtracting date of diagnosis from date of death. Microsoft Excel Program was used for all calculations providing the exact number of days, months, or years for analyses. Data were collected in five categories: (1) demographics, (2) cancer trajectory, (3) symptom control measures, (4) significant time points on cancer continuum, and (5) cancer death. A comprehensive list of variables and how they were computed are presented in Table 1.

Table 1 Comprehensive variable list
Full size table

Data for initial EOL discussion were obtained from multiple sources, including the hematology/oncology clinic shadow chart, the electronic medical record from all hospitalizations, the outpatient clinic nursing phone logs, and the social work files on each patient. Timing of discussions surrounding EOL care was defined as the first date that documentation about communication between a physician and patient/family occurred. Although documentation does not always accurately reflect what is happening clinically, it is the only measure by which to establish the beginning of what is ultimately a series of conversations that happen throughout the duration of the illness.

Sometimes documentation about EOL discussions was recorded in the social work documentation or nurse’s notes on the date that the family conference took place because these disciplines were present for the discussion between physician and family. The physician might simply add a progress note that indicated discontinuation of protocol with a focus on symptom control. Indicators in the charting systems that were used to define an initial “date of EOL conversation” were mainly (a) documentation by physicians regarding a shift in the goals of care from curative to palliative/EOL, (b) documentation by physicians, nurses, or social workers surrounding family conferences to address relapse or progression of disease and subsequent need to reconsider treatment goals, (c) documentation surrounding change in code status, (d) documentation in social work files regarding physician–patient disagreement regarding code status or goals of care (i.e., palliative/EOL versus curative), and (e) enrollment in hospice.

The documentation surrounding EOL care is not protocol-driven and is largely dependent on the documentation skills of an individual physician. Similar to documentation for billing and coding, there is much variability. Health care professionals are taught with increasing frequency in medicine that “if it isn’t documented, it didn’t happen”. While it is the goal of every practitioner to accurately document what is happening clinically, this documentation is not always achieved. Charts that were missing documentation of EOL discussions reflect possibly two scenarios: (1) conversations about death never occurred or (2) conversations occurred but were never documented.

Sample

The sample included data from 107 adolescents between 10 and 21 years old, treated at a Midwestern tertiary care center from 1982 to 2004 and subsequently died between 2000 and 2005. Four cases were removed from analyses because either cause of death was not documented and occurred several years after diagnosis, or death was accidental and therefore not cancer-related, leaving a final sample of 103 adolescents. The majority of the sample, N = 70, were between the ages of 10 and 21 at diagnosis and died between the ages of 10 and 21. The remaining 33 individuals were under the age of 10 at diagnosis but were between 10 and 21 years old at death. Characteristics of the adolescent sample at death are recorded in Table 2.

Table 2 Adolescent sample at death
Full size table

Results

Description of adolescent cancer death

The majority of adolescents (N = 58) died in the hospital; only 16 adolescents died at home. Place of death was missing for 29 adolescents. Nearly half (N = 24) of the hospital deaths occurred in ICU. Adolescents who died in ICU were more likely to have a treatment-related cause of death (p < 0.001).

Most adolescents died of disease progression (77%), with only 20% having a treatment-related death. Examples of a treatment-related death include sepsis, toxicity from treatment, tumor lysis syndrome, or poststem cell transplant complications such as graft versus host disease. In 3% of the sample (N = 3), cause of death was unknown or not documented; however, circumstances indicated that a cancer-related death was highly probable. These individuals were included in the analyses and cause of death was coded as system missing. Ten percent of the population experienced a second primary malignancy before death. Radiation and chemotherapeutic agents have been linked to secondary neoplasms years after successful treatment [12].

The majority of adolescents (86.6%) used pain medication at EOL, nearly half (46.3%) used an anti-emetic, and more than a third (38.8%) used an anxiolytic. Differences in use of symptom control measures among diagnostic groups and developmental groups are reported in Table 3 and in more detail in the sections below.

Table 3 Symptom control measures used at end of life
Full size table

Differences among diagnostic groups at death

Adolescent males died of leukemia/lymphoma or solid tumors at a higher rate than females (57.5% and 64%, respectively) and died at a lower rate (33%) from central nervous system (CNS)/brain tumors (p = 0.052). Aggressive life-sustaining measures such as using ventilators, pressors, or advanced cardiac life support (ACLS) code drugs, occurred in 12% of the adolescents at EOL. However, the majority of adolescents dying with aggressive life-sustaining measures were diagnosed with leukemia/lymphoma (70%). The timing of initial EOL discussions varied significantly across diagnostic groups (see Table 4). Adolescents with leukemia/lymphoma were more likely to have initial conversations about EOL in the last 7 days of life (p = 0.002), leaving very little time to prepare for death. Finally, place of death (home versus hospital) was not different across diagnostic groups; however, location of hospital death, that is, ICU versus cancer/general ward, was significantly different. More adolescents diagnosed with leukemia/lymphoma died in the ICU (p = 0.028).

Table 4 Timing of initial EOL discussions
Full size table

Symptom management techniques differed by cancer type. Adolescents dying of leukemia/lymphoma used more oxygen than adolescents dying of CNS/brain tumors (p = 0.054). Adolescents dying of CNS/brain tumors used corticosteroid or anticonvulsant medications more than adolescents dying of a solid tumor or leukemia/lymphoma (p < 0.001).

Differences across developmental age groups at death

Adolescence is a dynamic period of rapid physical, cognitive, psychosocial, and emotional development. Little research has been done to determine how stage of development affects the dying process during adolescence. This sample was divided into three developmental age groups consistent with Strasburger et al.: early adolescence (10–13 years old and generally attending middle school), middle adolescence (14–17 years old and generally attending high school), and late adolescence (18–21 years old and generally completing high school and transitioning to adulthood) [11]. At time of death, 42 persons in our sample (41%) were in early adolescence, 44 (43%) in middle adolescence, and 17 (16%) in late adolescence. There were no significant differences in gender distribution among developmental groups. However, more males died in early adolescence than females, 62% versus 38%, respectively.

There was little variation by stage of adolescence in use of pain medications or oxygen. Most adolescents used pain medications and few used oxygen; the use of anti-emetics was evenly distributed across groups (see Fig. 1). There was a significant difference in use of anti-anxiety medications. Relatively more individuals in late adolescence used anti-anxiety medications at EOL (p = 0.037) as noted in Fig. 2.

Fig. 1
figure 1

Percentage of adolescents using symptom control measures at EOL across developmental groups

Full size image
Fig. 2
figure 2

Difference in anxiolytic use across developmental groups at EOL

Full size image

Timing of initial EOL discussions and place of death were not significantly different among developmental groups. Additionally, there was not a significant difference in place of death. Nearly 80% of the adolescents in each developmental group died in a hospital setting.

End-of-life discussion

We identified whether EOL discussions with the adolescent and/or family were initiated by the primary oncologist or by another physician. The documentation rarely identified whether or not the adolescent was included in the discussion. A total of 46 adolescents and/or family members (68%) had initial EOL discussions with their primary oncologist. Although not significant (p = 0.071), when initial EOL conversations were conducted by the primary oncologist, death was more likely to occur at home.

There was consistent communication between the family’s referring primary-care physician and the oncologist. This was documented in 98% of charts, demonstrating coordinated efforts to keep the family’s primary-care physician aware of and involved in EOL care.

Discussion

The results of this study uncover six main issues in adolescents dying with cancer. This section will highlight each main issue with insights from clinical and medical perspectives.

Place of death and cause of death

Although studies show individuals with incurable cancer prefer to die at home [13, 14], most adolescents in this sample died of disease progression in a hospital setting. There are many factors that influence disease progression, such as relapse, secondary neoplasm, or refractory disease. Additionally, as shown in this study, there is much variability in time from diagnosis to death in pediatric oncology, making the prediction of the timing of death problematic. Health care providers interviewed in one study reported difficulty in “finding the true dying point” and agreeing collectively with other HCP and the family that curative efforts had been exhausted [15]. Similarly, there was discrepancy between parent and HCP understanding of a terminal cancer diagnosis in a large study conducted by Wolfe et al. Parental understanding lagged behind the physician by 3 to 4 months [16]. In the study of Wolfe et al., when both parent and physician recognized the prognosis of incurable cancer 50 days or more before death, there was a higher likelihood that both physician’s (odds ratio (OR) 5.17, p = 0.002) and parent’s (OR 6.56, p = 0.01) primary goal of treatment in the last month(s) of life was to lessen suffering [16].

Sometimes treatment decisions during disease progression are hindered if individuals do not fully understand the gravity of the situation. Parents queried in a study of quality of EOL care in one pediatric ICU setting agreed that honest and complete information was essential for making decisions that would be in the best interest of their child [17]. Similarly, studies conducted on adolescents with cancer have shown adolescents prefer to be informed about their diagnosis, treatment, and prognosis. They desire to fully participate in decisions [18, 19]. Dunsmore and Quine found adolescents want to be informed even when the prognosis is poor or treatment is ineffective [20]. When cure is no longer possible, discussions about goals of living while dying need to be addressed. However, knowing when to transition from curative therapy to palliative/EOL care is a challenge in pediatric oncology.

Initial EOL discussions

Discussions about death occurred in most cases (N = 80); however, 23 charts had no documentation about EOL discussions. Initial discussions occurred anywhere from 1 day before death to 880 days (2.4 years) before death, with 50% of the initial conversations beginning in the last 30 days of life, allowing minimal time for EOL preparation. Inadequate knowledge of communication strategies, difficulty dealing with personal emotions, and desire to protect the child/adolescent from the pain of separation have interfered with timely EOL conversations [19, 21–26]. Also, variability in the trajectory of a pediatric cancer diagnosis may complicate the timing of EOL discussions. In our sample, the time from diagnosis to death ranged from 21 days to almost 21 years, with 50% of the adolescents dying approximately 2 years after diagnosis.

Adolescents with a diagnosis of leukemia/lymphoma were more likely to have initial EOL discussions in the last week of life than those with other diagnoses and also were more likely to die in the ICU. It is unclear why the initial conversations occurred so close to death in adolescents with leukemia/lymphoma. However, there are a few factors that might explain the differences. Clinically, solid tumors and brain tumors are diagnostically visible. An adolescent with leukemia looks the same outwardly and the changes that are occurring in the blood stream are visible only through laboratory results. Additionally, more protocols are available for adolescents with leukemia/lymphoma, including stem cell transplants. This leads to a continuous push for cure and the line becomes even more blurred concerning when aggressive treatment is appropriate and when prolongation of life is no longer warranted.

Physical symptoms at EOL

Most of the adolescents in this study (87%) used some form of pain medication at EOL. From a developmental perspective, adolescents may be able to articulate their pain needs better than younger patients dying of cancer. Adequate pain control for all ages is essential. Accurate assessment of pain is an important component of quality care.

Psychological symptoms at EOL

According to Block, the developmental stage of an individual greatly influences his or her psychological response to a terminal or life-threatening illness [27]. Individuals facing EOL describe many fears; one fear is of inadequate pain control and another is of death itself [28–30]. Developmental age influenced the use of anti-anxiety medications in this study. There was an increase in the use of anxiolytics as the developmental age of the adolescent advanced. Similarly, a retrospective study by Theunissen et al. on the physical, psychological, and social symptoms of children and adolescents at EOL as reported by their parents revealed a significant difference in psychological symptoms for adolescents over the age of 12 compared to children under the age of 7. The most frequently reported psychological symptoms at EOL in the adolescent age group (over 12 years old) were sadness, loss of perspective, loss of independence, fear of physical symptoms, and fear of death [30]. Older adolescents with cancer reported in a recent study that because of some misconceptions about their cancer or treatment, they had lost at least some of their friends [18]. In a large prospective, EOL adult study, feelings of loneliness, and anxiety interfered with a peaceful death [31].

Cancer trajectory

In this sample, 72 adolescents (70%) had one or more relapses prior to death. There are time points during a cancer trajectory that provide opportunities to initiate conversations to address EOL uncertainties, goal setting, and meaningful activities. Relapse is an example of a time point when goals can be reexamined in light of prognosis and patients and families can become better prepared for the possibility of death. Another transitional time occurs when adolescents are facing the possibility of an SCT. In this study, 18 adolescents (17.5%) underwent an SCT. The mean time from SCT to death was 1.09 years. Although the purpose of an SCT is to cure, facing an SCT provides another opportunity for the HCP to help adolescents and their families address EOL uncertainties. In our study, 77% of EOL conversations with adolescents who had an SCT occurred in the last 32 days of life, leaving little time for the adolescents and their families to adjust to the likelihood of death.

Health care provider communication is extremely important during critical time points in a cancer trajectory. Many factors have an impact on the effectiveness of EOL communication, including the words used, the environment where conversations take place, and nonverbal cues. A qualitative study conducted with 51 families during family conferences in an ICU setting noted HCP words such as “there is nothing more we can do” can be misinterpreted as messages of abandonment [32]. In contrast, 36 family members interviewed after the death of their child said that continuity of care, having their child known as an individual, and having staff present through the child’s death provided great comfort during and after the death [33]. Adolescents and their families need to know that someone will walk beside them regardless of whether or not the cancer is curable. Helping adolescents and their families to hope for the best while simultaneously preparing for the worst is extremely challenging, but it is an essential part of advanced oncology care.

Do not resuscitate

The majority of adolescents (79%) had a DNR order in place at time of death; however, 20% had a conflict documented about DNR status. The conflict usually involved disagreement between the HCP and the parent(s), with the parent(s) not yet ready to agree to a change in DNR status. Conversations about death were quickly followed by initiation of a DNR order. The time from initial EOL discussion to DNR was only 1 day in 50% of the sample. Likewise, the time from DNR to death was only 7 days in 50% of the sample. It is not clear how long the physicians had believed that death was imminent, nor how long it took for the adolescents and families to accept that there was no realistic chance for cure.

Summary/conclusions

Most of the adolescents in this study died of disease progression in an acute care setting. Knowledge about how and where an adolescent with cancer dies is essential before interventions can be developed to better support adolescents and their families at EOL. Importantly, adolescents dying of cancer have unique developmental needs that should be considered when planning EOL care. Anxiety and psychosocial issues related to dying warrant further investigation. As adolescents progress through the stages of development and as abstract cognitive thinking matures, communication patterns as well as parent–adolescent relationships may change, making the reality of death even more devastating. More research is needed to identify how adolescents wish to be supported emotionally and psychologically at EOL and how communication or lack of communication affects distress for adolescents dying of cancer.

Limitations

This study has several limitations. The data were collected from one Midwestern tertiary care center in the USA, limiting the ability to generalize the findings. Since the study was a retrospective chart review, only information available in the chart was captured. Place of death may have influenced the number of missing variables in some charts. For example, if an adolescent went home to die, or died in an outlying hospital closer to the patient’s home, those records were not always included in the medical chart at the tertiary care center.

End-of-life discussions for this study were limited to physician conversations. If discussions about EOL occurred among nursing staff and adolescents and/or family members, the information was not captured. Finally, we do not know if symptom control measures alleviated distress. We simply know that symptom control measures were used. Despite these limitations, we believe these findings can inform researchers in planning future prospective EOL research in adolescents with incurable cancer and can guide clinicians to consider opportunities earlier in the cancer trajectory for initiating EOL discussions in adolescents dying of cancer.