In this study, we used GBTM to analyze the trajectory of ADL among patients with terminal cancer as assessed by FIM for 8 weeks before death. As a result, we identified four trajectories (Fig. 1). We also identified brain metastasis, bone metastasis, age over 65 years, and lung metastasis as related factors. To our knowledge, this is the first to show accurately the ADL trajectories of groups of patients with terminal cancer using a sample size sufficient for GBTM analysis and a standardized ADL measure.
ADL trajectory of patients with terminal cancer
Previous studies have reported that ADL among patients with terminal cancer begin to decline slowly at about 6 months before death  and sharply at between 3 months and 1 month before death [3, 4], resulting in most patients requiring full assistance from 1 week before death [4, 5]. Seow et al.  reported that palliative performance status (PPS) decreased to an average of 54.7% (often assisted level) at 1 month before death and to an average of 41.3% (mostly assisted level) at one week before death. PPS is based on a rating scale from 100% to 10% of the most applicable level for each of the following items: standing, activities and symptoms, ADL, oral intake, and level of consciousness. The lower the PPS score, the worse the condition. Lunney et al.  evaluated changes in the ADL of patients with cancer in the year before death based on whether they were able to perform seven ADL items independently. They reported that the average number of ADL items requiring assistance at 1 year before death was 0.77, but increased to 4.09 at 1 month before death. McCarthy et al.  evaluated the ADL of patients with rectal and non-small-cell lung cancer in the six months before death using a modified version of the Katz Index of independence in ADL, which assesses seven items: bathing, dressing, urinary control, transfers, toileting, eating, and walking. They reported that patients had five ADL disabilities from 1 month to 3 days before death, and seven disabilities at 3 days before death. However, these previous studies have shown changes in ADL according to the median or mean, and thus reported only one trajectory. In addition, PPS cannot evaluate each ADL item in detail. It is difficult to detect changes in the level of assistance required for ADL by using methods such as the Katz index, which evaluates only whether a person is independent.
To overcome these issues, in the present study, we evaluated ADL using the FIM and estimated multiple trajectories using GBTM. The FIM can capture sensitive changes in the level of assistance and indicate patients who require assistance with some ADL. We identified three different trajectories in addition to that of the “Rapid Decline” group, as in previous studies.
Concerning previous studies using GBTM, Gill et al.  reported five trajectories of ADL in 383 community-dwelling older adults with cancer, advanced dementia, organ failure, sudden death, frailty, and other conditions in the year before death. However, the participants in their study were older adults living in the community with various diseases, which made comparisons with our study difficult. On the other hand, Yasui et al.  analyzed ADL trajectories assessed by seven items in 22 patients with cancer using GBTM. Their results showed three trajectories in the month before death. They reported that the youngest group showed a rapid decline in ADL at 1 week before death, whereas the ADL in the other two groups, which consisted of older patients, had already declined.
In contrast to the report by Yasui et al. , four trajectories were revealed in the present study (Fig. 1). The reason for the different trajectories could be due to the evaluation measures and the number of participants. Yasui et al.  assessed whether patients could perform ADL independently, whereas in the present study, the FIM  was used to evaluate gradual declines, such as in the Moderate Disability and Slow Decline group. We were able to show this group because the FIM evaluates 18 items on a seven-point scale, which allowed us to detect gradual declines. This group could not be identified by independent or dependent assessment only. Also, according to Nagin , 300 to 500 cases are required for analysis using GBTM. The study by Yasui et al.  only included 22 patients, which is not a sufficient number of cases to identify the number of trajectory groups and their respective trajectories.
Factors associated with the trajectories
With reference to the No Decline group, multinomial logistic regression analysis revealed brain metastasis as a factor with an OR of 10.3, age over 65 as a factor with an OR of 2.93, and lung metastasis as a factor with an OR of 0.36 in the Severe Disability group (Table 2).
Patients with cancer with brain metastases or brain cancer have been reported to have neurological disorders and reduced ADL [15, 16]. When these patients reach the end of life, they tend to have symptoms of consciousness disorder, drowsiness, and cognitive impairment [17, 18]. Such patients have also been reported to have more nursing problems in regard to ADL than other patients with cancer . Patients with cancer with brain metastases are thought to require severe assistance in ADL because of not only impaired physical and mental functions, but also a decreased level of consciousness.
Another factor was that the patients in this study were aged 65 years or older (i.e., older), with a median age of 77 years. It is known that ADL before death are more likely to be lower in older than in younger patients with cancer. Yasui et al.  reported that among patients with terminal cancer, older patients adults (mean age, 78.5 years) had a lower ADL trajectory in the month before death than did their younger counterparts (mean age, 57.1 years). In addition, Costantini et al.  investigated ADL in patients with cancer at 52 weeks before death and reported that ADL tended to be lower in the 65–84 and 85+ age groups than in the 18–64 age group. In the present study, we considered that the ADL of older patients with cancer were low and extracted as a factor associated with the Severe Disability group.
By contrast, lung metastasis had a weak association with the Severe Disability group. Although a few reports have investigated the effect of lung metastasis on ADL, Yoshioka  reported finding no significant difference in the Barthel Mobility Index between patients with and without lung metastases. Patients with lung cancer often experience dyspnea , which may affect their mobility. However, unlike brain or bone metastases, lung lesions do not directly cause central nervous system or motor system disorders. Therefore, cognitive dysfunction such as attention and memory impairment and physical dysfunction such as motor paralysis and sensory impairment are less likely to occur; thereby, such associations would be lower.
On the other hand, bone metastasis was extracted as a factor in the Moderate Disability and Slow Decline group. It has been reported that patients with bone metastases have lower ADL than do those without . Pain, pathological fractures, spinal cord compression symptoms, and hypercalcemia, which can occur with bone metastases, are referred to as skeletal-related events (SREs) . When SREs occur and cause pain at rest and during movement, as well as motor and sensory paralysis of the limbs, the ability to sit, stand, and walk is impaired, and ADL are reduced [24, 25]. However, bone metastases tend to spread to the spine, pelvis, and femur , and when patients are able to use their upper extremities in bed or in a wheelchair, they may not need assistance with some activities such as eating and dressing. Therefore, bone metastasis may have been a factor in the trajectory group with mild to moderate ADL disability.
This study had several limitations. First, it was conducted at a palliative care unit. Of the four trajectories, the proportions of patients in the Moderate Disability and Slow Decline and Severe Disability groups were higher. The hospital environment, such as the room size and corridor distance, differs from that at home. In addition, in hospitals, patients may be assisted more than necessary to prioritize safety and help prevent falls. Therefore, the patients in this study may have had less independence in carrying out ADL.
Second, since the FIM is assessed every 2 weeks after admission, the assessment at just before death (0 weeks) can range from 1 to 13 days. Therefore, ADL are expected to decline further during this period. In particular, the No Decline group may have experienced a rapid decline in ADL during this period. We cannot deny the possibility that there may be a decline in ADL that was not revealed because of the range of assessment periods.
Third, the size of the metastases and presence of symptoms were not assessed. It is possible that small metastases with minimal influence and large metastases with severe impairment of physical and mental functions were equally assessed. In the future, the quality and impact of metastases should be evaluated.
Finally, each patient had a different number of FIM assessments. Therefore, we cannot exclude the possibility that this may have affected the model. However, if only patients with terminal cancer surviving for more than 8 weeks were included, biases in terms of the patients’ characteristics would likely occur. In this study, we were able to develop a more realistic ADL trajectory model by including patients with various hospitalization durations and using GBTM, which can withstand deficits.