Keywords:
Neoplasms - Palliative care - Prognosis - Hospitalization - Mortality
Palavras-chave:
Neoplasias - Cuidado paliativo - Prognóstico - Hospitalização - Mortalidade
INTRODUCTION
In countries where access to health is poor, most individuals with cancer arrive at
health services when it is already at an advanced stage[1]. Data from a specialized palliative care unit (PCU) in Brazil, indicate that most
patients entered the institution with advanced cancer and did not meet the eligibility
criteria to receive curative treatment[2].
In this context, clinical decisions related to procedures or surgeries, nutrition,
artificial hydration, etc., draw on the results of a prognostic assessment[3]. As a large part of this group of individuals experience reduced survival[4], especially in a hospital setting, it is essential for an adequate prognosis to
be made to ensure adequate care planning[5], with a view to optimizing treatment strategies, minimizing the risk of undertreatment
or approaches that are futile and/or disproportionate to the progression of the disease[6].
Thus, the use of good discrimination prognostic factors is essential for inpatients,
since a shorter survival time and higher mortality rates are observed in the hospital
setting. Therefore, this study aims to identify prognostic factors and their discriminatory
capacity for advanced cancer patients hospitalized at the PCU of a specialized cancer
hospital.
MATERIAL AND METHODS
This clinical, observational, prospective cohort study was conducted with patients
admitted to the PCU in Brazil. The focus of treatment in the PCU is symptom control
and promotion of quality of life and death. It begins when antitumor treatment is
interrupted due to ineffective response and/or serious side effects, so none of the
patients in the cohort were receiving any curative cancer treatment. The study was
approved by the Ethics Committee of INCA (3,550,658) and the participants were included
in the research after agreeing and signing the free and informed consent term.
Patients were continuously enrolled from October 2019 to May 2021, evaluated by trained
researchers within 72 hours of the first hospital admission, and monitored until the
outcome of their hospital stay. Eligibility criteria for the study were: having advanced
malignant neoplasm regardless of location, age ≥20 years, Karnofsky Performance Status
(KPS) ≥30%, and being able to provide the required information. The KPS is a percentage
scale that classifies the individual as to their ability to perform normal daily activities,
active work, self-care, and need for regular medical care due to greater evidence
of disease (100%: full function; 0%: death)[7].
Independent variables
The following variables were obtained from the electronic medical records: sociodemographic
(age [<60 vs. ≥60 years] and sex [male vs. female]); clinical (diagnosis [gynecological
cancer vs. breast vs. gastrointestinal tract (GIT) vs. lung vs. head and neck vs.
connective bone tissue vs. others] and distant metastasis [no vs. yes]), functionality
(KPS [30% vs. ≥40%]), nutritional (overall score of the patientgenerated global subjective
assessment short-form [PG-SGA SF©] <9 vs. ≥9). For laboratory characteristics, it
was considering these cutoff: albumin (<3 vs. ≥3g/dL), C-reactive protein (CRP, <5
vs. ≥5mg/L), C-reactive protein albumin ratio (CAR, <2 vs. ≥2), and the modified Glasgow
Prognostic Score (mGPS, 0 vs. 1+2).
After permission, the translated Portuguese version of the PG-SGA SF©, available at
pt-global.org (©FD Ottery, 2005, 2006, 2015) was used. The instrument, comprising
the first four domains of the complete tool, was administered by trained researchers
in order to assess: (1) change in body weight (score from 0 to 5); (2) food intake
(score from 0 to 4); (3) presence of symptoms of nutritional impact (score from 0
to 24); (4) functional capacity (score from 0 to 3). At the end of the evaluation,
a numerical score was generated based on the sum of each of the items in the questionnaire,
ranging from 0 to 36. The higher the score, the worse the nutritional status, with
9 being the cutoff point for classification of nutritional risk[8]
[9]. As directed by the tool, patients with cutoff ≥9 points need an urgent interventional
nutrition to control symptoms.
Outcome
The outcome evaluated was death within 30 days, based on information collected from
the medical records.
Statistical analysis
Statistical analysis was performed using Stata 13.0 (Stata Corp., College Station,
Texas, USA); p-values<0.05 were considered statistically significant.
Descriptive statistics were presented as percentages (number of observations/frequency,
%) and the death rate between groups was compared using the chi-square test for proportions.
The log-rank test was used to compare survival differences between the groups and
Kaplan Meier curves were constructed to assess the probability of survival for selected
variables (log-rank p-value<0.050).
In addition, Cox proportional regression analyses were used to identify prognostic
factors, with the hazard ratio (HR) and confidence interval (95%CI)
as measures of effect. The variables considered in the multivariate analysis were
the ones for which p≥0.20 in the univariate analysis, and were removed one by one, in descending order
of p-value. Only those with p<0.050 were retained in the final model.
The agreement statistic (C-statistic) was used to assess the discrimination of the
factors associated with the dependent variables. A C-statistic of 0.5 indicates that
the model predicts the outcome as well as chance (equal numbers of true and false
positives), 0.7 to <0.8 indicates acceptable discrimination, 0.8 to <0.9 indicates
excellent discrimination, 0.9 to <1.0 is remarkable discrimination, and 1.0 is perfect
prediction[10].
RESULTS
A total of 136 patients were included in the study. Most of them were older (≥60 years:
55.2%), female (68.4%), with gynecological as the primary site of malignancy (23.5%),
and had distant metastasis (83.1%). In general, they had low functionality (KPS 30%:
59.6%), nutritional risk (PGSGA SF© ≥9 points: 83.1%), and exacerbated systemic inflammation
(albumin <3g/dL: 59.6%, CRP ≥5mg/L: 80.4%, CAR ≥2: 70.9%, mGPS 1+2: 56.4%) ([Table 1]).
Table 1
General characteristics and median survival of inpatients with advanced cancer (n=136)
Variables
|
Total
|
Death inpantiens
|
pa
|
Survival (days)
|
No (n=59; 43.4%)
|
Yes (n= 77; 56.6%)
|
Median
|
IQR
|
pb
|
Age (years)
|
|
|
|
|
|
|
<60
|
61 (44.8%)
|
28 (45.9%)
|
33 (54.1%)
|
0.593
|
14
|
8-23
|
0.327
|
≥60
|
75 (55.2%)
|
31 (41.3%)
|
44 (58.7%)
|
|
13
|
7-20
|
|
Gender
|
|
|
|
|
|
|
Male
|
43 (31.6%)
|
20 (46.5%)
|
23 (53.5%)
|
0.617
|
13
|
7-22
|
0.935
|
Female
|
93 (68.4%)
|
39 (41.9%)
|
54 (58.1%)
|
|
14
|
8-21
|
|
Tumor type
|
|
|
|
|
|
|
Gynecological
|
32 (23.5%)
|
8 (25.0%)
|
24 (75.0%)
|
0.001
|
12
|
7-19
|
0.056
|
Breast
|
30 (22.0%)
|
22 (73.3%)
|
8 (26.7%)
|
|
16
|
14-19
|
|
GIT
|
27 (19.9%)
|
7 (25.9%)
|
20 (74.1%)
|
|
9
|
6-15
|
|
Lung
|
12 (8.8%)
|
6 (50.0%)
|
6 (50.0%)
|
|
12
|
7-18
|
|
HN
|
10 (7.3%)
|
3 (30.0%)
|
7 (70.0%)
|
|
13
|
11-19
|
|
CBT
|
9 (6.6%)
|
3 (33.3%)
|
6 (66.7%)
|
|
11
|
5-22
|
|
Othersc
|
16 (11.9%)
|
10 (62.5%)
|
6 (37.5%)
|
|
21
|
13-21
|
|
Distant metastasis
|
|
|
|
|
|
|
No
|
23 (16.9%)
|
8 (34.8%)
|
15 (65.2%)
|
0.361
|
13
|
6-14
|
0.244
|
Yes
|
113 (83.1%)
|
51 (45.1%)
|
62 (54.9%)
|
|
14
|
8-22
|
|
KPS (%)
|
|
|
|
|
|
|
30%
|
81 (59.6%)
|
29 (35.8%)
|
52 (64.2%)
|
0.030
|
11
|
6-19
|
0.025
|
≥40%
|
55 (40.4%)
|
30 (54.5%)
|
25 (45.5%)
|
|
15
|
11-22
|
|
PG-SGA SF© (points)
|
|
|
|
|
|
|
<9
|
23 (16.9%)
|
16 (69.6%)
|
7 (30.4%)
|
0.005
|
22
|
14-27
|
0.002
|
≥9
|
113 (83.1%)
|
43 (38.0%)
|
70 (62.0%)
|
|
12
|
7-18
|
|
Albumin (g/dL)
|
|
|
|
|
|
|
<3
|
65 (59.6%)
|
21 (32.3%)
|
44 (67.7%)
|
0.001
|
11
|
6-18
|
0.006
|
≥3
|
44 (40.4%)
|
28 (63.6%)
|
16 (36.4%)
|
|
17
|
11-24
|
|
CRP (mg/L)
|
|
|
|
|
|
|
<5
|
20 (19.6%)
|
12 (60.0%)
|
8 (40.0%)
|
0.135
|
20
|
7-27
|
0.127
|
≥5
|
82 (80.4%)
|
34 (41.5%)
|
48 (58.5%)
|
|
14
|
8-21
|
|
mGPS
|
|
|
|
|
|
|
0
|
44 (43.6%)
|
23 (52.3%)
|
21 (47.7%)
|
0.170
|
17
|
8-24
|
0.252
|
1+2
|
57 (56.4%)
|
22 (38.6%)
|
35 (61.4%)
|
|
14
|
8-22
|
|
CAR
|
|
|
|
|
|
|
<2
|
23 (29.1%)
|
15 (65.2%)
|
8 (34.8%)
|
0.051
|
21
|
8-24
|
0.017
|
≥2
|
56 (70.9%)
|
23 (41.1%)
|
33 (58.9%)
|
|
12
|
8-18
|
|
n = Number of observations; IQR = Interquartile range; GIT =
Gastrointestinal tract; HN = Head and neck; CBT = Connective bone tissue; KPS = Karnofsky
Performance Status; PG-SGA SF© =
Patient-Generated Subjective Global Assessment short form; CRP =
C-reactive protein; mGPS = Modified Glasgow Prognostic Score; CAR =
C-reactive protein albumin ratio.
Notes: ap-value refers to chi-square test or Fisher's exact; bp-value refers to the log-rank test; cLeukemia, lymphoma, myeloma, central nervous system, kidney and urinary tract, male
genitals, peritoneum, mediastinum, and unrecognized site.
Among the patients evaluated, 77 (56.6%) died within 30 days and the median overall
survival was 10 (interquartile range [IQR]: 6-14) days. Death rates were higher in
patients with gynecological and GIT tumors (p=0.001), KPS=30%
(p=0.030), PG-SGA SF© ≥9 points (p=0.005), albumin <3g/dL (p=0.001), and CAR ≥2 (p=0.051)
at admission ([Table 1]). The survival medians were statistically lower in these same groups ([Table 1] and [Figure 1]).
Figure 1 Survival curves of inpatients with advanced cancer according to selected variables
(n=136). Abbreviations: n = Number of observations; KPS = Karnofsky Performance Status;
PG-SGA SF© = Patient-Generated Subjective Global Assessment short form; CAR = C-reactive
protein albumin ratio.
According to the Cox univariate regression analyses, the primary tumor site, KPS,
PG-SGA SF©, serum albumin concentration, CRP, and CAR were candidates for the multiple
model. In the multivariate analysis, the primary tumor site located in the GIT (HR:
1.61, 95%CI: 1.11-2.82), KPS=30% (HR: 1.73, 95%CI: 1.09-3.00), PG-SGA SF© ≥9 points
(HR: 4.58, 95%CI: 1.62-12.92), and serum albumin concentrations <3g/dL (HR: 1.88,
95%CI: 1.05-3.34) were retained as prognostic factor within 30-day. However, only
albumin showed acceptable discrimination, with a C-statistic value of 0.75 ([Table 2]).
Table 2
Prognostic factors within 30-day among inpatients with advanced cancer (n=136)
Variables
|
HR
|
Univariate 95% CI
|
pa
|
HR
|
Multivariate 95% CI
|
pb
|
C-statistic
|
Types of tumor
|
|
|
|
|
|
|
|
GIT
|
1.83
|
1.10-3.05
|
0.021
|
1.61
|
1.11-2.82
|
0.049
|
0.67
|
Others
|
1.00
|
|
|
1.00
|
|
|
|
KPS (%)
|
|
|
|
|
|
|
|
30%
|
1.69
|
1.05-2.73
|
0.031
|
1.73
|
1.09-3.00
|
0.042
|
0.69
|
≥40%
|
1.00
|
|
|
1.00
|
|
|
|
PG-SGA SF© (points)
|
|
|
|
|
|
|
|
<9
|
1.00
|
|
0.005
|
1.00
|
|
0.004
|
0.69
|
≥9
|
3.10
|
1.41-6.77
|
|
4.58
|
1.62-12.92
|
|
|
Albumin (g/dL)
|
|
|
|
|
|
|
|
<3
|
2.15
|
1.21-3.81
|
0.009
|
1.88
|
1.05-3.34
|
0.033
|
0.75
|
≥3
|
1.00
|
|
|
1.00
|
|
|
|
CRP (mg/L)
|
|
|
|
|
|
|
|
<5
|
1.00
|
|
0.141
|
-
|
-
|
-
|
-
|
≥5
|
1.76
|
0.83-3.72
|
|
|
|
|
|
mGPS
|
|
|
|
|
|
|
|
0
|
1.00
|
|
|
|
|
|
|
1+2
|
1.36
|
0.79-2.34
|
0.264
|
-
|
-
|
-
|
-
|
CAR
|
|
|
|
|
|
|
|
<2
|
1.00
|
|
|
|
|
|
|
≥2
|
2.45
|
1.12-5.36
|
0.024
|
-
|
-
|
-
|
-
|
Abbreviations: n = Number of observations; HR = Hazard ratio; CI =
Confidence interval; GIT = Gastrointestinal tract; KPS = Karnofsky Performance Status;
PG-SGA SF© = Patient-Generated Subjective Global Assessment short form; CRP = C-reactive
protein; mGPS = Modified Glasgow prognostic score; CAR = C-reactive protein albumin
ratio.
Notes: ap-value refers to univariate Cox proportional hazard model; bp-value refers to multivariate Cox proportional hazard model.
DISCUSSION
In this study, we identified the prognostic factors and their discriminatory capacity
in patients with advanced cancer hospitalized at a specialized PCU at a reference
hospital for cancer care. Our results showed that the patients with advanced GIT cancer
who presented at the time of hospitalization with impaired functionality, reduced
serum albumin, and nutritional risk had a worse prognosis. Among these prognostic
factors, serum albumin concentration showed better discriminatory.
Prognostic assessments are challenging for health professionals and researchers, especially
for inpatients with advanced cancer. Despite the availability of validated objective
tools and prognostic factors, consideration should be given to the method to be applied,
since its accuracy may vary according to the population, environment, and forecast
period. Death is a prevalent outcome for patients hospitalized with advanced cancer,
calling for the use of specific prognostic factors to guide important personal and
clinical decisions[11]. However, there is not a great deal of scientific evidence at the present time that
focuses exclusively on cancer inpatients in palliative care.
The hospitalized patients in palliative cancer care from our study had low survival
(10 days [IQR: 6-14]), which is consistent with the findings of other studies[5]
[12]. In an Argentine cohort, hospitalized cancer patients in palliative care were found
to have a higher risk of death than those receiving outpatient follow-up (HR: 1.87,
95%CI: 1.24-2.84, p-value: 0.003)[13].
The cutoff points selected for biomarkers analysis (albumin<3.0[5], CAR≥2.0[14], CRP≥5[15], and scores GPS 1+2[16] were based on previous studies, which observed lower survival when using them. Considering
a context of lower survival, we used more severe cutoff points to better support the
care plan for these patients.
Turning to tumor location, other cohort studies carried out at the same referral center
found a high prevalence of GIT and gynecological cancer[4]
[17]
[18]. The multivariate Cox regression used in our study found that patients with GIT
cancer had a worse prognosis (HR: 1.61, 95%CI: 1.11-2.82, p-value: 0.049), which is similar to the findings of Martin et al. (2010)[19] (HR: 1.69, 95%CI: 1.30-2.19, p-value<0.001). This could be attributed to the fact that this type of cancer is associated
with a greater nutritional impact, and consequently with repercussions on overall
survival.[20] In a Brazilian multicentric study, upper digestive cancer had a strong association
with malnutrition [odds ratio (OR): 4.51, 95%CI: 3.31-6.1,
p-value<0.001].[21]
Functional capacity is recognized as a relevant prognostic factor in different health
contexts, including in cancer patients, where it is considered a strong independent
predictor of survival.[22] In their multivariate analysis, Fiorin de Vasconcellos et al. (2019)[23] found that patients with solid tumors at an advanced stage and with worse functionality
had a higher risk of death within 30 days (HR: 2.01,
95%CI: 1.14-3.53, p-value: 0.016).
Furthermore, the median KPS at admission was significantly lower in those who progressed
to death than in those whose outcome was discharge.[24] These data corroborate our findings, as our multivariate analysis revealed that
the presence of KPS=30% was a prognostic factor within 30 days (HR: 1.73, 95%CI: 1.09-3.00,
p-value: 0.042). Consistent with this, some specific prognostic tools for patients
in palliative care – such as the palliative performance scale, the palliative prognostic
score, and the palliative prognostic index – include functionality as a crucial variable
in their composition.[25]
As expected, most of our inpatients (83.1%) had an overall PG-SGA SF© score ≥9 points
(HR: 4.58, 95%CI: 1.62-12.92, p-value: 0.004), which indicates nutritional risk, considered an important prognostic
factor. A previous publication, based on research developed at the same PCU, demonstrated
that PG-SGA SF© was associated with lower survival and a higher risk of 90-day mortality,
making it an indicator of a worse prognosis[16]. This cutoff point, despite not being validated for the Brazilian cancer population,
is widely used to classify patients at nutritional risk.[26]
As for albumin, in addition to its good prognostic power, it is also a marker of nutritional
status and a simple parameter capable of reflecting inflammatory status.[27] In a study of patients with inoperable advanced esophageal cancer, survival was
significantly shorter in those with a serum concentration of this acute-phase protein
was lower than 3.5g/dL.[28] In a cohort study developed at the same PCU, the presence of hypoalbuminemia was
also an independent prognostic factor within 90-day (HR: 2.04, 95%CI: 1.16-3.58, p-value: 0.013).[16] Furthermore, in a systematic review and meta-analysis, Dolan et al. (2017)[29] found studies in which patients with serum albumin levels <3.0g/dL had lower survival
(HR: 1.57, 95%CI: 1.26-1.95, p-value<0.0001), corroborating our findings (HR: 1.88, 95%CI: 1.05-3.34, p-value<0.033).[29]
There are some limitations of our study that deserve to be highlighted. First, the
study was undertaken at a single site and had a small sample size, which could interfere
with the power of the statistical tests used. Larger studies would be needed to overcome
this limitation. Meanwhile, the study's strength is that it analyzes simple and objective
elements for prognostic assessment that could be used by any member of the multidisciplinary
team, making it more easily applied in the clinical setting.
CONCLUSION
Inpatients with advanced cancer in the GIT (primary site), impaired functionality,
reduced serum albumin, and nutritional risk at admission have a worse prognosis. Serum
albumin concentration has better discriminatory ability than the other factors identified.
Although this relationship is well explored in the literature, the use of these variables
and, some of them, with more severe cutoff points, specifically for hospitalized patients,
is an issue that has not yet been explored, which can better support the care plan
in this specific group.
Bibliographical Record
Karla Santos da Costa Rosa, Amanda Soares Oliveira, Raphael de Paiva Cypriano, Livia
Costa de Oliveira. Prognostic factors in inpatients with advanced cancer at a palliative
care unit. Brazilian Journal of Oncology 2022; 18: e-20220344.
DOI: 10.5935/2526-8732.20220344