Abstract
Background: There has been no construction of nomograms
specifically geared towards pblueicting the prognosis of older patients
with rectal cancer. Our objective was to create a nomogram that can
accurately pblueict the chances of cancer-specific survival (CSS) and
overall survival (OS) for older patients with rectal cancer,stratifying
them into distinct risk groups.
Methods: Data of older than 65 years with rectal cancer from
the Surveillance, Epidemiology, and End Results (SEER) database between
2000 and 2020 were extracted.
Results: A total of 12,780 patients were collected in order to
identify the independent prognostic factors for cancer-specific survival
(CSS) and overall survival (OS) and construct nomograms. The factors
included age, sex, race,year of diagnos, marital status, grade, tumor
size, CEA, T classification, N classification, Months from diagnosis to
treatment, surgery, chemotherapy, and radiation. LASSO regression was
used for this purpose. The results showed that grade, chemotherapy, and
months from diagnosis to treatment were identified as independent
prognostic factors for OS, while grade, chemotherapy, and tumor size
were identified as independent prognostic factors for CSS. The
concordance index of the CSS nomogram was 0.609 (95% confidence
interval [CI], 0.598-0.619) in the training cohort and 0.682
(95%CI, 0.665-0.698) in the validation cohort. The concordance index of
the OS nomogram was 0.697 (95%CI, 0.686-0.708) in the training cohort
and 0.605 (95%CI, 0.589-0.621) in the validation cohort.
Conclusion: Caution should be taken when administering adjuvant
therapy after surgery in older patients with rectal cancer,as we have
developed and validated a new pblueictive nomogram for CSS and OS.
Keywords: elderly rectal cancer, nomogram, prognosis, survival
Introduction
From 2000 to 2016, there has been a noticeable increase in the rates of
occurrence and mortality associated with rectal cancer. In China, the
incidence rate of rectal cancer climbed to the second position, while
the mortality rate ranked fourth. In the United States, among the
147,950 individuals diagnosed with rectal cancer in 2020, 54% were aged
65 and above. This same age group accounted for 68% of rectal cancer
deaths [1-2]. Additionally, the 2020 statistics on rectal cancer
highlight that it is the third most commonly diagnosed cancer and the
third leading cause of cancer-related deaths for both males and females
[3]. Recent data from the National Center for Health Statistics
reveals that adults aged 65 and older comprised 44% of newly diagnosed
rectal cancer cases in the US in 2020 [4]. Considering Ireland, a
substantial majority of rectal cancer patients were even older, with
42% of them being over 70 years old between 2014 and 2017 [5].
During the past three decades, there have been significant advancements
in the management of locally advanced rectal cancer (LARC) through the
use of multimodal therapy, which includes neoadjuvant chemoradiotherapy
and total mesorectal excision. As a result, there has been a notable
decrease in the occurrence of local recurrence [6]. The use of total
neoadjuvant therapy, also known as TNT, which involves preoperative
systemic treatment followed by radiation, could enhance adherence and
potentially decrease distant metastasis. While several recent
large-scale trials have shown promising outcomes with TNT, these
findings may not be entirely applicable to older individuals diagnosed
with rectal cancer. For instance, the French trial PRODIGE 23
demonstrated an improvement in three-year disease-free survival (DFS)
from 69% to 76%, but subgroup analysis did not reveal a discrepancy in
the DFS benefits between patients aged 70 or younger and those over 70
[7]. The overall survival and acute toxicity were enhanced through
the trial POLISH II, however, the TNT group had a median age of only 60
and 75% of the participants were 66 years old or younger. This poses a
challenge in implementing this approach for older patients [8].
Although, rectal cancer treatment typically involves a combination of
therapies and is more frequently associated with higher morbidity rates
compared to colon cancer [9], Compared to younger patients, older
individuals diagnosed with rectal cancer exhibit physical decline, have
a higher prevalence of concurrent ailments, and experience a lengthier
recovery period following surgery [10]. Clinical trials for cancer
often do not include a sufficient number of elderly patients, resulting
in a lack of efficacy data and ambiguous guidance for doctors when
making treatment decisions. In the prognosis prediction of rectal cancer
patients, the eighth edition of the American Joint Committee on Cancer
commonly uses the TNM staging system [11]. Nevertheless, there are
certain constraints when it comes to predicting postoperative overall
survival. The outcome for individuals diagnosed with rectal cancer can
greatly differ, even if they are in the same TNM stage. Numerous factors
other than tumor stage, including tumor location, histological type,
age, gender, microsatellite status, and RAS/RAF mutation, can influence
the overall survival of patients with stage I-III CRC. It is important
to note that no individual factor can provide an accurate prediction of
survival in individuals with rectal cancer [12-13]. Important
prognostic factors have been incorporated and illustrated in nomograms,
which are widely utilized in clinical oncology. These nomograms serve as
reliable tools for estimating numerical probabilities for individual
patients [14-17]. This has been demonstrated in several types of
cancers[18-20], the nomograms’ predictions may be more precise in
comparison to the traditional TNM staging systems for different types of
tumors [21-22]. Furthermore, we have thus far found relatively few
nomograms for predicting survival in older patients with rectal cancer.
The objective of this study was to create a predictive model that
incorporates all relevant prognostic factors and can determine the
overall survival (OS) and cancer-specific survival (CSS) for elderly
patients with rectal cancer. In addition, we aimed to classify patients
into various risk groups using this nomogram. By providing personalized
prognostic information, this tool can assist healthcare professionals
and patients in making well-informed decisions regarding treatment and
management.
- Methods
- Data SourceThe SEER database comprises multiple tumor registries across various
regions, providing comprehensive cancer-related information,
including treatment details. The SEER data is accessible to the
public. Our study focused on collecting clinical data of women who
were diagnosed with rectal cancer from 2000 to 2020, specifically
from the SEER database submission of November 2022, known as SEER
17. This data encompasses various parameters such as gender, age at
diagnosis, year of diagnosis, race, marital status, grade, CEA
levels, tumor size, TMN stage, treatment information, and follow-up
data.
- Study PopulationTo extract data from the SEER database (https://seer.cancer.gov), we
utilized SEER*-Stat 8.4.2 software from the National Cancer
Institute in Bethesda, MD, USA. Our study included patients
diagnosed with primary rectal cancer, confirmed by positive
histology, in American Joint Committee on Cancer (AJCC) stages I to
III between 2000 and 2020. We only included patients with complete
clinical and demographic information. Patients diagnosed between
2000 and 2010 were staged using the 6th edition of AJCC rectal
cancer, while those after 2010 were staged using the 7th edition.
Patients who lacked clinical and follow-up information were excluded
from the study.
- Variables and endpointsThe variables obtained from the SEER database comprised of age,
gender, ethnicity, year of diagnosis, marital status, tumor grade,
tumor size, CEA levels, T classification, N classification, time
period between diagnosis and treatment, surgical procedures,
chemotherapy administration, and radiation therapy. The study
primarily examined overall survival (OS), rectal cancer-specific
survival (BCSS), and the hazard ratio (HR). OS indicated the
duration of survival from the diagnosis of rectal cancer until death
from any cause, while BCSS denoted deaths specifically attributed to
rectal cancer.
- Propensity score matchingTo minimize selection bias, propensity score matching (PSM) was
carried out for all relevant confounding factors, as proposed by
Rosenbaum and Rubin in 1983. This method of matching is not
constrained by the number of events [22-23]. In our research, we
employed the techniques of nearest available neighbor matching and
caliper matching in PSM [24-25]. For the patients who received
surgery or combination therapy (surgery plus RT or CT), to balance
the effects of confounding a PSM with a 1:1 ratio and the caliper of
0.02 was set. Only when the propensity score of the control group
(surgery plus RT or CT) is within a certain distance (0.02), the
control group will be matched with the case. Matched covariates
include sex, age, race, year of diagnosis, marital status, grade,
CEA, tumor size, stage and months from diagnosis to treatment.
- Statistical AnalysisIn order to ensure the accuracy of the model, we randomly divided
the patients into two groups, a development cohort and a validation
cohort, with a ratio of 7:3. We compared the categorical
characteristics between the training cohort and the validation
cohort using the Pearson chi-square test. Then, we used univariate
and multivariate Cox regression models to determine the independent
prognostic factors for overall survival (OS) and cancer-specific
survival (CSS) in the training cohort. The potential risk factors
identified in the multivariate analysis of the training cohort were
further analyzed using the least absolute shrinkage and selection
operator (LASSO) regression algorithm. We created dummy variables
for categorical variables and used cross-validation to determine the
appropriate tuning parameters for LASSO logistic regression.
Finally, LASSO selected the most significant features.
We performed multivariate Cox proportional hazards analyses using
the most significant features selected by LASSO from the training
dataset. Variables with a P<0.05 by multivariate analysis
were then included in the construction of nomograms. These nomograms
were designed to predict the 1-, 3-, and 5-year CSS and OS rates.
The results of the subgroup analysis were presented using a forest
plot. To evaluate the discrimination of the nomograms, we calculated
the concordance index (C-index) and area under the receiver
operating characteristic (ROC) curve (AUC) values. Additionally,
calibration curves were plotted at 1, 3, and 5 years using a
bootstrap involving 100 resamples to assess the consistency between
the actual prognosis and the survival predicted by the nomograms.
Furthermore, we determined the optimal cutoff value based on the
risk scores obtained from the nomograms using time-dependent ROC
(tdROC) curve analysis, next, the patients were categorized into
low-risk and high-risk categories based on the established
threshold. Subsequently, the Kaplan-Meier method was employed to
create OS and CSS curves, and the log-rank test was utilized to
compare the survival between the groups. All the statistical
analysis mentioned above were conducted using R software
(http://www.R-project.org, The R Foundation). The Free Statistics
software versions 1.8 were employed for all the statistical
analyses. A p-value of less than 0.05 was deemed to represent a
statistically significant difference.
- Results
- Patient baseline characteristicsFigure 1 illustrates the data filtering process. Following a
rigorous application of inclusion and exclusion criteria, a cohort
of 12,780 individuals diagnosed with rectal cancer was selected for
the study. Through the utilization of R software, these patients
were randomly split into a training cohort and a validation cohort
in a 7:3 ratio, resulting in 8948 patients in the training cohort
and 3832 patients in the validation cohort. The baseline
characteristics of the patients in both cohorts were found to be
similar, as showed by Table 1.