KR20210061066A - Method and apparatus for predicting prognosis of epithelial ovarian cancer - Google Patents

Abstract

Disclosed are a method and apparatus for predicting a prognosis of epithelial ovarian cancer. According to the present invention, the apparatus for predicting the prognosis of epithelial ovarian cancer comprises: a processor; and a memory connected to the processor. The memory receives an input of a predetermined clinical variable of a patient, and stores program commands executed by the processor for calculating a predicted score of platinum anti-cancer drug reactivity related to epithelial ovarian cancer, three-year progression-free survival rate, and five-year survival rate through the clinical variable. In addition, the input of the clinical variable and the output of the calculated predicted score are conducted through a nomogram corresponding to each of the platinum anti-cancer drug reactivity, three-year progression-free survival rate, and five-year survival rate. The clinical variable is determined through a successive application of a first and a second screening and a forward and backward stepwise selection based on the area under the receiver operating characteristic curve (AUC) of a first candidate clinical variable determined by using clinical pathological characteristics collected for a plurality of patients, first treatment details, surgical findings during surgery, and survival results. The present invention aims to provide a method and apparatus for predicting a prognosis of epithelial ovarian cancer which are able to precisely predict the reactions and survival results of patients.

Description

TECHNICAL FIELD [Method and apparatus for predicting prognosis of epithelial ovarian cancer]

The present invention relates to a method and apparatus for predicting prognosis of epithelial ovarian cancer.

Ovarian cancer has the lowest 5-year survival rate (46.5%) among gynecological malignancies. Worldwide, there are 226,000 new cases and 158,00 cancer deaths each year, and the incidence rate is gradually increasing in Korea.

Because of the lack of cancer-specific symptoms and effective screening tools, the rate of diagnosis of ovarian cancer inevitably end-stage, recurrence and death is increasing.

The majority (90%) of ovarian cancer is epithelial ovarian cancer (EOC).

For patients with advanced EOC, maximal cytoreductive surgery is performed after taxane and platinum-based chemotherapy as the first-line treatment. Nevertheless, 80% of patients with a complete response after first-line treatment eventually experience disease recurrence.

With the era of precision medicine, it is necessary to find a model that predicts the exact prognosis of EOC as the first step in individualized treatment.

However, until now, models that predict the prognosis of epithelial ovarian cancer have analyzed only fragments of clinical pathological factors, and their predictive ability is low or limited.

Teramukai S, Ochiai K, Tada H, Fukushima M; Japan Multinational Trial Organization OC01-01. PIEPOC: a new prognostic index for advanced epithelial ovarian cancer--Japan Multinational Trial Organization OC01-01. J Clin Oncol. 2007;25: 3302-6.

In order to solve the problems of the prior art, a method and apparatus for predicting the prognosis of epithelial ovarian cancer capable of accurately predicting a patient's treatment response and survival result is proposed.

In order to achieve the above object, according to an embodiment of the present invention, there is provided an apparatus for predicting prognosis of epithelial ovarian cancer, comprising: a processor; And a memory connected to the processor, wherein the memory receives a predetermined clinical variable of a patient, and through the clinical variable, a platinum anticancer drug responsiveness related to epithelial ovarian cancer, a 3-year progression-free survival rate, and a 5-year survival rate are predicted. To calculate a score, program instructions executed by the processor are stored, and the input of the clinical variable and the output of the calculated prediction score are nomograms corresponding to the platinum anticancer drug responsiveness, 3-year progression-free survival rate, and 5-year survival rate, respectively. The clinical variable is AUC (area under) of the first candidate clinical variable determined using clinical pathological characteristics, primary treatment details, intraoperative surgical findings, and survival results collected for a plurality of patients. The receiver operating characteristic curve) based first, second screening, and forward and backward stepwise selection are determined through sequential application of an epithelial ovarian cancer prognosis prediction apparatus.

The clinical variables input to the nomogram for predicting the reactivity of the platinum anticancer drug related to epithelial ovarian cancer are the log serum CA-125 level at diagnosis, the FIGO stage (International Federation of Gynecology and Obstetrics stage). , Histologic type, neoadjuvant chemotherapy (NAC), pleural effusion, ascites or peritoneal washing cytology, omentum, uterus , Colon except rectosigmoid, metastasis to the liver surface, and residual tumor size after primary debulking surgery/interval debulking surgery. have.

Clinical variables input to the nomogram for predicting the 3-year progression-free survival rate are the number of hemoglobin, the number of lymphocytes, the number of monocytes, and the log-converted serum CA-125 level at the time of diagnosis, ascites or peritoneal washing. Cell pathology, NAC, ovarian surface, tube, serous and small bowel and mesentery metastases, and residual tumor size after primary/intermediate debulking surgery.

Clinical variables input to the nomogram for predicting the 5-year survival rate are: lymphocyte count, monocyte count, log-converted serum CA-125 level, pleural effusion, ascites or peritoneal lavage cell pathology, NAC, uterus, duct, serous, rectal sinus It may be colon except rectosigmoid, small intestine and mesenteric metastases.

Among the first candidate clinical variables, a second candidate clinical variable with a p-value equal to or less than a preset threshold is first screened through univariate analysis, and a preset threshold value through M-fold cross-validation among the second candidate clinical variables A third candidate clinical variable having the above area under the receiver operating characteristic curve (AUC) may be screened for the second time.

In the forward and backward stepwise selection, the AUC may be calculated by a leave-one-out cross-validation (LOOCV).

According to another aspect of the present invention, there is provided a method for predicting prognosis of epithelial ovarian cancer including a processor and a memory, the method comprising: receiving a predetermined clinical variable of a patient; And calculating predicted scores of platinum anticancer drug responsiveness, 3-year progression-free survival rate, and 5-year survival rate through the clinical variable, wherein the input of the clinical variable and the output of the calculated prediction score are the platinum. It is achieved through a nomogram corresponding to each of the anticancer drug reactivity, 3-year progression-free survival rate, and 5-year survival rate, and the clinical variables are the clinical pathological characteristics collected for multiple patients, primary treatment details, and intraoperative surgical The first and second screenings based on the area under the receiver operating characteristic curve (AUC) of the first candidate clinical variable determined using the discovery and survival results, and the forward and backward stepwise selection are determined through sequential application. A method for predicting prognosis of epithelial ovarian cancer is provided.

According to another aspect of the present invention, there is provided a computer program stored on a recording medium for performing the above method.

According to the present invention, there is an advantage of being able to accurately predict the prognosis of patients with epithelial ovarian cancer through clinical variables selected through statistical processing.

1 is a diagram showing the configuration of an apparatus for predicting prognosis of epithelial ovarian cancer according to an embodiment of the present invention.
2 is a view showing a nomogram for predicting the reactivity of a platinum anticancer drug according to an embodiment of the present invention.
3 is a diagram showing a nomogram for predicting a 3-year progression-free survival rate according to an embodiment of the present invention.
4 is a diagram showing a nomogram for predicting a 5-year survival rate according to an embodiment of the present invention.
5 is a diagram illustrating a process of determining a clinical variable according to an embodiment of the present invention.
6 shows the clinical pathological characteristics of the patient.
7 shows the results before and during surgery according to the anatomical site.
8 shows the patient's OS and PFS.
9 is a diagram showing that patients in the PDS group had significantly longer OS and PFS times than those in the NAC group.

In the present invention, various modifications may be made and various embodiments may be provided, and specific embodiments will be illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to a specific embodiment, it should be understood to include all changes, equivalents, and substitutes included in the spirit and scope of the present invention.

The present invention aims to provide a model for predicting an accurate prognosis of epithelial ovarian cancer as a first step in individual patient treatment.

1 is a diagram showing the configuration of an apparatus for predicting prognosis of epithelial ovarian cancer according to an embodiment of the present invention.

The apparatus for predicting prognosis of epithelial ovarian cancer according to the present embodiment may include a processor 100 and a memory 102.

The processor 100 may include a central processing unit (CPU) capable of executing a computer program or a virtual machine.

The memory 102 may include a nonvolatile storage device such as a fixed hard drive or a removable storage device. The removable storage device may include a compact flash unit, a USB memory stick, or the like. Memory 102 may also include volatile memories such as various random access memories.

According to an embodiment of the present invention, a predetermined clinical variable of a patient is input to the memory 102, and through the clinical variable, platinum anticancer drug responsiveness related to epithelial ovarian cancer (platinum sensitivity), 3-year progression-free survival rate (3- Program instructions executed by the processor 100 are stored to calculate predicted scores of year progression-free survival (PFS) and 5-year overall survival (OS).

In the prediction of the prognosis of epithelial ovarian cancer according to the present embodiment, the input of the clinical variable and the output of the prediction score calculated through the clinical variable are nomograms corresponding to the above-described platinum anticancer drug responsiveness, 3-year progression-free survival rate, and 5-year survival rate, respectively. The clinical parameters described above include: Patients'clinicopathologic characteristics collected for multiple patients, details of primary treatment, and intra-operative surgical findings. ) And survival results are determined through sequential application of AUC (area under the receiver operating characteristic curve)-based primary and secondary screening and forward and backward stepwise selection.

Here, the clinical pathological characteristics of the patient include a personal history such as age, obstetric history, a family history of breast cancer or gynecological malignancies, a history such as high blood pressure or diabetes, and the serum CA-125 level and neutrophils, lymphocytes, monocytes and Includes the number of individual blood cells, including platelets. It also includes information on the World Federation of Obstetrics and Gynecology (FIGO) staging, histology, and tumor differentiation.

The first-line treatment details include the use of neoadjuvant chemotherapy (NAC), the extent and individual procedure of debulking surgery, and the administration and frequency of taxane and platinum-based chemotherapy after surgery.

Intraoperative surgical findings include residual tumor size and site after debulking, preoperative images, and intraoperative surgical findings.

2 is a view showing a nomogram for predicting the reactivity of a platinum anticancer drug according to an embodiment of the present invention.

Referring to FIG. 2, the clinical variables input to the nomogram for predicting the reactivity of the platinum anticancer agent according to the present embodiment are the log serum CA-125 level at diagnosis, and the FIGO stage (International Federation of Gynecology and Obstetrics stage), histologic type, neoadjuvant chemotherapy (NAC), pleural effusion, ascites or peritoneal washing cytology, omentum , Uterus, colon except rectosigmoid, liver surface, and residual tumor size after debulking surgery.

Some clinical variables, such as serum CA-125 levels, are log transformed to relieve skewness.

3 is a diagram showing a nomogram for predicting a 3-year progression-free survival rate according to an embodiment of the present invention.

Referring to FIG. 3, the clinical variables input to the nomogram for predicting the 3-year progression-free survival rate according to the present embodiment are the number of hemoglobin, the number of lymphocytes, the number of monocytes, and log conversion at the time of diagnosis. Serum CA-125 levels, ascites or peritoneal lavage cytopathy, NAC, ovarian surface, tube, serous and small bowel and mesentery metastases and primary/intermediate debulking surgery (primary) It is the residual tumor size after debulking surgery/interval debulking surgery, PDS/IDS).

4 is a diagram showing a nomogram for predicting a 5-year survival rate according to an embodiment of the present invention.

Referring to FIG. 4, the clinical variables input to the nomogram for predicting the 5-year survival rate according to the present embodiment are the number of lymphocytes, the number of monocytes, the log-converted serum CA-125 level, pleural effusion, ascites or peritoneal lavage cell pathology, NAC, uterus, duct, serous, colon except rectosigmoid, small intestine and mesenteric metastases.

5 is a diagram illustrating a process of determining a clinical variable according to an embodiment of the present invention.

Referring to FIG. 5, the clinical pathological characteristics, primary treatment details, surgical findings and survival results during surgery are collected for a plurality of patients (step 500).

Thereafter, a first candidate clinical variable is determined using the collected information (step 502).

Among the first candidate clinical variables determined in step 502, a second candidate clinical variable whose p-value is less than or equal to a preset threshold is first screened through univariate analysis (step 504).

For example, in step 502, the threshold of the p-value may be set to 0.05 because it may become an important variable when a variable having a large area under the AUC condition is statistically significant.

Next, among the second candidate clinical variables, a third candidate clinical variable having an area under the receiver operating characteristic curve (AUC) greater than or equal to a preset threshold through M-fold cross-validation is secondarily screened (step 506).

In step 506, 10-fold cross-validation may be used, and the threshold for AUC may be 0.55, but is not limited thereto.

A predictive model is generated through the third candidate clinical variable (step 508).

To generate a predictive model, a logistic regression model for platinum anticancer drug responsiveness and a Cox regression model for PFS and OS were fitted for stepwise variable selection using AUC.

For the Cox regression model, a time dependent receiver operating characteristic (ROC) curve is constructed and a time dependent AUC is calculated. AUC is calculated by LOOCV (leave-one-out cross-validation) as follows.

Estimates the parameters of the prediction model from n-1 samples (where n is the sample size) and uses the estimated parameters from n-1 samples to obtain predictions for the remainder.

It is repeated for all samples and has n predicted values based on n estimated predictive models. Next, the AUC is calculated using the predicted and observed values of the response variable.

According to an embodiment of the present invention, in the process of generating the predictive model in step 508, final clinical variables for inputting a predictive nomogram for prognosis of epithelial ovarian cancer are determined.

Here, the selection of the final clinical variable is performed through forward and backward stepwise selection.

First, to select the first clinical variable, a number of univariate models are fitted as many as the number of predictors.

AUC is calculated for each clinical variable, and the one with the highest AUC value is selected.

To search for the model with the highest AUC, the forward and backward stepwise selection method is used.

를 m번째 단계에서 선택할 임상 변수라 하고,

는

로 구성된 세트이다.

Is the clinical variable to be selected in the mth stage,

Is

It is a set consisting of.

는

를 이용한 AUC 값을 나타낸다. 즉 m-1개까지 선택된 변수 세트에 변수

를 추가한 모델의 AUC 값을 의미한다.

Is

Shows the AUC value using. In other words, the variable set is selected up to m-1

It means the AUC value of the model to which is added.

를

에 추가하였을 때의

이 계산된다. (1) Advance phase: each

To

When added to

Is calculated.

에 속하는

후보들에 대해서 가장 큰

값을 가지면서, 그 값이

보다 크면 최대값

을 가지는

를 선택하고,

및

를 설정한다. (1)단계에서 새로운 변수가 추가되면 2단계를 진행한다.after that

Belong to

The largest for the candidates

Having a value, that value is

Greater than the maximum

Having

Select,

And

Is set. If a new variable is added in step (1), proceed to step 2.

If a new variable is no longer selected, proceed to step 3.

에 대해

가 계산되며, 여기서

는

가 제외된 세트

을 나타낸다.

에 속하는

후보들에 대해서

가 제외된 모델들의

값 중 그 값이 가장 크고

보다 크면 최대값

을 가지는

를 삭제한다. (2)단계에서 기존 변수 중 하나가 삭제된 새로운 모델이 선택되면 다시 1단계를 진행한다.(2) Reverse step: each

About

Is calculated, where

Is

Set without

Represents.

Belong to

About the candidates

Of models excluding

Among the values, that value is the largest

Greater than the maximum

Having

Delete. In step (2), if a new model in which one of the existing variables has been deleted is selected, step 1 is performed again.

If there is no further deletion, proceed to step 3.

(3) Stop the step-by-step method for variable selection.

The forward and backward selection method according to the present embodiment is an algorithm that repeats the forward and backward steps by measuring AUC. When the updated model does not increase the AUC value than the previous model, the selection process is terminated to determine the prediction model.

Hereinafter, an experimental procedure for predicting the prognosis of epithelial ovarian cancer according to the present embodiment will be described.

Study population

Two tertiary hospital ovarian cancer cohort databases included patients who met the following inclusion criteria.

(1) patients over 18 years of age;

(2) Patients who were diagnosed with EOC at Seoul National University Hospital (SNUH) or Asan Hospital (AMC) from January 2007 to August 2016 and were treated mainly.

However, (1) patients with malignant tumors other than EOC; And (2) Excluding those with insufficient clinical data, information related to 866 patients who met the criteria was collected.

Data collection

A vast amount of patient clinical pathologic data was collected, including 108 variables across 16 domains. Data collection included personal history such as age, obstetric history, family history of breast cancer or gynecological malignancies, history such as hypertension or diabetes.

Serum CA-125 levels and individual blood cell counts including neutrophils, lymphocytes, monocytes and platelets at initial diagnosis were obtained.

The World Federation of Obstetrics and Gynecology (FIGO) stages, histology and tumor differentiation and the use of neoadjuvant chemotherapy (NAC), the degree and individual procedures of debulking surgery, and the administration and frequency of taxane and platinum-based chemotherapy after surgery. Details of the primary treatment, including, were also obtained.

After debulking surgery, the residual tumor size and site were investigated, and optimal debulking surgery was considered when the residual tumor size was less than 1 cm in the longest diameter. Preoperative images and intraoperative results were also investigated.

All patients receive CT scans and measurements of serum CA-125 levels every 3 cycles during adjuvant chemotherapy, and every 3 months for 1 year, then every 6 months for the next 3 years during the observation period after first line therapy. CT scans and measurements of serum CA-125 levels were taken. Response to anticancer drug therapy was evaluated by Response Evaluation Criteria in Solid Tumors (RECIST).

For survival analysis, for patients with measurable disease, RECIST ver. Progression-free survival (PFS) as assessed by 1.1 was defined as the time elapsed from the start of the first treatment (the date of the first debulking surgery (PDS) date or the date of the first cycle of the NAC) to the date of disease progression. . In the case of patients with unmeasurable disease, disease progression was confirmed using the Gynecologic Cancer InterGroup criteria using serum CA-125 levels.

Overall survival (OS) was defined as the time elapsed from the date of initial diagnosis to the date of cancer-related death or study termination. For all patients, survival status was investigated using institutional medical records and social security mortality index.

For the evaluation of platinum anticancer drug responsiveness, only patients who actually received taxane and platinum-based anticancer drug therapy as primary treatment were included. Among them, those who relapse are classified as either a PSR defined as recurrence of at least 6 months after completion of first-line treatment or a PRR defined as recurrence of less than 6 months. In addition to the PSR, those who completed taxane and platinum-based chemotherapy and did not experience disease recurrence for more than 6 months following the follow-up were also considered to be sensitive to platinum.

Patient's clinical pathological characteristics

The clinical pathological characteristics of a total of 866 patients are shown in FIG. 6.

The average age of the patients was 53.5 years. Overall, 584 patients (67.4%) had FIGO stage III-IV disease, and the most common histological type was serous type (61.1%).

Details of the primary treatment are presented in FIG. 6.

A total of 712 patients (82.2%) received PDS, while the other 154 (17.8%) received NAC followed by IDS.

The optimal volume reduction surgery rate was 90.3%. After surgery, 792 patients (91.5%) received taxane and platinum-based chemotherapy after surgery. Among them, 616, 108, and 13 patients showed complete remission, partial remission and stable disease, respectively, while 55 patients showed progressive disease.

7 shows the results before and during surgery according to the anatomical site.

Ascites analysis of peritoneal lavage cytopathy was performed in 785 patients (90.6%), and malignant cells were detected in 482 patients (55.7%). Meanwhile, 76 patients (8.8%) at the time of diagnosis had pleural effusion.

Patient's Survival Outcomes and Treatment Response

The mean follow-up period of 866 patients was 42.4 months (interquartile range, 25.7 to 69.9 months), during which 441 patients (50.9%) experienced disease recurrence.

The patient's OS and PFS are shown in FIG. 8.

While the median OS has not been reached, the median PFS is 32.6 months. The 5-year OS and 3-year PFS rates are 68.4% and 47.8%, respectively.

Significant differences were observed not only in PFS (p <0.001) but also in OS (p <0.001) according to FIGO stage.

The 5-year OS rates for FIGO stages I-II and III-IV were 91.9% and 55.8%, respectively, and the 3-year PFS rates were 84.9% and 30.6%, respectively. For stage III-IV, the average OS was 76.8 months and the median PFS was 17.9 months.

In addition, survival analysis was performed according to the primary treatment strategy.

Patients in the PDS group had significantly longer OS (5 year survival rates, 73.3% vs 45.6%; p <0.001) and PFS (median values, 59.6 vs 15.9 months; p <0.001) compared to patients in the NAC group. Had (Fig. 9).

Of the recurrences (n=441), 433 patients (98.2%) received taxane and platinum-based chemotherapy after surgery, and 285 patients and 148 patients had PSR and PRR, respectively. In the evaluation of platinum anticancer drug reactivity, 562 patients (64.9%) and 148 patients (17.1%) were classified as platinum sensitive and resistant patients, respectively.

The above-described embodiments of the present invention are disclosed for the purpose of illustration, and those skilled in the art who have ordinary knowledge of the present invention will be able to make various modifications, changes, and additions within the spirit and scope of the present invention, and such modifications, changes and additions It should be seen as falling within the scope of the following claims.

Claims (8)

As an epithelial ovarian cancer prognosis prediction device,
Processor; And
Including a memory connected to the processor,
The memory,
Receive a predetermined patient's clinical parameters,
To calculate the predicted scores of platinum anticancer drug reactivity, 3-year progression-free survival rate, and 5-year survival rate through the above clinical variables,
Store program instructions executed by the processor,
The input of the clinical variable and the output of the calculated prediction score are made through a nomogram corresponding to each of the platinum anticancer drug reactivity, 3-year progression-free survival rate, and 5-year survival rate,
The clinical variable is an area under the receiver operating characteristic curve (AUC) of the first candidate clinical variable determined using clinical pathological characteristics, primary treatment details, intraoperative surgical findings, and survival outcomes collected for a plurality of patients. ) A device for predicting prognosis of epithelial ovarian cancer that is determined through sequential application of the based first, second screening and forward and backward stepwise selection. The method of claim 1,
The clinical variables input to the nomogram for predicting the reactivity of the platinum anticancer drug related to epithelial ovarian cancer are the log serum CA-125 level at diagnosis, the FIGO stage (International Federation of Gynecology and Obstetrics stage). , Histologic type, neoadjuvant chemotherapy (NAC), pleural effusion, ascites or peritoneal washing cytology, omentum, uterus , Colon except rectosigmoid, metastasis to the liver surface, and epithelium including residual tumor size after primary debulking surgery/interval debulking surgery Device for predicting prognosis of sexual ovarian cancer. The method of claim 1,
Clinical variables input to the nomogram for predicting the 3-year progression-free survival rate are the number of hemoglobin, the number of lymphocytes, the number of monocytes, and the log-converted serum CA-125 level at the time of diagnosis, ascites or peritoneal washing. Cell pathology, NAC, ovarian surface, tube, serous and small bowel and mesentery metastases, and the residual tumor size after primary/intermediate debulking surgery, a device for predicting the prognosis of epithelial ovarian cancer . The method of claim 1,
Clinical variables input to the nomogram for predicting the 5-year survival rate are: lymphocyte count, monocyte count, log-converted serum CA-125 level, pleural effusion, ascites or peritoneal lavage cell pathology, NAC, uterus, duct, serous, rectal sinus A device for predicting the prognosis of epithelial ovarian cancer, which is colon except rectosigmoid, small intestine and mesenteric metastases. The method of claim 1,
Among the first candidate clinical variables, a second candidate clinical variable whose p-value is less than or equal to a preset threshold is first screened through univariate analysis,
The apparatus for predicting prognosis of epithelial ovarian cancer in which a third candidate clinical variable having an area under the receiver operating characteristic curve (AUC) equal to or greater than a preset threshold value is screened for a second time through M-fold cross-validation among the second candidate clinical variables. The method of claim 5,
The device for predicting prognosis of epithelial ovarian cancer in the forward and backward stepwise selection, wherein the AUC is calculated by a leave-one-out cross-validation (LOOCV). A method for predicting prognosis of epithelial ovarian cancer comprising a processor and a memory,
Receiving a predetermined clinical parameter of the patient; And
Comprising the step of calculating a predicted score of a platinum anticancer drug reactivity, a 3-year progression-free survival rate, and a 5-year survival rate through the clinical variables,
The input of the clinical variable and the output of the calculated prediction score are made through a nomogram corresponding to each of the platinum anticancer drug reactivity, 3-year progression-free survival rate, and 5-year survival rate,
The clinical variable is an area under the receiver operating characteristic curve (AUC) of the first candidate clinical variable determined using clinical pathological characteristics, primary treatment details, intraoperative surgical findings and survival outcomes collected for a plurality of patients. ) A method for predicting the prognosis of epithelial ovarian cancer, which is determined through sequential application of the based primary and secondary screening and forward and backward stepwise selection. A computer program stored on a recording medium for performing the method according to claim 7.

Images