Endometrial Cancer Endometrial Cancer Risk Prediction Model - - PowerPoint PPT Presentation

Endometrial Cancer Endometrial Cancer Risk Prediction Model

Jennifer Prescott, PhD Channing Division of Network Medicine Advances in Endometrial Cancer Epidemiology and Biology Symposium March 17, 2014 Supported by National Cancer Institute 1U01CA167763-01A1

Utility of risk prediction models Utility of risk prediction models

 Aid in clinical management  Aid in clinical management  Identify high risk populations for intervention studies

Identify high risk populations for intervention studies

 Understand biologic process of carcinogenesis

Understand biologic process of carcinogenesis

2

Exposures change rate of tissue aging

Colorectal cancer Endometrial cancer

e/100,000 e/100,000 Incidence rate Incidence rate

Breast cancer Ovarian cancer

0,000 0,000 ence rate/100 ence rate/100 3

Sources: Pike 1983 Nature Pike 1987 J Chron Dise Pike 2004 Oncogene

Incid Incid

Reproductive events in breast tissue aging aging

4

Sources: Pike 1983 Nature Pike 1987 J Chron Dise Pike 2004 Oncogene

Can we improve endometrial cancer p risk prediction by incorporating timing

• f exposures?
• f exposures?

5

Study population Study population

 Nurses’ Health Study (NHS)  Nurses Health Study (NHS)

 Ongoing prospective cohort  121,700 female registered nurses from 11 U.S. states  30-55 years of age at baseline in 1976  Biennial self-administered questionnaires

NHS

 High follow-up rates

6

Detailed risk factor information Detailed risk factor information

Type of female female hormones first reported

http://www.channing.harvard.edu/nh

NHS

7

Case ascertainment Case ascertainment

 Self report incident diagnosis of endometrial cancer  Self-report incident diagnosis of endometrial cancer  Women were asked permission to obtain medical records  Records reviewed by gynecologic oncologist or pathologist

 Histologic subtype  Stage  Grade

 Analysis restricted to medical-record confirmed invasive

ith li l d t i l d i (St IA IV) epithelial endometrial adenocarcinoma (Stages IA-IV)

8

Exclusions at baseline (1978) Exclusions at baseline (1978)

 Cancer diagnosis (except non-melanoma skin cancer;  Cancer diagnosis (except non melanoma skin cancer;

N=4,180)

 Hysterectomy (N=24,542)  Surgical or unknown menopausal status (N=13,673)  Missing risk factor data (N 12 519)  Missing risk factor data (N=12,519)  Eligible women at baseline: 66 786

9

Eligible women at baseline: 66,786

Risk factors modeled Risk factors modeled

 Timing of reproductive events

Timing of reproductive events

 Exogenous hormone use  BMI in low vs. high estrogen environment

(menopause/hormone use status) (menopause/hormone use status)

 Smoking  Family history of endometrial cancer

10

 Personal history of diabetes or hypertension

Statistical analysis Statistical analysis

 Cox proportional hazards regression model  Cox proportional hazards regression model  Similar exclusions throughout follow-up

g p

 Follow-up ended in June 1, 2010  Total incident cases diagnosed throughout follow-up: 648

11

Statistical analysis Statistical analysis

 Relative risks (RR)  Relative risks (RR)

 Dichotomous variables: exp(β)  Continuous variables: exp(β*contrast in risk factor)

 C-statistic for overall discriminatory ability

12

Preliminary Results Preliminary Results

13

NHS endometrial cancer incidence rates rates

140 80 100 120

per 100,000 ears

40 60 80

dence rate p person-ye

20 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Incid Age categories (years) Age categories (years) 14

Duration of premenopausal period period

Age at natural menopause RR 45 years of age 1.00 50 years of age 1.75

15

g

Number and timing of births Number and timing of births

Age at bi th RR births RR Nulliparous 1.00 20 0.68 20 23 0 53 20, 23, 26,29 0.53 29 0 48

16

29 0.48

Oral contraceptive use Oral contraceptive use

Duration RR Never 1.00 10 years 0.42 y

17

Current menopausal hormone use use

Type, Duration RR Never 1.00 Oral E, 5 years 3.91 , y Oral E+P, 5 years 2.06 years Other, 5 years 2.72

18

Body mass index (BMI) Body mass index (BMI)

BMI RR 22 kg/m2 1.00 30 kg/m2 3.58 g

19

Pack years of smoking Pack-years of smoking

Duration RR Never 1.00 20 pack- 0.92 p years

20

Medical history Medical history

Condition RR None 1.00 Family history y y

• f

endometrial cancer 1.48 Diabetes 1.34

21

Diabetes 1.34 Hypertension 1.19

Comparison of discriminatory ability ability (Internal population)

Pfeiffer et al. model Current model Difference between models P- value 0.766 (±0.009) 0.793 (±0.009)

• 0.027

(±0.007) 0.0000 3

22

Limitations Limitations

Source: http://www.win.niddk.nih.gov/statistics/#b Source: Zbuk 2012 J Epidemiol Community Health

23

Conclusions Conclusions

 Incorporating timing of exposures may improve risk  Incorporating timing of exposures may improve risk

prediction of endometrial cancer

 Reproductive years may be particularly relevant period for

intervention

24

Future directions Future directions

 Further explore relation with timing of birth  Further explore relation with timing of birth  Validate model in independent population (PLCO)  Can we improve upon base model?

p p

 Suggestive factors (e.g., coffee intake)  Plasma biomarkers and/or genetic variants  By histologic or molecular tumor subtype  By histologic or molecular tumor subtype

First dietary t Blood (1989 90) Blood (2000 01) Buccal cell (2001 04) assessment (1989-90) (2000-01) (2001-04) Tumor tissue

NHS

25

Acknowledgments Acknowledgments

NHS PLCO

 Immaculata De Vivo  Nicolas Wentzensen  Immaculata De Vivo  Bernard Rosner  Shelley Tworoger  Nicolas Wentzensen  Shelley Tworoger  Akila Viswanathan

Participants of the Nurses’ Health Study

26