Cervical Cancer

Disease Overview

Cervical cancer is a female-specific cancer. It is prevalent globally, ranked as the fourth-most common cancer in women. In 2018, about 106,430 new cervical cancer cases are diagnosed in China and about 47,739 cervical cancer deaths occur annually in China. For women at 50-54 years of age, the incidence of cervical cancer reaches its maximum value, 30 cases per 100,000 person-years. The deaths due to cervical cancer increase over age, and have the highest value 29 per 100,000 person-years in elder people who aged above 75 years. [HPV-and-related-disease-2019-summary-report]

ICD codes for cervical cancer

Cause	ICD10
Benign cervical cancer	D06 (D06.0, D06.1, D06.7, D06.9), N87.1
Invasive cervical cancer	C53 (C53.0, C53.1, C53.3, C53.4, C53.8, C53.9)

GBD 2017 Modeling Strategy

The cervical cancer model includes both benign and invasive state. Benign cervival cancer is modelled together with uterine cancer, namely benign and in situ cervical and uterine neoplasms. GBD directly processed the combined clinical informatics data in DisMod so they don’t have a way to split it in cause level. However, it seems doable if the clinical informatics team could remap this cause to benign cervical and uterine separately in next round.

State definition

State name	Definition	Notes
Susceptible	Individuals who don’t have HPV infection nor benign cervical cancer.
hrHPV-infected	Individuals who are infected with HPV 16 or 18	Our model restricts high-risk HPV infection to be subtypes 16 and 18 only.
Benign cervical cancer (BCC)	High grade cervical lesions or worse (CIN2+), including cervical intraepithelial neoplasia grade 2-3 and cervical carcinoma without invasion of basement membrane.	ICO/IARC HPV Information Centre
Invasive cervival cancer (ICC)	The high-grade precancerous cells invade the basement membrane.	ICC stages range from 1 to 4 according to [FIGO-cancer-stage-2018-report]
Recovered	Recovered from invasive cervical cancer

Cause hierarchy of cervical cancer in GBD

GBD cervical cancer cause hierarchy

Cause name	GBD cause id	Level	Sequelae
All causes	c_294	0
All NCDs	c_409	1
Neoplasms	c_410	2
Cervical cancer	c_432	3	diagnosis_and_primary_therapy_phase_of_cervical_cancer (s_282), controlled_phase_of_cervical_cancer (s_283), metastatic_phase_of_cervical_cancer (s_284), terminal_phase_of_cervical_cancer (s_285)

Restrictions

The following table describes any restrictions on the effects of this cause (such as being only fatal or only nonfatal), as well as restrictions on the age and sex of simulants to which different aspects of the cause model apply.

Restrictions

Restriction Type	Value	Notes
Male only	False
Female only	True
YLL only	False
YLD only	False
YLL age group start	15 to 19	GBD age group id 8
YLL age group end	95 plus	GBD age group id 235
YLD age group start	15 to 19	GBD age group id 8
YLD age group end	95 plus	GBD age group id 235

Vivarium Modeling Strategy

Things to consider:

1. Within GBD 2017 data, there is no remission rate for invasive cervical cancer.

2. After diagnosis of invasive cervical cancer if a patient survives more than 10 years, they are considered cured for calculating disability. Additionally, per GBD 2017, the patients also do not have excess mortality rate after 10 years. In vivarium simulation model, we will remit them back to a recovered state.

3. Keep simulants in benign cervical cancer state and don’t go into remission after successful treatment unless literature tells us otherwise.

4. Most of the benign cervical cancer cases are resutling from a disease state called hrHPV-infected, where only high risk subtypes 16 and 18 of HPV infection are considered in our model. Though we do include the transition from susceptible state to benign cervical cancer state without high-risk HPV infection.

Todo

Add more assumptions and limitations.

Compartmental Diagram

Note

Regression of BCC will not be included in our Vivarium cause model, this is because we have little evidence to tell how it varies by age and subtypes of HPV infection. This measure may confound our assumption on duration of BCC. For simulants have BCC caused by high-risk HPV infection (subtypes 16 and 18), it brings more complexity to allow for two transition pathways: 1) regress from BCC_C_hrHPV to hrHPV-infected; 2) regress from BCC_C_hrHPV to Susceptible. Notably, the detection of BCC cases would change if we add it back for future model improvement.

State and Transition Data Tables

State Data

State	Measure	Value	Notes
Susceptible	prevalence	1 - (prev_hrHPV + prev_BCC_and_S_hrHPV + prev_BCC_and_C_hrHPV + prev_ICC_and_S_hrHPV + prev_ICC_and_C_hrHPV)	derived, used only at initialization
Susceptible	excess mortality rate	0	No EMR for susceptible state
Susceptible	disabilty weights	0	No disability weights for susceptible state
hrHPV-infected	prevalence	/ihme/costeffectiveness/vivarium_csu_cancer	used only at initialization
hrHPV-infected	excess mortality rate	0	assume zero death due to high risk HPV infection
hrHPV-infected	disabilty weights	0
BCC, S_hrHPV	prevalence (prev_BCC_and_S_hrHPV)	\(\text{prev\_BCC}\times(1-PAF\times\frac{\text{RR\_hrHPV}}{\text{RR\_hrHPV}-1})\)	prev_BCC, PAF, and RR_hrHPV are specified in Data sources
BCC, S_hrHPV	excess mortality rate	0	assume no EMR in BCC state
BCC, S_hrHPV	disability weight	0
BCC, C_hrHPV	prevalence (prev_BCC_and_C_hrHPV)	\(\text{prev\_BCC}\times\text{PAF}\times\frac{\text{RR\_hrHPV}}{\text{RR\_hrHPV}-1}\)	prev_BCC, PAF, and RR_hrHPV are specified in Data sources
BCC, C_hrHPV	excess mortality rate	0	assume no EMR in BCC state
BCC, C_hrHPV	disability weight	0
ICC, S_hrHPV	prevalence (prev_ICC_and_S_hrHPV)	\(\text{prev\_c432}\times(1-PAF\times\frac{\text{RR\_hrHPV}}{\text{RR\_hrHPV}-1})\)	prev_c432, PAF, and RR_hrHPV are specified in Data sources
ICC, S_hrHPV	excess mortality rate	\(\frac{\text{csmr\_c432}}{\text{prev\_c432}}\)
ICC, S_hrHPV	disability weights	\(\frac{\displaystyle{\sum_{s\in\text{s\_c432}}}\scriptstyle{\text{disability\_weight}_s\,\times\,\text{prev}_s}}{\displaystyle{\sum_{s\in\text{s\_c432}}}\scriptstyle{\text{prev}_s}}\)	weighted average of cervical cancer disability weight over all sequelae including ids s_282, s_283, s_284, s_285
ICC, C_hrHPV	prevalence (prev_ICC_and_C_hrHPV)	\(\text{prev\_c432}\times\text{PAF}\times\frac{\text{RR\_hrHPV}}{\text{RR\_hrHPV}-1}\)	prev_c432, PAF, and RR_hrHPV are specified in Data sources
ICC, C_hrHPV	excess mortality rate	\(\frac{\text{csmr\_c432}}{\text{prev\_c432}}\)
ICC, C_hrHPV	disability weights	\(\frac{\displaystyle{\sum_{s\in\text{s\_c432}}}\scriptstyle{\text{disability\_weight}_s\,\times\,\text{prev}_s}}{\displaystyle{\sum_{s\in\text{s\_c432}}}\scriptstyle{\text{prev}_s}}\)	weighted average of cervical cancer disability weight over all sequelae including ids s_282, s_283, s_284, s_285

S = susceptible; C = with condition

Transition Data

Transition	Source state	Sink state	Value	Notes
i_hrHPV	Susceptible	hrHPV-infected	hrHPV incidence	i_hrHPV is specified in Data sources.
r_hrHPV	hrHPV-infected	Susceptible	hrHPV clearance/remission	r_hrHPV is specified in Data sources.
i_BCC_hrHPV+	hrHPV-infected	BCC, C_hrHPV	\(\text{incidence\_BCC}\times(1-PAF)\times\text{RR\_hrHPV}\)	incidence_BCC, PAF, and RR_hrHPV are specified in Data sources.
i_BCC_hrHPV-	Susceptible	BCC, S_hrHPV	\(\text{incidence\_BCC}\times(1-PAF)\)	incidence_BCC and PAF are specified in Data sources.
i_hrHPV	BCC, S_hrHPV	BCC, C_hrHPV	hrHPV incidence
r_hrHPV	BCC, C_hrHPV	BCC, S_hrHPV	hrHPV clearance/remission
i_ICC	BCC, S_hrHPV	ICC, S_hrHPV	1 / duration_BCC	duration_BCC is specified in Data sources.
i_ICC	BCC, C_hrHPV	ICC, C_hrHPV	1 / duration_BCC	duration_BCC is specified in Data sources.
i_hrHPV	ICC, S_hrHPV	ICC, C_hrHPV	hrHPV incidence
r_hrHPV	ICC, C_hrHPV	ICC, S_hrHPV	hrHPV clearance/remission
r	ICC, S_hrHPV	Recovered	0.1 per person-years regardless of age	remission rate from ICC to R = 1 divided by duration of cervical cancer (10 years) = 0.1 per person-years regardless of age
r	ICC, C_hrHPV	Recovered	0.1 per person-years regardless of age	remission rate from ICC to R = 1 divided by duration of cervical cancer (10 years) = 0.1 per person-years regardless of age

prev = prevalence; i = incidence; r = remission; RR = relative risk; PAF = population attributable fraction

Data sources

Measure	Sources	Notes
crude-prevalence ratio of BCC	derived from marketscan data	see below for prevalence ratio calculation
prev_BCC	derived from incidence_c432 and duration of BCC	prev_BCC(age) = incidence_c432(age) * duration_BCC
duration_BCC	extracted from [Burger-et-al-2020-cause]	10 years
incidence_BCC	derived from incidence_c432	incidence_BCC(age) = incidence_c432(age + duration_BCC)
prev_c432	forecasted for future years 2020-2040	/ihme/costeffectiveness/vivarium_csu_cancer
csmr_c432	forecasted for future years 2020-2040	/ihme/costeffectiveness/vivarium_csu_cancer
incidence_c432	forecasted for future years 2020-2040	/ihme/costeffectiveness/vivarium_csu_cancer
remission_c432	GBD 2017	remission rate of cervical cancer = 1/10 per person-years for all ages
Disability weights for cervical cancer sequelae	[GBD-2017-YLD-Appendix-Cervical-Cancer]	total breast cancer disability weight over all sequelae with ids s_282, s_283, s_284, s_285
ACMR	forecasted for future years 2020-2040	/ihme/costeffectiveness/vivarium_csu_cancer
Population	demography for 2017	mid-year population
prev_hrHPV	derived from Abie’s dismod	/ihme/costeffectiveness/vivarium_csu_cancer
incidence_hrHPV	derived from Abie’s dismod	/ihme/costeffectiveness/vivarium_csu_cancer
remission_hrHPV	derived from Abie’s dismod	/ihme/costeffectiveness/vivarium_csu_cancer
RR_hrHPV	extracted from [Naucler-et-al-2007-cause]	relative risk of HPV 16/18 causing CIN2+ = 27.4 (95%CI 19.7 to 38.0)
PAF	derived from prev_hrHPV and RR_hrHPV	PAF = \(\frac{\text{prev\_hrHPV}\times(\text{RR\_hrHPV}-1)}{\text{prev\_hrHPV}\times(\text{RR\_hrHPV}-1)+1}\)

Todo

Describe dismod approach to estimate consistent rates for:

• prevalence, incidence, and remission of high risk HPV infection.

• prevalence, incidence, and regression of benign cervical cancer

Prevalence ratio calculation:

1. MarketScan research databases capture person-specific clinical utilization, expenditures, and enrollment across inpatient, outpatient, prescription drug and carve-out services. Currently GBD estimates bundle benign and in situ cervical and uterine neoplasms. Thus, we use external marketScan data source to calculate ratio of benign to malignant cervical cancer.

2. Outpatient year 2016 and 2017 data were pulled with following ICD 10 codes: C53 Malignant neoplasm of cervix uteri, C53.0 Malignant neoplasm of endocervix, C53.1 Malignant neoplasm of exocervix, C53.8 Malignant neoplasm of overlapping sites of cervix uteri, C53.9 Malignant neoplasm of cervix uteri, D06 (CIN3) Carcinoma in situ of cervix uteri, D06.0 Carcinoma in situ of endocervix, D06.1 Carcinoma in situ of exocervix, D06.7 Carcinoma in situ of other parts of cervix, D06.9 Carcinoma in situ of cervix, N87.1 (CIN2) Moderate cervical dysplasia, Z12.4 Encounter for screening for malignant neoplasm of cervix.

3. Non-medicare (age 0-65) & medicare (subset age 65+ only) were merged together to include all ages and limited to screened female patients only. After concatenating 2016& 2017 outpatient data, duplicates were removed based on enrolid and data were grouped by 5-year age band to align with GBD age pattern. Prevalence ratio was calculated using benign cervical cancer counts over invasive cervical cancer counts within each age group. Result shows younger age groups have larger ratio with wider uncertainty level. This ratio pattern is consistent with a study [Sun-et-al-2010] , that is BCC prevalence is higher than ICC prevalence for younger and middle age groups, but the specific ratio values are a little off.

prevalence ratio

Age Group	Prevalence Ratio
15_to_19	26.5
20_to_24	89.6
25_to_29	36.8
30_to_34	22.2
35_to_39	11.5
40_to_44	7.2
45_to_49	4.98
50_to_54	3.75
55_to_59	2.5
60_to_64	1.92
65_to_69	1.26
70_to_74	0.71
75_to_79	0.48
80 plus 0.5	0.5
all ages	8.83

Validation Criteria

Fatal outcomes

• Deaths

  • EMR_hrHPV = EMR_BCC = 0

  • ACMR = CSMR_c432 + CSMR_other

• YLLs

  • YLLs_hrHPV = YLLs_BCC = 0

  • YLLs_total = YLLs_c432 + YLLs_other

Non-fatal outcomes

• YLDs

  • YLDs_hrHPV = YLDs_BCC = YLDs_other = 0

  • YLDs_total = YLDs_c432

• Prevalence

  • add formula here once we identified marketscan data

• Incidence

  • add formula here once we identified marketscan data

Todo

1. Compare forecast data in 2020 against GBD 2017 (2019) results.

2. Compare prevalence, incidence, CSMR of cervical cancer, and ACMR over year with GBD age-/sex- stratification that calculated from simulation baseline to forecast data.

3. Check outcomes such as YLDs and YLLs in 2020 yield from simulation baseline against GBD 2017 (2019) all causes and cervical cancer results.

References

[GBD-2017-YLD-Appendix-Cervical-Cancer]

Supplement to: GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018; 392: 1789–858 (pp. 310-317)

[FIGO-cancer-stage-2018-report]

FIGO Cancer Report 2018: Cancer of the cervix uteri https://obgyn.onlinelibrary.wiley.com/doi/epdf/10.1002/ijgo.12611

[HPV-and-related-disease-2019-summary-report]

Ferlay J, Ervik M, Lam F, Colombet M, Mery L, Piñeros M, Znaor A, Soerjomataram I, Bray F (2018). Global Cancer Observatory: Cancer Today. Lyon, France: International Agency for Research on Cancer.

[Sun-et-al-2010]

Sun Z-R, Ji Y-H, Zhou W-Q, Zhang S-L, Jiang W-G, Ruan Q. Characteristics of HPV prevalence among women in Liaoning province, China. International Journal of Gynecology & Obstetrics 2010; 109: 105–9.

[Burger-et-al-2020-cause]

Burger EA, de Kok IMCM, Groene E, et al. Estimating the Natural History of Cervical Carcinogenesis Using Simulation Models: A CISNET Comparative Analysis. J Natl Cancer Inst 2020; 112: 955–63.

[Naucler-et-al-2007-cause]

Naucler P, Ryd W, Törnberg S, et al. HPV type-specific risks of high-grade CIN during 4 years of follow-up: a population-based prospective study. Br J Cancer 2007; 97: 129–32.