2.2.1 Model description.

Fig 1 describes the disease pathway of cervical cancer, along with interventions preventing the progress of the disease. It was adapted and modified based on existing cervical cancer microsimulation models, including Policy1-Cervix and MISCAN-cervix [24,25]. At a susceptible stage, HPV infections are most commonly acquired through sexual contact [26]. Most sexually active women and men eventually become infected; some may develop persistent infections. In Malaysia, the prevalence of high-risk HPV infection is estimated at 4.33% to 4.75% [13]. Acknowledging that HPV can lead to several types of cancers among both sexes [27], the model was bound and concentrated on HPV infection among females, focusing on cervical cancer disease.

Low-grade cervical abnormalities may be clinically evident in screening with productive HPV infection (low-grade squamous intraepithelial lesions (LSIL) or cervical intraepithelial neoplasia (CIN) grade 1 (CIN1)). They are usually temporary and disappear without intervention within 1–2 years [26]. The majority (90%) of contracted HPV infections typically become untraceable within a span of 1 to 2 years. Left untreated, a small percentage of HPV infections persist for over a year, increasing the likelihood of the development of cervical pre-cancer, such as high-grade squamous intraepithelial lesions (HSIL) or CIN grades 2 or 3 (CIN2/3), and the potential progression to cancer. As shown in the diagram, at this precancerous stage (CIN1 –CIN3), the abnormal cervical cells may either progress or regress spontaneously or intervene through treatment, either progressing a susceptible woman to the cancer stage or regressing to the susceptible pool. However, once it reaches cervical cancer, the regression no longer occurs [26].

Identifying HPV as the causative agent of cervical cancer has resulted in ground-breaking improvements in cervical cancer prevention, including preventive HPV vaccination. HPV vaccination has shown high efficacy and maximal effectiveness [28]. In Malaysia, the vaccine was given only to 13-year-old girls [3]. The quadrivalent vaccine, Gardasil-4 (Merck, Kenilworth, NJ, USA), which protects against HPV genotypes 6, 11, 16, and 18, was modelled since this vaccine was predominantly used in Malaysia [3]. Vaccinated cohorts received lifelong protection against HPV 16 and 18 in the model [29]. This represents 85–91% of the cohort who receives the HPV vaccination yearly [27,30], removed from the cancer pathway. Regarding the screening approach, Malaysia began transitioning from LBC tests to HPV testing in 2019. Besides high sensitivity and specificity, this method will be able to remove the obstacles encountered in the current screening method for cervical cancer [7]. In the model, women attending screening who were detected at the infectious or precancerous stage were assumed to receive optimised treatment or intervention, thus removed from the cancer path. At the same time, those lost to follow-up were returned to the initial disease stage.

The model combined the disease pathway with population dynamics, including declining birth rate and increasing life expectancy. The population began at birth, with the female cohort receiving and protected from the disease path at age 14 (upon completion of both vaccine doses). The population was stratified according to age groups to accommodate differing death, infection, disease progression rates, and policy approaches. The age group classifications are as follows: 0–13, 14 (completed vaccination group), 15–19, 20–29, 30–39, 40–49, 50–65, and above 65. While the diagram in Fig 1 is simplified for comprehensibility, all disease pathways and intervention components were stratified and modelled accordingly by each age group.

2.2.2 Model assumptions.

The baseline model was simulated based on several assumptions, as follows:

1. Vaccination assumptions: Two types of HPV vaccines were approved in the MOH drug formulary: Cervarix (GlaxoSmithKline, London, UK) and Gardasil-4 (Merck, Kenilworth, NJ, USA), with the vaccine selection reviewed periodically upon purchase contract renewal [3,31]. The efficacy used in the model was based on Gardasil-4 with 100% protection against four HPV genotypes, 16/18/6/11, without cross-protection for other genotypes [29], considering the more extended usage of Gardasil-4. The full-dose vaccination was assumed to provide life-long protection [29]. Two doses of HPV vaccination were given to girls aged 13 with a 6-month interval between doses. Therefore, complete protection and exiting the disease path were assumed to be achieved at age 14. The vaccination coverage was set at 90% yearly population coverage.
2. Screening assumptions: Following the national screening guideline, the model simulated screening for women aged 30 to 65 [7]. The screening coverage was assumed to be equally distributed across different age groups. Pap Smear was introduced into the model in 1998 following the launch of the national screening program and converted to LBC from 2014 onwards. The screening rate was set at 25% in 2015, mimicking actual practice, including three-yearly screenings following a negative LBC result. HPV testing was introduced into the model starting in 2019, using a screen-treat path. The screen-triage-treat was tested in the alternative scenario analysis. The HPV screening cycle was every five years following a negative result, as currently practised [7].
3. Herd immunity: Malaysia achieved a high vaccination coverage (90%), contributing to a 100% reduction in HPV prevalence among the female population [32].
4. Susceptibility to HPV infection: The susceptibility rate was based on National Health Morbidity Survey (NHMS) data from 2011–2019, whereby about 76–80% of women who were ever married or cohabiting with a partner were assumed to be at risk [33–35].
5. Adherence to follow-up following positive screening: Assumed to be at a rate of 80 to 90%. Lower commitment to follow-up was simulated as an alternative scenario.
6. Screening costs: MYR 58.66 (USD13.33) per LBC test and MYR 102.86 (USD23.38) per HPV test done through in-house laboratory processing. MYR 84.55 (USD19.22) per HPV test done through outsourcing the laboratory processing was done as an alternative scenario. The details in the cost calculation are reported elsewhere [36]. The monetary figures in this study were denominated in the local currency unit (MYR) and converted to USD, utilising the World Bank’s official exchange rate for 2022 (1 USD equivalent to 4.4 MYR) [37]. Constant screening costs without discounted rates were applied in the simulation.
7. Disease pathway: The transition rates were averaged across all high-risk HPV genotypes.

2.2.3 Data preparation.

Whenever possible, Malaysian secondary data was used in the model. Data on the prevalence of HPV infection between 2019–2021 was analysed from the Malaysian cervical cancer screening database by the Family Health Development Division, MOH Malaysia (BPKK), published elsewhere [13]. Information on cervical cancer incidence and stratification by age group was extracted from the Malaysian National Cancer Registry Report 2012–2016 [2] with additional data provision by the Department of National Cancer Registry for 2017–2018. The impact of treatment and survival rate at different cancer stages were based on the Malaysian Study on Cancer Survival (MySCan) 2018 report [38]. Vaccination data from 2010–2021 was from the Family Health Development Division, MOH Malaysia, while population data was obtained from the Department of Statistics Malaysia (DOSM) [39]. Data on women susceptible to HPV infection was derived from the National Health and Morbidity Survey (NHMS) 2011–2019 [33–35]. Permission and access to use otherwise not publicly available data was obtained. Most disease progression and regression rates were extracted from various evidence in the literature. Other data for Malaysia was estimated by a group of experts consisting of two public health specialists, two obstetrics and gynaecologists, one medical officer, two nurses, two cervical cancer data custodians, and one academician (expert in cervical cancer studies). Details on the list of data sources can be referred to in Appendix A in S1 File.

2.2.4 Model development.

The system dynamics model was built using Vensim Professional version 9.3.4 (Ventana, 2022). Fig 2 represents the stock-flow diagram of the model based on the conceptual illustration described earlier, segregated into different sectors for comprehensibility.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Stock-flow models of the Malaysian cervical cancer progression, segregated according to the different parts of the model.

https://doi.org/10.1371/journal.pone.0307880.g002

The primary differential equations of integral calculus approximations were used in the system dynamics model as shown in Eq (1) below: (Eq 1)

Where t represents the stock, t₀ represents the initial value of the stock, and dt is a time step that represents the rate of change concerning time [15].

The stock variables are state variables, illustrated as rectangles in Fig 2 [15]. It accumulates the number of women at various stages in the model by balancing between those entering and leaving the stocks. The inputs and outputs of stocks are linked by flow variables, which indicate rate variability, shown as double-lined arrows. Single-lined arrows show interrelationships between variables. Stock variables include women population from birth till the oldest age group, vaccinated cohort, precancerous and cancerous stages, and cancer incidence. Some of the main equations in the model are shown as follows.

Women from different age groups in the population who were exposed to the infection were calculated as in Eq 2: (Eq 2)

Women infected with HPV were expressed in Eq 3 as: (Eq 3)

The cohort receiving HPV vaccination was expressed in Eq 4 as: (Eq 4)

Women screened every year were calculated in Eq 5 as follows: (Eq 5)

Women progressing to cervical cancer were calculated in Eq 6 as follows: (Eq 6)

Screening at the infectious and precancerous stages was drawn as stocks in the model to allow women successfully detected and treated for having either the infection or precancerous cells to leave the path progressing towards cancer development.

Auxiliary variables are intermediate factors whose values represent significant effects and are calculated by the model [15]. However, it does not capture the cumulative population figures for that specific state; instead, it reflects the count at one point in time. In the model, susceptible women attending screening but found negative were drawn as auxiliaries as the women did not leave the original stock, but need to be calculated to estimate the total number of screening tests needed to be performed. Other auxiliaries include the prevalence of HPV infection, the total number of screening tests performed, cancer staging, and the number of deaths due to cancer. The constants include infectious rates, progression and regression rates at the precancerous stage, progression rate at the cancerous stage, follow-up adherence rate, sensitivity detection rate of screening tools, screening coverage, screening rate, vaccination coverage, vaccination efficacy, and survival rates of different cancer stages. Birth, death, and ageing rates were entered using a lookup function to accommodate the changing population dynamics.

Altogether, 91 parameters were used in the model. Appendix A in S1 File summarises all main parameters with the evidence source. The crucial parameters comply with the HPV-FRAME Standards of modelling HPV vaccination and cervical cancer screening [40] (Appendix B in S2 File). The model was simulated for 100 years from 1990, as the national screening program was launched in 1998 in Malaysia, to 2090, to allow enough time for elimination to be included in the simulation. The time step was set at one year, and calibration was done in years.

2.2.5 Model calibration and validation.

The developed model was presented and discussed with two obstetrics gynaecologists and one academician who validated that the disease and intervention structures were accurate through four separate sessions between September 2022 and April 2023. Subsequently, structural and behavioural testing were conducted to validate the model. For structural validation, the unit-checking function in Vensim confirmed that the units used throughout the model were error-free [15], indicating that the equations for all paths were correctly specified according to the variables involved. Next, the extreme condition test was performed, whereby scenarios where no vaccination and no screening were performed were simulated. The model demonstrated a continuous increase in cervical cancer incidences as the population grew, confirming the expected behaviour.

The model was calibrated and validated against several data: Malaysian female population data, national HPV infection prevalence data, number of screening tests performed, national age-standardised cervical cancer incidence rates, national cervical cancer incidences, and global estimation of cervical cancer elimination. To measure the closeness between the simulated trend and actual data, the statistical method Mean Absolute Percent Error (MAPE) was calculated using the following formula in Eq 7: (Eq 7)

A value of less than or equal to ten per cent (≤ 10%) was considered a good fit [41,42]. Fig 3 compares simulated behaviours of cancer incidence cases and rates with the historical data. Both parameters had ≤10% of the mean absolute percentage errors with the Malaysian data, with behavioural progression akin to actual behaviours. Differences in the fluctuation patterns were most likely due to the considerable yearly variations in the actual data. In contrast, most parameters in the model were either held constant or changed in a linear or stepwise pattern. The validation of other main parameters, including population growth, the prevalence of HPV infection, the number of screening tests performed, and the distribution of cervical cancer by age groups, are reported in Appendix C in S3 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Structural and behaviour validations of the main parameters in the model.

Note: MNCR = Malaysian National Cancer Registry.

https://doi.org/10.1371/journal.pone.0307880.g003

2.3.1 Establishing the scenarios.

Once the model development, calibration and validation processes were completed, the model was considered stable to simulate different policy scenarios. Continuous and iterative expert engagement sessions were conducted to establish several scenarios to be projected by the model. Four sessions were conducted between September 2022 and April 2023 with seven experts (two public health specialists, one gynaecologist, two medical officers, and two nurses). Considering the capacity and potential, the scenarios were set based on current screening performance and expected screening expansion deemed achievable for Malaysia. The Action Plan Towards the Elimination of Cervical Cancer in Malaysia 2021–2030 guided the engagement sessions [6]. The scenarios were divided into two sections: first, to look at the impact of vaccination, followed by the effects of screening.

2.3.2 Assessing the impact of HPV vaccination.

To evaluate the impact of vaccination in the absence of screening, the year of elimination was projected for the following scenario: Annual HPV vaccination of girls aged 13 at a population coverage of 90% compared to 70%, all without implementing screening.

To examine the implications of the absence of HPV vaccination, the simulation for the following scenario was conducted: LBC screening at 25% coverage versus 70% coverage, with no HPV vaccination.

To investigate the combined effects of HPV vaccination and screening, the simulation for the following scenario was conducted: Annual HPV vaccination of girls aged 13 at 90% population coverage and LBC screening at 25% coverage. This scenario also served as the baseline in subsequent assessment of the impact of screening.

2.3.3 Assessing the impact of screening.

Subsequently, four scenarios incorporating the targeted year for a full transition from LBC to HPV testing, along with an increase in screening coverage, were simulated as follows:

• In Scenario A, the transition from LBC to five-yearly HPV testing occurred gradually till a complete transition was achieved in 2030, with the screening coverage retained at 25% from 2023 onwards.
• In Scenario B, the transition from LBC to five-yearly HPV testing occurred gradually till a complete transition was achieved in 2030, with the screening coverage now increased and retained at 40% in 2030.
• In Scenario C, the transition from LBC to five-yearly HPV testing occurred gradually till a complete transition was achieved in 2030, with the screening coverage increased up to and retained at 70% in 2040.
• In Scenario D, the transition from LBC to HPV testing occurred gradually till a complete transition was achieved in 2030, with the screening coverage now increased up to and retained at 70% in 2040. In this scenario, HPV testing was done twice in a lifetime, first by the age of 40 and second by the age of 50, in concordance with WHO recommendations [1]. The ages for the two screening episodes were adjusted to accommodate the wider age groups screened in Malaysia between 30–65 years.

The screening approach was assumed to be sustained at the same coverage following elimination in all scenarios. The vaccination rate was held constant at a yearly 90% population coverage. In addition, an ideal scenario with 100% HPV vaccination coverage and a high screening rate of 90% was simulated.

Two additional simulations were added to each scenario, projecting:

• the delay in transitioning, whereby the full transition from LBC to HPV was delayed to 2040 and
• the delay in screening coverage increments, whereby achieving 40% and 70% coverage was delayed to 2040 and 2050.

As listed in Table 1, multiple simulated outcomes were examined across scenarios. It included a baseline setting, which mirrors the screening approach through LBC alone at a screening coverage retained at 25%. This baseline configuration served as a reference point, reflecting Malaysia’s screening practices before the introduction of HPV testing in 2019. The baseline scenario was also employed as a comparative benchmark against all transitional scenarios, allowing assessment of the changes brought about by transitioning to HPV testing instead of persisting with LBC alone.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of scenarios and outcomes simulated.

https://doi.org/10.1371/journal.pone.0307880.t001

2.3.2 Simulating alternative scenarios in screening practice.

To address several uncertainties, a simulation of the following alternative scenarios to screening practice was performed:

• A combination of screen-treat for HPV 16/18 and screen-triage-treat pathway for non-HPV 16/18 following a positive HPV testing. Three approaches in Scenario A were used for comparison: i) HPV was done five-yearly HPV screening at up to 40% coverage, ii) HPV was done twice a lifetime, up to 70% coverage, and iii) HPV was done five-yearly at 70% coverage.
• A lower adherence to follow-up upon a positive screening test at 50–60% (as opposed to 80–90%). In this approach, 70% of women screened with HPV five-yearly were used for comparison.

In addition to these scenarios, sensitivity analyses were conducted on several crucial parameters, including a range of non-adherence rates to follow-up after a positive screening test, varying HPV infection rates across different age groups, and varying unvaccinated rates. The sensitivity analyses were performed based on the baseline model whereby screening was done via LBC sustained at 25% coverage.

3.2.1 Impact towards the year of cervical cancer elimination.

We analysed four scenarios compared with a baseline scenario and the additional effect of delays in transition and increasing screening coverage. In Scenario A (Fig 5, Diagram A), the model projected a complete transition from three-yearly LBC to five-yearly HPV testing by 2030, with the screening coverage retained at 25% would accelerate the elimination of cervical cancer in Malaysia by two years as opposed to the baseline (2068 vs. 2070). A ten-year delay in transitioning from LBC to HPV testing would delay the elimination by one year (2069 vs. 2068).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The simulated year of elimination based on different transition and screening expansion scenarios.

https://doi.org/10.1371/journal.pone.0307880.g005

In Scenario B (Fig 5, Diagram B), a complete transition from a three-yearly LBC to a five-yearly HPV test combined with increased screening coverage to 40% would accelerate the elimination by 14 years compared to the baseline (2056 vs. 2070). However, a ten-year delay in transitioning from LBC to HPV testing would delay the elimination by 13 years (2069 vs. 2056). In contrast, a delay in increasing the screening coverage from 25% to 40% would result in no delay of elimination.

In Scenario C (Fig 5, Diagram C), a complete transition from three-yearly LBC to five-yearly HPV testing combined with increased screening coverage to 70% would accelerate the elimination by 18 years compared to the baseline (2052 vs. 2070). A delay in in transitioning from LBC to HPV testing would result in delaying the elimination by four years (2056 vs. 2052). In contrast, a delay in increasing the screening coverage from 25% to 40% would result in a three-year delay (2055 vs. 2052).

In Scenario D (Fig 5, Diagram D), a complete transition from three-yearly LBC to twice-in-a-lifetime HPV testing in combination with increased screening coverage to 70% would accelerate the elimination by 14 years as opposed to the baseline (2056 vs. 2070). A delay in transitioning from LBC to HPV testing would delay the elimination by 12 years (2068 vs. 2056). In contrast, a delay in increasing the screening coverage to 70% would result in no elimination delay.

3.2.2 Impact towards the costs, number of screening tests needed and cancer avoidance.

This study extended the scenario analysis by projecting the impact towards different outcomes. As shown in Table 2, the total costs required to perform the screening until each scenario’s respective elimination year were either similar or cheaper in the baseline (MYR 1.39 billion, equivalent to USD 315 million) than all other scenarios. The exception was when HPV testing was done twice in a lifetime at 70% coverage (MYR 1.13 billion, equivalent to USD 257 million), which was projected to save MYR 254 million or USD 58 million. In addition, the total costs for sustaining each approach until 2070 were higher in all approaches than the cost of maintaining the baseline.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The detailed costs, screening tests needed to perform, and differences in cervical cancer incidence based on different approaches.

https://doi.org/10.1371/journal.pone.0307880.t002

Despite the increased cost, the number of tests necessary to perform, calculated from 2023 to the respective year of elimination, was lower in all approaches compared to the baseline, except if HPV testing was done five-yearly and at 70% coverage. This will require, on average, an additional 131,505 (26%) tests per year instead of the baseline (Table 2). In contrast, HPV testing increased up to 70% coverage twice in a lifetime would require 144,868 fewer tests than baseline to achieve elimination.

All approaches would retain a similar number or avoid new cervical cancer cases instead of baseline. The most prominent avoidance was seen with HPV five-yearly combined with 70% screening coverage, leading to a 5,987 lesser incidence between 2023 and 2070, as opposed to if baseline were to continue. Although fewer, HPV twice in a lifetime would reduce the incidence compared to the baseline by 1070 cases by 2070.