Correlation studies are essential for achieving World Health Organization cervical screening targets

 

Eric J Suba MD

Global Cervical Cancer Prevention Project

San Francisco, California, USA

eric.suba@gmail.com

Greta S Neethling MSc

Stellenbosch University

Cape Town, South Africa

greta.neethling@nhls.ac.za

Stephen S Raab MD

Global Cervical Cancer Prevention Project

San Francisco, California, USA

raabss@gmail.com

 

Mwesigwa Boaz CT(IAC),CMIAC

Mulago Hospital

Kampala, Uganda

mwesigwaboaz155@gmail.com

Dang Van Duong MD

Institute of Cancer Research Cooperation and Community Health Development

Hue City, Viet Nam

dvduong@gmail.com

Chiara Altavilla-Sugrue MBA MS

NYU Langone Health

Mineola, New York, USA

caltavilla2020@gmail.com

Amber D Donnelly PhD, MPH, SCT(ASCP)

University of Nebraska Medical Center

Omaha, Nebraska, USA

addonnelly@unmc.edu

Dana M Grzybicki MD PhD

Global Cervical Cancer Prevention Project

San Francisco, California, USA

tronagirl@ymail.com

Corresponding author:

Eric J Suba MD

Global Cervical Cancer Prevention Project

2295 Vallejo Street

San Francisco, California, USA 94123

eric.suba@gmail.com

phone: +415.922.2364

ABSTRACT

BACKGROUND: The World Health Organization global strategy to accelerate the elimination of cervical cancer as a public health problem sets two targets regarding cervical screening: 70% of the world’s women should be screened with a high-performance test, and 90% of women with cervical disease should receive appropriate follow-up care. Correlation studies are essential for assessing whether those targets are met.

METHODS: Correlation studies are methods of error analysis that compare cervical screening test results with follow-up cervical biopsy test results and assess whether those pairs of results are concordant, discordant, or missing. Correlation studies assess whether women with abnormal cervical screening test results have received appropriate follow-up care. Correlation studies determine whether cervical screening tests are high-performing by assessing whether cervical disease-detection rates are in control, relative to local benchmarks. We conducted correlation studies in South Africa.

RESULTS: Correlation studies show that loss to follow-up is crippling cervical cancer prevention efforts in South Africa. Root cause analyses that include fishbone diagrams and process mapping are currently in progress in South Africa in an effort to inform data-driven workflow modifications that may reduce loss to follow-up.

CONCLUSIONS: Correlation studies provide information critical to assuring the effectiveness of cervical cancer prevention efforts. Expanding information-sharing networks among clinical practitioners and sharing best practices for correlation studies and root cause analyses will help meet cervical screening targets set by WHO and thereby improve the effectiveness of global cervical cancer prevention efforts, particularly in low- and middle-income countries.

KEYWORDS: cervical cancer; low- and middle-income countries; quality management; correlation studies

BACKGROUND

Cervical screening prevents cervical cancer by detecting and eradicating pre-cancerous cervical lesions before they progress to cancers. Three types of cervical screening methods may be used to detect cervical pre-cancerous lesions and cervical cancers: visual inspection with 4% acetic acid (VIA), Papanicolaou cytology tests, and human papillomavirus (HPV) tests. Visual screening methods such as VIA were introduced to clinical practice by pathologist Walter Schiller during the 1930s. Visual screening methods suffered from insurmountable quality-assurance challenges and were largely abandoned after pathologist George Papanicolaou introduced the eponymous Pap test to clinical practice during the 1940s. HPV tests were introduced to clinical practice during the 1990s. The World Health Organization (WHO) Global Strategy for cervical cancer prevention() sets specific targets regarding global cervical screening policy: 70% of the world’s women should be screened with a high-performance test, and 90% of women with cervical disease should receive appropriate follow-up care.

WHO has advised that most of the world’s premature deaths can be prevented with simple, available interventions; what is not clear is how to make these interventions more widely available to the people who need them.() William Foege has observed that a lack of management skills appears to be the single most important obstacle to improving health throughout the world.() Quality assurance is a system of management activities that promotes higher level functioning of specific processes that occur in a health care system.() Measurements of health care outcomes, structures, and processes are cornerstones of effective quality assurance, but deciding which outcomes, structures, and processes to measure is a major problem.() Correlation studies are quality assurance methods that use process measurements to provide a wealth of data that may be used to improve cervical screening processes.() Correlation studies compare screening test results with follow-up biopsy results to determine whether those pairs of results are concordant, discordant, or missing.

In this article, we explain why correlation studies are essential for achieving WHO cervical screening targets. Because 95% of global cervical cancer deaths currently occur in low- and middle-income countries (LMICs),() we illustrate the use of correlation studies with real-world examples from South Africa. The WHO Global Strategy recommends that all countries should adopt HPV-based cervical screening as soon as it is feasible.() Consistent with WHO recommendations, South Africa is currently transitioning from primary cervical screening using Pap tests to primary cervical screening using HPV tests.()  

METHODS

Process measurements: sensitivity rates and disease-detection rates

The goal of a cervical screening test is to detect a true disease state of high-risk cervical neoplasia. High-risk cervical neoplasia includes cervical intraepithelial neoplasia (CIN) grade 2, CIN 3, and carcinoma and is often abbreviated as “CIN2/3+.” A screening test, by definition, does not define a true disease state. The definition of a true disease state is the function of a diagnostic test. Cervical tissue biopsies are the diagnostic tests used in cervical screening programs. The performance level of a screening test is defined in relation to the ability of that screening test to detect the true disease state present in the individual being screened.

In research settings, the most important process measurement to characterize the performance level of a cervical screening test is its sensitivity rate, which is defined as:

 

[endif]-->

A “true positive” is defined as a woman with a positive cervical screening test who has a follow-up biopsy showing CIN2/3+. A “false negative” is defined as a woman with a negative cervical screening test who has a follow-up biopsy showing CIN2/3+. To directly measure sensitivity rates, all women with negative screening test results must undergo follow-up diagnostic biopsies to define the true disease state.

However, in real-world settings, very few women with negative cervical screening tests undergo follow-up biopsies. In real-world settings, the most important process measurement to characterize the performance level of a cervical screening test is not its sensitivity rate but its disease-detection rate, which is defined as:

[endif]-->

 

An ideal screening test will demonstrate a sensitivity rate of 100% in any locality. However, there is no theoretical ideal for a disease-detection rate, which is locality-specific. In real-world settings, benchmarks provide the yardsticks for measuring how good something is. Each locality-specific process measurement has a corresponding locality-specific benchmark against which it should be regularly compared. Screening tests that demonstrate disease-detection rates that align with local benchmarks are considered “in control” and high-performing. If those locality-specific disease-detection rates do not align with their corresponding benchmarks, screening tests are considered “out of control” and quality-improvement interventions are recommended.

The Bethesda System of Pap test reporting specifies 5 categories of abnormal Pap test results:

1. ASCUS (atypical squamous cells of uncertain significance)
2. AGC (atypical glandular cells)
3. LSIL (low-grade squamous intraepithelial lesion)
4. HSIL (high-grade squamous intraepithelial lesion)
5. MALIGNANT (carcinoma, sarcoma, or lymphoma)

The Bethesda System specifies one category of negative Pap test result:

1. NILM (negative for intraepithelial lesion or malignancy).

The Pap test has features of both a screening test and a diagnostic test. Pap test results of HSIL or MALIGNANT may be considered diagnostic for a true disease state of CIN2/3+ with no requirement for additional testing. For those reasons, locality-specific HSIL/MALIGNANT Pap test-positivity rates may be considered surrogate measurements for the locality-specific prevalence of CIN2/3+.

The locality-specific HSIL/MALIGNANT Pap test-positivity rate is defined as:

[endif]-->

Locality-specific HSIL/MALIGNANT Pap test-positivity rates may be used as locality-specific benchmarks against which to compare locality-specific disease-detection rates, although the choice of benchmarks may evolve as clinical experience accumulates. All other things being equal, the demonstration of an increased disease-detection rate in a given locality will be the most important real-world criterion for the benefit of any novel screening approach.()

Correlation studies using Pap tests for primary screening or HPV triage: Principles (Table 1)

Correlation studies are methods of medical error detection that are used to evaluate failures in cervical screening.(6)

Because it is essential to prioritize detection of medical errors most likely associated with harm, it is essential to prioritize correlation studies for all Pap test results of HSIL or MALIGNANT.(Table 1) In any setting, correlation studies may be extended to other Pap test abnormalities as clinical experience accumulates.

Table 1 outlines principles of error detection and analysis using correlation studies. Table 1 is applicable to Pap tests that are used as primary screening tests. It is also applicable to Pap tests that are used for the triage of positive HPV tests in primary HPV screening programs. Each Pap test result along with its follow-up biopsy result constitute subjects of a correlation study. Each Pap test/cervical biopsy correlation study requires collecting 12 different data items: 

1. Pap test case number

2. Sign-out Pap test result

3. Sign-out cytologist

6. Review Pap test result

7. Review cytologist

8. Cervical biopsy case number

9. Sign-out cervical biopsy diagnosis

10. Sign-out cervical biopsy pathologist

11. Review cervical biopsy diagnosis

12. Review cervical biopsy pathologist

 

TABLE 1: PRINCIPLES OF ERROR DETECTION AND ANALYSIS USING CORRELATION STUDIES(5)
PAP TEST RESULT	CERVICAL BIOPSY RESULT (CORRELATION TYPE)	CORRELATION REQUIREMENT	POST-CORRELATION ERROR CATEGORY	POTENTIAL FOLLOWUP ACTION(S)
MALIGNANT	NONE (MISSING)	Review Pap test	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Cytology interpretation accurate: Patient has not received appropriate follow-up care	Root cause analysis +/- fishbone diagrams and process mapping*
	(DISCORDANT)	Review Pap test and biopsy	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Biopsy interpretive undercall	Correlate with pathologist ID and monitor
			Cytology and biopsy interpretations accurate; clinical sampling error possible	ALERT colposcopist of possible incomplete clinical evaluation*
	MALIGNANT (CONCORDANT)	NONE
HSIL	NONE (MISSING)	Review Pap test	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Cytology interpretation accurate: Patient has  not received appropriate follow-up care	Root cause analysis +/- fishbone diagrams and process mapping*
	(DISCORDANT)	Review Pap test and biopsy	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Biopsy interpretive undercall	Correlate with pathologist ID and monitor
			Cytology and biopsy interpretations accurate; clinical sampling error possible	ALERT colposcopist of possible incomplete clinical evaluation*
	³CIN 2 (CONCORDANT)	NONE
AGC	NONE (MISSING)	Review Pap test	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Cytology interpretation accurate: Patient has not received appropriate follow-up care	Root cause analysis +/- fishbone diagrams and process mapping*
	(DISCORDANT)	Review Pap test and biopsy	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Biopsy interpretive undercall	Correlate with pathologist ID and monitor
			Cytology and biopsy interpretations accurate; clinical sampling error possible	ALERT colposcopist of possible incomplete clinical evaluation*
	³CIN 2 (CONCORDANT)	NONE
LSIL	NONE (MISSING)	LOCALITY-SPECIFIC POLICIES
	(DISCORDANT)	Spontaneous regression possible; review of Pap test and biopsy optional	Cytology interpretive overcall	Correlate with cytologist ID and monitor
			Biopsy interpretive undercall	Correlate with pathologist ID and monitor
			Cytology and biopsy interpretations accurate; clinical sampling error possible	NONE (possible spontaneous regression)
	³CIN 1 (CONCORDANT)	NONE
ASC-US	NONE (MISSING)	LOCALITY-SPECIFIC POLICIES
	(CONCORDANT)	NONE
	³CIN 2 (DISCORDANT)	Review Pap test and biopsy	Cytology and biopsy interpretations accurate; clinical sampling error possible	Correlate with smear collector ID and monitor
			Cytology interpretive undercall	Correlate with cytologist ID and monitor
			Biopsy interpretive overcall	Correlate with pathologist ID and monitor
NILM	NONE (MISSING)	NONE
	(CONCORDANT)	NONE
	³CIN 1 (DISCORDANT)	Review Pap test and biopsy	Cytology and biopsy interpretations accurate; clinical sampling error possible	Correlate with smear collector ID and monitor
			Cytology screening undercall	Correlate with cytologist ID and monitor
			Biopsy interpretive overcall	Correlate with pathologist ID and monitor

*THIS ERROR IS ASSOCIATED WITH HARM.

Errors detected by correlation studies that are associated with harm may be targeted for further study. In Table 1, errors associated with harm are highlighted with BOLD ALL CAPS. In South Africa, patient loss to follow-up is a harm that is being targeted for further study in the form of root cause analysis. Root cause analysis is a qualitative method that focuses on system and human factors in order to discover the underlying cause of an error. Root cause analyses involve those who are the most familiar with the situation and, through a persistent series of "why" questions, determine the levels of health care processes at which failure occurs.() Results of root cause analyses then guide the planning of appropriate interventions aimed at preventing the error from happening again.

Root-cause analyses may include fishbone diagrams and process mapping. A fishbone diagram is a visual brainstorming tool that maps out all potential causes of a specific problem, organizing them into major categories (e.g. Laboratory, Patients, Personnel, Policies, Equipment, Communication) to reveal root causes rather than just symptoms. A fishbone diagram resembles a fish skeleton, with the problem as the head and the skeleton comprised of potential factors that are causing the problem organized into major categories. Once the cause categories have been identified, an interrogative approach is used to drill down into each category looking for possible causes of the problem.() A process map is a visual diagram of a workflow, showing steps and decisions from start to finish. Process mapping is conducted to help identify bottlenecks, reduce waste, and clarify responsibilities in the context of cervical screening workflows. In Cape Town, process maps are constructed using sticky notes that are annotated during brainstorming conversations conducted with those most familiar with the clinical situation.

Correlation studies using Pap tests: Practice (Table 2)

The South African National Health Laboratory Service (NHLS) is the country’s largest diagnostic pathology service, serving the public health sector and approximately 80% of the total population of 64 million people.() In 2010, the South African NHLS began the implementation and rollout of TrakCare, a laboratory information system through which all public-sector South African laboratory data can be accessed. All national test results are collected centrally and archived within a single Central Data Warehouse, which is a large server able to store, manage and analyze all laboratory information system data from all tests generated.

 

To begin a correlation study, it is necessary to select a locality where cervical screening is conducted and a time period during which Pap tests were reported at that locality. It is also necessary to select an age range for the women who will be included in the correlation study. The laboratory information system available at the selected locality is then searched to obtain the total number of Pap tests, whether abnormal or negative, reported during the selected time period from women in the selected age range. It is possible that some women will have had more than one Pap test reported during the time period that is being studied. If so, the number of repeat Pap tests reported is subtracted from the total number of Pap tests reported to arrive at the total number of women screened during the time period selected. The resulting number is entered as process measurement “A” in Table 2, corresponding to the total number of women screened. In South Africa, process measurement “A” is available on request from the Central Data Warehouse. The total number of women comprising the target age range in the selected locality is obtained from local population registers and entered as process measurement “B” in Table 2.    

The laboratory information system is then searched to obtain all reports containing Pap test results of HSIL or MALIGNANT that were issued for women in the selected age range during the selected time interval. It is possible that some women will have had more than one Pap test report with a result of HSIL or MALIGNANT during the selected time interval. Repeat Pap test results of HSIL or MALIGANT reported from the same woman during the selected time interval is subtracted from the total number of Pap test results of HSIL or MALIGNANT. The resulting total number of reports containing Pap test results of HSIL or MALIGNANT is entered as process measurement “C” in Table 2. It is assumed that appropriate follow-up care for women with Pap test results of HSIL or MALIGNANT includes colposcopy with cervical biopsy and/or endocervical sampling.

The laboratory information system is then searched to obtain all cervical biopsy reports that were issued for any woman with a previous Pap test result of HSIL or MALIGNANT. A time interval will elapse between the date when an abnormal Pap test result is reported and the date when the corresponding cervical biopsy result is reported. The length of that time period will vary. It is prudent to assume that time period may be as long as 24 months. The location of the laboratory that reports the cervical biopsy results following abnormal Pap test results may also vary. If more than one cervical biopsy report has been issued on any given patient, only the highest-risk cervical biopsy result is included in the correlation study. The resulting number of cervical biopsy reports containing any result is entered as process measurement “D” in Table 2, corresponding to the number of women who received appropriate follow-up care. In South Africa, process measurement “D” is produced by Central Data Warehouse analysts using custom software. The laboratory information search that produces process measurement “D” will yield all of the cervical biopsy reports that were issued within 24 months of a Pap test result of HSIL or MALIGNANT. Those reports form the basis for the error detection and analysis studies summarized in Table 1.

The cervical biopsy reports that constitute process measurement “D” are then searched for any reports containing diagnoses of CIN2/3+.  The number of non-duplicate reports containing diagnoses of CIN2/3+ is entered as process measurement “E” in Table 2, corresponding to the number of biopsy-confirmed cases of CIN2/3+ among women with Pap test results of HSIL or MALIGNANT.

 

TABLE 2: CORRELATION STUDIES ARE ESSENTIAL FOR ACHIEVING WHO PAP SCREENING TARGETS
LOCALITY	Target demographic group coverage rate (WHO target = 70%)	Follow-up care rate  (WHO target = 90%)	Benchmark disease-detection rate (WHO target: high-performing)	Measured disease-detection rate (WHO target: high-performing)
Locality name	A/B	D/C	C/A	E/A

A = total number of women screened

B = number of women in target age range

C = total number of women with Pap test results of HSIL or MALIGNANT

D = number of cervical biopsy reports containing any result among women with Pap test results of HSIL or MALIGNANT

E = number of cervical biopsy reports containing CIN2/3+ among women with Pap test results of HSIL or MALIGNANT

Table 2 provides the data needed to assess whether or not a Pap screening program in a particular locality is achieving WHO cervical screening targets. Correlation studies were conducted in South African and are summarized in Table 5 and Figure 1.

Correlation studies using HPV primary screening tests: Principles (Table 3)

Consistent with WHO recommendations,(8) South Africa is currently transitioning from primary cervical screening using Pap tests to primary cervical screening using HPV tests.

In South Africa, the primary screening algorithm for HPV tests recommends that:(9)

1. All women positive for HPV types 16, HPV 18, or HPV 45 are referred for colposcopy.
2. All women who are age 40 or over, who are negative for human immunodeficiency virus (HIV), and who are positive for any high-risk HPV type other than 16/18/45 are referred for colposcopy.
3. All women living with HIV (WLWHIV) who are positive for any high-risk HPV type other than 16/18/45 receive a reflex Pap test.
4. All women under age 40 who are positive for any high-risk HPV type other than 16/18/45 receive a reflex Pap test.

Because it is essential to prioritize detection of medical errors most likely associated with harm, it is essential to conduct correlation studies for all HPV tests that are positive for HPV types 16, 18, and 45. Correlation studies may be extended to other types of positive HPV tests as clinical experience accumulates.

Each abnormal HPV test along with its follow-up biopsy result constitute subjects of a correlation study. Each HPV test/cervical biopsy correlation study requires collecting 8 different data items:

1. HPV test case number

2. HPV test result

3. Review HPV test result

4. Cervical biopsy case number

5. Sign-out cervical biopsy diagnosis

6. Sign-out cervical biopsy pathologist

7. Review cervical biopsy diagnosis

8. Review cervical biopsy pathologist

Table 3 outlines principles of error detection and analysis using correlation studies and is applicable to primary HPV screening in South Africa.

TABLE 3: PRINCIPLES OF ERROR ANALYSIS USING CORRELATION STUDIES FOR PRIMARY HPV SCREENING IN SOUTH AFRICA
HPV TEST RESULT	CERVICAL BIOPSY RESULT (CORRELATION         TYPE)	CORRELATION REQUIREMENT	POST-CORRELATION ERROR CATEGORY	POTENTIAL FOLLOWUP ACTION(S)
POSITIVE FOR HPV TYPES 16/18/45 OR POSITIVE FOR ANY OTHER HIGH-RISK HPV TYPE AND AGE 40 OR OVER AND HIV NEGATIVE	NONE (MISSING)	Review HPV test result	HPV test negative on review: HPV laboratory error	Inform HPV laboratory and monitor
			HPV test positive on review: patient has been lost to follow-up	Root cause analysis +/- fishbone diagrams and process mapping*
	(DISCORDANT)	Review HPV test result	HPV test negative on review: HPV laboratory error	Inform HPV laboratory and monitor
			HPV test positive on review: clinical sampling error possible	ALERT colposcopist of possible incomplete clinical evaluation*
	³CIN 2 (CONCORDANT)	NONE
POSITIVE FOR ANY HIGH-RISK HPV TYPE OTHER THAN 16/18/45 AND UNDER AGE 40 OR WLWHIV	BECAUSE WOMAN IS REFERRED FOR CYTOLOGY TRIAGE, REFER TO TABLE 1
HPV NEGATIVE	NONE (MISSING)	NONE
	(CONCORDANT)	NONE
	³CIN 2 (DISCORDANT)	Review HPV test result	HPV test negative on review	Inform HPV laboratory and monitor
			HPV test positive on review: HPV laboratory error	Inform HPV laboratory and monitor

*THIS ERROR IS ASSOCIATED WITH HARM.

Errors detected by correlation studies that are associated with harm may be targeted for further study. In Table 4, errors associated with harm are highlighted in BOLD ALL CAPS.

 

Correlation studies using HPV tests: Practice (Table 4)

Performing correlation studies requires laboratory information systems that have the ability to retrieve results for cervical biopsies that are obtained from women following abnormal HPV test reports. Laboratory information systems vary in their abilities to collect those data items. To begin a correlation study, it is first necessary to select a locality where cervical screening is conducted and a time period during which Pap tests were reported at that locality. It is also necessary to select an age range for the women who will be included in the correlation study.

The laboratory information system available at the selected locality is then searched to obtain the total number of HPV tests, whether abnormal or negative, reported during the selected time period from women in the selected age range. It is possible that some women will have had more than one HPV test reported during the time period that is being studied. If that has happened, the number of repeat HPV tests reported is subtracted from the total number of HPV tests reported to arrive at the total number of women screened during the time period selected. The resulting number is entered as process measurement “F” in Table 4, corresponding to the total number of women screened. The total number of women comprising the target age range in the selected locality is obtained from local population registers and entered as process measurement “G” in Table 4.    

The laboratory information system is then searched to obtain the total number of HPV tests positive for HPV types 16, 18, or 45. Repeat HPV tests positive for HPV types 16, 18, or 45 reported from the same woman during the selected time interval is subtracted from the total number. The resulting total number of HPV tests positive for HPV types 16, 18, or 45 is entered as process measurement “H” in Table 4. It is assumed that appropriate follow-up care for a woman with an HPV test positive for HPV type 16, 18, or 45 includes colposcopy with cervical biopsy and/or endocervical sampling.(9)

The laboratory information system is then searched to obtain all cervical biopsy reports containing any result that were issued for any woman with an HPV test positive for HPV type 16, 18, or 45. A time interval will elapse between the date when an abnormal HPV test result is reported and the date when the corresponding cervical biopsy result is reported. The length of that time period will vary. It is prudent to assume that time period may be as long as 24 months. The location of the laboratory that reports the cervical biopsy results following abnormal HPV test results may also vary. If more than one biopsy report has been issued on any given patient, only the highest-risk cervical biopsy result is included in the correlation study. The resulting number of cervical biopsy reports is entered as process measurement “J” in Table 4, corresponding to the total number of women who were biopsied following an HPV test positive for HPV type 16, 18, or 45. The laboratory information search that produces process measurement “J” will yield all of the cervical biopsy reports that were issued within 24 months of an HPV test positive for type 16, 18, or 45. Those reports form the basis for the error detection and analysis studies summarized in Table 3.

The cervical biopsy reports that constitute process measurement “D” are then searched for any reports containing diagnoses of CIN2/3+.  The number of non-duplicate reports containing diagnoses of CIN2/3+ is entered as process measurement “K” in Table 2, corresponding to the total number of biopsy-confirmed cases of CIN2/3+ among women with positive HPV screening tests.

 

TABLE 4: CORRELATION STUDIES ARE ESSENTIAL FOR ACHIEVING WHO HPV SCREENING TARGETS
LOCALITY	Target demographic group coverage rate (WHO target = 70%)	Follow-up care rate  (WHO target = 90%)	Benchmark disease-detection rate (WHO target: high-performing)	Measured disease-detection rate (WHO target: high-performing)
Locality name	F/G	J/H	C/A (from Table 2)	K/F

F = total number of women screened

G = total number of women in target demographic group

H = total number of women with positive screening tests for HPV types 16, 18, or 45

J = total number of women who were biopsied following an HPV test positive for HPV type 16, 18, or 45

K = total number of biopsy-confirmed cases of CIN2/3+ among women with HPV tests positive for HPV types 16, 18, or 45

Table 4 provides the data needed to assess whether or not an HPV screening program in a particular locality is achieving WHO cervical screening targets. Because primary HPV screening is new in South Africa, locality-specific benchmarks for disease-detection rates are not available. For the time being, locality-specific benchmarks for Pap screening disease-detection rates should be used as locality-specific benchmarks for HPV screening disease-detection rates. Demonstration of an increase in locality-specific disease-detection rates is the best proof for the improved quality of any modifications in cervical screening approaches.()

RESULTS

Correlation studies using methods described for Table 2 produced results presented in Table 5:

 

TABLE 5: CORRELATION STUDIES ARE ESSENTIAL FOR ACHIEVING WHO PAP SCREENING TARGETS
LOCALITY/ TIME PERIOD/ TARGET AGE RANGE	Target demographic group coverage rate (WHO target = 70%)	Follow-up care rate  (WHO target = 90%)	Benchmark disease-detection rate (WHO target: high-performing)	Measured disease-detection rate (WHO target: high-performing)
Western Cape/March 2005 through December 2016/Women age 30+	78%()	43% - 88%(see Figure 1)	4.3%(15)	1.9% - 3.8%

 

Correlation studies using methods described for Table 1 and Table 2 indicated significant loss to follow-up in South Africa. Subsequently, correlation studies using methods described for Table 2 were conducted at 19 different localities in Cape Town. The results of those correlation studies are presented in Figure 1.

Figure 1: Follow-up care rates at difference health care facilities in Cape Town()

DISCUSSION
Correlation studies show that loss to follow-up is crippling cervical cancer prevention efforts in South Africa. On average, more than 5% of women screened through the National Laboratory Network in South Africa have Pap test results of HSIL or MALIGNANT, with some districts reporting that more than 10% of women screened have Pap test results of HSIL or MALIGNANT.(13) In the Western and Northern Cape, 55% of women with Pap test results of HSIL or MALIGNANT were lost to follow-up.() In Johannesburg, 6.2% of all women screened had Pap test results of HSIL or MALIGNANT and, of those, 73% were lost to follow-up.() Root cause analyses that include fishbone diagrams and process mapping are currently in progress in Cape Town in an effort to inform data-driven workflow modifications that may reduce loss to follow-up. In 2025, two of the health care facilities included in Figure 1 created positions for health care workers specifically tasked with assuring follow-up of women with Pap test results of HSIL or MALIGNANT. Future correlation studies will measure whether that intervention impacts follow-up care rates.

In order to make correlation studies more manageable, not all cases of CIN2/3+ detected by cervical screening efforts are included in the correlation methods described. Cases of CIN2/3+ are detected among women who present with Pap test results other than HSIL or MALIGNANT. Cases of CIN2/3+ are detected among women who present with HPV tests that are positive for HPV types other than 16, 18, or 45. Correlation studies may be extended to lower-grade Pap test abnormalities and/or to abnormal HPV test results other that HPV tests positive for types 16, 18, or 45 as clinical experience accumulates.

Performing correlation studies requires laboratory information systems that have the ability to retrieve results for cervical biopsies that are obtained from women following abnormal Pap test and HPV test reports. Laboratory information systems vary in their abilities to collect those data items. In Vietnam, reports for Pap test results, HPV test results, and cervical biopsy results are stored on a computerized laboratory information system. In Uganda, reports for Pap test results and cervical biopsy results are stored on a computerized laboratory information system, but HPV test results are stored as hard copies only. LMICs would benefit from sharing best practices for searching their laboratory information systems to collect data essential for achieving WHO cervical screening targets.

Visual screening experts emphasize that close monitoring of disease-detection rates is essential to maintain good standards of visual screening.() However, combining visual screening methods with immediate cryotherapy produces no biopsy results with which to measure locality-specific disease-detection rates.() Current WHO guidelines recommend that communities with visual screening programs ”transition rapidly” away from visual screening ”because of the inherent challenges with quality assurance,”() but do not suggest a destination for that transition in settings where HPV screening is unaffordable.(20) It appears that Pap screening is the prudent destination for that rapid transition for the following reasons:(20)

• It appears that HPV screening is not widely affordable in many LMICs. For example, in 2023, the US National Cancer Institute acknowledged “the WHO Cervical Cancer Elimination strategy calls for screening the majority of women with a high-performance HPV test twice in their lifetime. Realization of that goal using current commercial HPV tests is unlikely.”() The US National Cancer Institute has provided no timeline to affordability for HPV screening in LMICs.

• In 2001, medical leaders in East, Central and Southern Africa reported “95% of institutions at all health care levels in East, Central, and Southern African countries had the basic infrastructure to carry out Pap test screening.”() In 2005, the WHO Head of Cancer Screening emphasized “Our results clearly show that good-quality Pap smear screening can be implemented even in a rural setting of a developing country with reasonable investment.”()

• he 20th-Century campaign to roll out routine Pap screening services to millions of women in the United States required the sustained political will for nothing short of a total national mobilization.() Root cause analysis shows that critical real-world obstacles to successful cervical cancer prevention in LMICs involve people far more than technology, and are attributable to lapses of political will and quality management to which all preventive interventions are vulnerable.()

•  

• Pap tests remain important components for the management of women with positive HPV primary screening tests in primary HPV screening programs that have been implemented in high-income countries.() It is plausible that Pap tests will also remain important components for HPV primary screening programs in LMICs.

Communities with visual screening programs or no screening whatsoever could “transition rapidly” to good-quality Pap screening until better-quality HPV screening becomes widely affordable.(20) The implementation of good-quality Pap screening in LMICs will enable the development of centralized laboratories together with the associated information systems that will be needed to conduct correlation studies essential for global cervical cancer prevention efforts. Alternatively, communities in which HPV screening is unaffordable may continue with visual screening programs or no screening whatsoever pending the unlikely realization of affordable, better-quality HPV screening approaches that will not require any Pap tests or any Pap screening infrastructure. The latter scenario would be inconsistent with the appropriate public health goal of saving as many lives as quickly as possible.(20)

CONCLUSIONS: Correlation studies provide information critical to assuring the effectiveness of cervical cancer prevention efforts. Expanding information-sharing networks among clinical practitioners and sharing best practices for correlation studies and root cause analyses will help meet cervical screening targets set by WHO and thereby improve the effectiveness of global cervical cancer prevention efforts, particularly in low- and middle-income countries.

DECLARATIONS:

Ethics approval and consent to participate

The Health Research Ethics Committee of Stellenbosch University approved this study.

Consent for publication

Not applicable.

Availability of data and materials

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare that they have no competing interests.

Funding

This study was supported by the Brocher Foundation (Geneva, Switzerland).

Authors' contributions

This study was conceptualized by GSN, who provided overall supervision and conducted data collection, analysis, and interpretation. All authors reviewed the collected data and contributed to its interpretation. All authors provided critical feedback and helped shape the analysis and manuscript. EJS took the lead in writing the manuscript.

Acknowledgements

Not applicable.

References: