A United States study of self-collected urine vs clinician-collected cervical sample for HPV and cervical intraepithelial neoplasia grade two or worse detection: Cross-sectional agreement and diagnostic accuracy study

Abstract

Objectives

Urine can provide a non-invasive screening method for high-risk human papillomavirus (hrHPV) infections. We aimed to compare detection rates of hrHPV and cervical intraepithelial neoplasia grade two or worse (CIN2+) using first-void urine (FVU) and clinician-collected cervical samples. We assessed whether the time of day of FVU collection impacts hrHPV detection.

Methods

We recruited 188 participants, 30–65 years, before their colposcopy or routine screening from September 2020 through February 2022. Each collected FVU twice on the same day: at home and in the clinic before their appointments. We used prevalence-adjusted bias-adjusted kappa (PABAK) scores to compare hrHPV type-specific agreement, followed by McNemar's Exact test. We assessed both absolute and relative sensitivity and specificity to evaluate the accuracy of the collection techniques in predicting CIN2+ by hrHPV type.

Results

95 hrHPV infections were detected in 70/188 (37.2 %) women. The PABAK agreement for hrHPV16 detection between FVU and the speculum was 0.95 (95 % CI: 0.91, 1.0). The agreement for hrHPV detection was almost perfect between home or clinic urine collection and the speculum specimen. The accuracy of CIN2+ detection measured by the sensitivity ratio of the clinic first-void specimen compared to the speculum specimen for hrHPV16 was 1.00 (0.62, 6.20). Most specificity ratios for all other hrHPV types were anchored at 1.0 with narrow confidence intervals. The time of day for FVU gave similar “any hrHPV” detection rates (27.5 % vs 32 %, p > 0.05).

Conclusion

FVU hrHPV testing offers a potential cervical cancer screening alternative for women who cannot tolerate other collections.

Highlights

• •Urine and cervical samples have a 95 % Human Papillomavirus-type 16 agreement.
• •First-void urine equally detects human papillomavirus-16 regardless of time of day.
• •First-void urine and cervical samples equally detect pre-cancerous disease.

1. Introduction

Persistent infection with high-risk human papillomavirus (hrHPV) is a necessary cause of cervical cancer. The disease typically develops over years through a sequence of pre-cancerous lesions (e.g., cervical intraepithelial neoplasia grade two or three (CIN2 and CIN3)), which can be identified and removed through organized screening programs.

Many studies from around the world indicate that women prefer urine to vaginal self-sampling for cervical cancer screening, especially among those who refuse to undergo a speculum exam for screening, including groups of women in the United States (De Pauw et al., 2025; Guetterman et al., 2025; Hendrickx et al., 2025; Nyabigambo et al., 2022; Pils et al., 2022; Sy et al., 2017; Tranberg et al., 2025). Despite several research centers using the VALHUDES protocol (VALidation of HUman papillomavirus assays and collection DEvices for HPV testing on Self-samples and urine samples) (Van Keer et al., 2025) and investigating this globally, there is little evidence of the validity of urine HPV collection among US populations.

Urine serves as a readily accessible biomarker offering valuable information on many aspects of human health. Urine is also a high-pressure stream of fluid that carries secretions of desquamated cells from cervical, vaginal, and vulvar epithelia, providing a non-invasive method to screen for human papillomavirus (hrHPV) infections. Research in hrHPV detection from urine exploded at the time of the hrHPV vaccine studies two decades ago, looking for a non-invasive manner to monitor hrHPV infection status, especially in young children (Cheng et al., 2024). HRHPV detection in urine quickly spread to studies for screening or surveillance for cervical cancer (Jacobson et al., 2000; Pattyn et al., 2019; Vorsters et al., 2012; Hsiao et al., 2025),

Early studies demonstrated that, in addition to hrHPV DNA from desquamated epithelial cells, very small fragments of the hrHPV E1 gene could be detected in cell-free hrHPV from first void urine (FVU) specimens when the external genitalia were not cleansed before sampling (Pathak et al., 2014; Sahasrabuddhe et al., 2014). By 2021, results showed that FVU, not clean catch, was necessary to obtain sufficient DNA concentrations (National Institute for Health and Care Excellence (NICE), 2025). 10 or 20 mL of urine is sufficient volume for HPV detection. Using the Colli-pee device,seven mL of universal collection medium (UCM) is pre-filled, leaving 13 mL of FVU to fill the remaining tube (Teblick et al., 2021). The hrHPV detection assay cut-off for urine testing has not been optimized for the detection of CIN2+, but instead, studies used the cut-off for cervical samples when target amplification-based DNA hrHPV assays were used (Arbyn et al., 2022; Cho et al., 2022; Poljak et al., 2023) Modeling studies suggest that increasing the uptake of cervical cancer screening by 10 % could translate into a 40 % reduction in cervical cancer mortality (Knudsen et al., 2023). Hence, it is important to evaluate urinary-based screening with the potential to reach individuals that cannot tolerate vaginal or speculum-based screening options.

We aimed to evaluate the FVU and the speculum-collected cervical specimen as collection techniques for type-specific hrHPV detection for 1)to assess overall and HPV genotype-specific agreement between FVU and clinician-collected cervical samples; 2) to evaluate the absolute and relative clinical accuracy of HPV testing on FVU compared to cervical samples for the detection of CIN2+.

2. Methods

We define “first void” as the initial urine expelled from the urethra during voiding, regardless of the time of day (FVU). This is a cross-sectional diagnostic accuracy study of FVU collected by each person both at home and immediately prior to the clinic visit, where disease ascertainment occurs. Urine and cervical samples were paired for all comparisons except the second FVU done either at a morning or afternoon appointment.

2.1. Participant identification and first-morning FVU collection

The University of Michigan approved this clinical study (HUM00163299, 2019). We invited all women aged 30–65 who attended a clinical examination that included either a colposcopy or a speculum-based cervical cancer screening between September 15, 2020, and February 16, 2022, at least one week before their appointment (Fig. 1). After signing the electronic consent form, the study team mailed a collection kit to each participant.

Fig. 1.

Population enrolled by screening technique, among people assigned female at birth, ages 30–65 years, between 2020 and 2022, in Michigan, United States.

2.1.1. The Colli-pee device (Novosanis, Genotek, OraSure technologies, Wijnegem, Belgium) diverts the first 20 mL of urine into a screw-top tube containing seven

mL of preservative UCM, ensuring DNA stability [Appendix Fig. 1]. Participants brought their first-morning FVU (U1) collected at home that morning to the clinic. There was, at most, an 8-h time difference between the FVU (U1) and the clinician specimen, and at most, less than an hour between FVU (U2) and the clinician specimen.

2.2. Standard of care participant exam and second FVU collection

Upon check-in, the study team asked all participants to use an additional Colli-pee collection device in the clinic bathroom to obtain a second FVU specimen (U2). We recorded the time of this collection as either mid-morning or afternoon. After the clinic collected the FVU sample, participants proceeded to their scheduled clinical exams. We recorded the time of this collection as either mid-morning or afternoon. After the second FVU, participants proceeded to their scheduled clinical exams. Those attending a cervical cancer screening exam with normal or negative results participated solely in the hrHPV agreement study. Meanwhile, those undergoing colposcopy had biopsy results as a histologic endpoint, which contributed to the hrHPV accuracy study.

2.3. hrHPV detection procedure from urine

The hrHPV PCR test. The hrHPV polymerase chain reaction (PCR)-Mass Array assay is a validated laboratory-developed test (LDT) used for population screening (Brouwer et al., 2022a; Brouwer et al., 2022b) that includes a PCR step that amplifies the DNA E6 HPV region for any of the 15 discrete hrHPV types. The second PCR step is a single-base extension that discriminates between each hrHPV type and the competitor sequence added to the reaction. The latter step provides high sensitivity to the assay. The Matrix-Assisted Laser Desorption/Ionization Time-of-flight is the spectrometry instrument that measures/reads the fragments generated by the first steps.

Sample processing. The research lab transferred the FVU, composed of 13 mL urine and seven ml preservative, to a 50 ml conical tube and centrifuged for 10 min at 1600 ×g at room temperature prior to DNA extraction. We used a two ml input volume of FVU for DNA extraction. We isolated DNA from urine using two extraction kits from Qiagen: The QIAamp DNA Micro Kit and the QIAamp Viral RNA Mini Kit. The isolated DNA was determined to be of sufficient quality to amplify a 164-base pair fragment of the human β-globin gene. Both kits demonstrated high yields and purities for the recovered DNA, with no PCR inhibitors detected. The temperature at which the urine was stored did not impact the necessary DNA yield for analysis (El Bali et al., 2014). We adhered to the supplier's protocol for DNA extraction and standard PicoGreen quantitation.

2.3.1 We conducted hrHPV detection and type identification using multiplex PCR-Mass Array (Brouwer et al., 2022a; Brouwer et al., 2022b). The threshold for being positive was one molecule (copy) of HPV in the majority of replicate wells (3/4 or 2/3). This threshold was also employed for the cervical specimens described in Section 2.4.2.

2.4. hrHPV Detection Procedure from Cervical Samples

2.4.1 All cervical samples for research were collected during the speculum-based examination before the clinical exam (e.g., before the routine cervical cancer screening or the colposcopy).

2.4.2 The hrHPV PCR MassArray assay detects and identifies 15 high-risk hrHPV subtypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, and 73) as previously described (Brouwer et al., 2022a; Brouwer et al., 2022b; National Cancer Institute, 2025a). This in-house LDT targets hrHPV E6 DNA. The test includes interrogation of human glyceraldehyde-3-phosphate dehydrogenase as a control for sample DNA quality and assay validity for each hrHPV type. Additionally, the assay employs type-specific, multiplex, competitive PCR to amplify the E6 region of hrHPV, followed by a probe-specific single-base extension to differentiate between naturally occurring hrHPV present in the sample and the synthetic competitors included in the reaction. Matrix-assisted laser desorption/ionization-time of flight mass spectroscopy allows the separation of products on a matrix-loaded silicon chip array. When possible, the research lab runs all samples in quadruplicate, along with appropriate positive and negative controls for human use.

2.4.3. Colposcopy standards. Every woman referred for colposcopy had an exam with bright light and magnification that aided in the visualization of the transformation zone. Acetowhite epithelium after the application of 3 % acetic acid was biopsied if seen. If the examination did not show obvious lesions, at least one random biopsy was taken. An endocervical curettage was taken if the transformation zone was not satisfactorily visualized or disease extended into the os.

2.5. Statistical analysis

The index test was the hrHPV result on FVU; the comparator test was the hrHPV result from the cervical collection; and the reference standard for disease was CIN2+, as defined in 2.5.2.

2.5.1. hrHPV detection agreement

2.5.1.1 Urine and cervical specimens. The following sets of hrHPV tests were compared. FVU (U1) was compared to the clinical speculum-collected cervical specimen. The clinical FVU before the exam (U2) was compared to the clinical speculum-collected cervical specimen. The U1 was compared to the U2 and stratified by clinic urine collection time, either mid-morning or afternoon.

2.5.1.2 We presented the hrHPV agreement between urine and cervical sampling as a prevalence-adjusted bias-adjusted kappa (PABAK) statistic, adjusting for the potential low prevalence of the hrHPV type and any potential bias between the two sampling methods (Sahasrabuddhe et al., 2014). In the context of a 2-by-2 table, PABAK is simply the usual kappa statistic calculated for a revised table where the original overall (diagonal) concordance and (off-diagonal) discordance are replaced by the average concordance and discordance, respectively. Interestingly, these adjustments render PABAK to be linearly related to the overall concordance, estimated as

$$\text{PABAK}=2\left(\text{overall concordance}\right)-1$$

We further used Exact McNemar's test for paired proportions to estimate any systematic differences in marginal detection probability. The positive percent agreement and negative percent agreement were calculated as secondary measures.

The statistical methods used in the hrHPV agreement study are similar to those we have reported (Pathak et al., 2014). We defined PABAK thresholds as follows: 0–0.2 as poor agreement, 0.21–0.4 as fair agreement, 0.41–0.6 as moderate agreement, 0.61–0.8 as substantial agreement, and 0.81–1.0 as almost perfect agreement (National Institute for Health and Care Excellence (NICE), 2025). We used the Wilson 95 % confidence interval calculation.

2.5.1.3 Our bellwether indicator for hrHPV agreement was the PABAK of hrHPV16 detection from urine and the cervix. We, a priori, defined the agreement for hrHPV16 to be 90 % with a lower bound of its confidence interval at 81 %, which is nearly perfect agreement. This bellwether serves as an indicator to calculate all other hrHPV genotype agreements. If the hrHPV16 agreement did not meet this threshold, no further hrHPV agreement analysis would be done.

2.5.2. Accuracy of CIN2+ Detection

2.5.2.1 Disease definition. In our population, the diagnosis of CIN2+ disease means all tissue was histologically positive for p16 immunostaining, creating a uniform high-grade disease, a true cancer precursor endpoint. The presence or absence of CIN2+ disease defined the accuracy statistics. Women who went to colposcopy had a biopsy-proven outcome of CIN2+ or < CIN2. Women enrolled in the primary care clinics who did not need colposcopy were considered negative/normal and part of the non-diseased group. We explicitly included the women screening negative as part of the non-diseased group to avoid biasing the accuracy statistics by using only a positive screening test.

2.5.2.2 We compared the accuracy of CIN2+ detection by high-risk hrHPV genotype detection through three comparisons: between the U1 specimen and the U2 specimen (regardless of the time of collection), between the U1 hrHPV detection and the clinical speculum-collected cervical specimen, and between the U2 specimen and the clinical speculum-collected cervical specimen.

2.5.2.3 The statistical methods used in the hrHPV accuracy study are similar to those previously reported (Young et al., 2025). We measured the absolute and relative sensitivity and specificity of hrHPV detection from urine versus cervical samples for CIN2+ disease by genotype, where non-disease was considered <CIN2 on biopsy or a negative screening test. We also calculated the false positive and false negative rates (Latsuzbaia et al., 2023). We compared the sensitivity and specificity ratios for significant deviance from 1.0 or non-overlapping 95 % confidence intervals calculated with the Delta method.

We used SAS 9.4 for all statistical analyses (SAS Institute, Cary, NC).

3. Results

3.1. Population

3.1.1 We enrolled 188 participants, of whom 43 % (80/188) had a prior positive hrHPV test within the past five years (Table 1). There was no difference in age, race/ethnicity, gravidity, or parity between those with and without a history of hrHPV positivity. Although many races were represented, the majority were White (82 %). HPV vaccination did not differ by a history of hrHPV positivity. There was a significant difference in the proportion of women with a current hrHPV infection (U1, U2, speculum collected) by whether they had a past hrHPV infection (52 % vs 13 %, 56 % vs 14 %, and 63 % vs 20 %, chi-square > 30, p < 0.01).

Table 1.

Demographics of the population of people assigned female at birth, ages 30-65 years, between 2020 and 2022 in Michigan, United States.

	Positive speculum-collected cervical HPV test in the past 5 years
	Yes N = 80 (42.6 %)	No N = 108 (57.4 %)
	Mean (SD)	Mean (SD)
Age (yrs)	46.1 (10.3)	46.1 (11.0)
Race	N (%)	N (%)
Asian	2 (2.8)	3 (3.2)
Black	7 (9.9)	6 (6.4)
Middle Eastern North African (MENA)	1 (1.4)	1 (1.1)
Multiracial	1 (1.4)	0 (0.0)
Multiracial/Hispanic	1 (1.4)	2 (2.1)
White	65 (81.3)	89 (82.4)
White/Hispanic	0 (0.0)	1 (1.1)
Other	1 (1.4)	4 (3.7)
Other/Hispanic	2 (2.8)	2 (2.1)
Reproductive	Mean (SD)	Mean (SD)
Gravidity	2.4 (1.9)	2.2 (1.9)
Parity	1.5 (1.2)	1.7 (1.3)
Population Recruitment	N (%)	N (%)
Routine screening population	19 (23.8)	74 (68.5)
Colposcopy population	61 (76.3)	34 (31.5)
HPV positive detection⁎	N (%)	N (%)
Morning first void- U1-any hrHPV positive	37/71 (52.1)	12/95 (12.6)
First void of clinic urine -U2 – any hrHPV positive	40/71 (56.3)	15/105 (14.3)
Speculum-based collection- any hrHPV positive	49/80 (61.3)	20/108 (18.5)
HPV Vaccinated women	n = 9 (11 %)	n = 13 (12 %)
	Mean (SD)	Mean (SD)
Age at first prophylactic HPV vaccine	28.6 (15.6)	25.5 (12.5)
Vaccine valency	N (%)	N (%)
Gardasil4	5 (55.6)	6 (46.2)
Gardasil9	4 (44.4)	7 (53.8)
First HPV vaccine at age 18 years or younger	4 (44.4)	4 (30.8)
Completion of three doses	4 (44.4)	9 (69.2)

HPV means human papillomavirus.

hrHPV means high-risk human papillomavirus.

There was no difference in age, race/ethnicity, gravidity, or parity between those with and without a history of hrHPV positivity. Although many races were represented, the majority were White (82 %). HPV vaccination did not differ by a history of hrHPV positivity.

^⁎There was a significant difference in the proportion of women with a current hrHPV infection (U1, U2, speculum collected) by whether they had a past hrHPV infection (52 % vs 13 %, 56 % vs 14 %, and 61 % vs 19 %, chi-square > 30, p < 0.01).

Most routine screening population had no prior hrHPV infection in the past five years (68.5 %), whereas most attending colposcopy had prior hrHPV infections (76.3 %). 95 hrHPV infections were detected in 70/188 (37.2 %) women by at least one collection technique [Supplemental Table 2].

3.1.2. HRHPV detection by type. Of the 188 participants, 70 (37.2 %) women tested positive on one or more collection devices. Of those with a prior five-year history of hrHPV positivity, the percentage positive was significantly higher than the percentage without a prior five-year hrHPV positivity regardless of specimen collection, U1 (37/71 (52.1 %) vs. 12/95 (12.6 %)), U2 (40/71 (56.3 %) vs. 15/105 (14.3 %)), and the speculum (50/80 (62.5 %) vs. 22/108 (20.4 %), p < 0.01) (Table 1).

All fifteen single hrHPV-type infections were detected by at least one of the three devices. (Supplemental Table 1). Women had double infections in sixteen different combinations across all devices: one woman had three hrHPV types, and two women had four hrHPV types. No collection device detected all hrHPV types for all women.

hrHPV16/18 mono-infections were the most frequently detected types, followed by non-vaccine-specific mono-types (35, 39, 68, 51, 56, 59, 66, 73) (Supplemental Table 2a). Among the 22 hrHPV-vaccinated women, three had vaccine-specific hrHPV16 or hrHPV18 infections detected; all other hrHPV infections were not vaccine-associated (Supplemental Table 2b).

3.1.3 hrHPV Detection in the Population by Device. While 188 participants provided a speculum-based cervical specimen for hrHPV typing, only 166 (88.3 %) provided the U1 specimen, and 182 (96.8 %) provided U2 (Fig. 1). In sum, we have 166 paired samples between U1 and cervical samples; 182 paired samples between U2 and cervical samples, and 160 paired U1 and U2 samples Ninety-five (50.5 %) participants underwent a colposcopic biopsy, and there was no difference in the proportions of CIN2+ disease between those with and without past five-year hrHPV positivity (7/73 (9.6 %) vs 3/22 (13.7 %)). Most of the no-disease results (90.5 %) came from women with no past five-year hrHPV positivity, whose screening results were hrHPV-negative and cytology-negative for intraepithelial lesions or malignancy (NILM).

There were 72/188 (38.3 %) women with positive speculum-based hrHPV tests; 49/166 (29.5 %) women had first-morning first-void (U1) positive hrHPV tests, and 55/182 (30.2 %) women had positive clinic FVU (U2) hrHPV tests. Of the U2 tests, 23/82 (28.1 %) women had positive mid-morning tests, and 32/100 (32 %) had positive afternoon hrHPV tests (Table 1).

For type-specific infections, there was no difference in the percentage of women positive for hrHPV16 or hrHPV18. There were 27/188 (14.4 %) women whose speculum specimens were positive for hrHPV16, and 8/188 (4.3 %) women who were positive for hrHPV18, compared to 18/166 (10.8 %) who were hrHPV16-positive and 4/166 (2.4 %) who were hrHPV18-positive by U1 tests. The U2 tests resulted in 23/182 (12.6 %) hrHPV16-positive and 4/182 (2.2 %) hrHPV18-positive cases.

3.1.4. HRHPV vaccination. Similarly, there was no difference in the age at which the first prophylactic hrHPV vaccine was administered, nor in the number of individuals who received their vaccine at an age younger than 18, between those with and without a prior hrHPV-positive test. Among the 13 women who received three doses of the hrHPV vaccine, only one who began the series at 42 years had hrHPV16 detected as part of a dual infection; all others had non-vaccine-specific hrHPV types detected (Supplemental Table 2c).

Vaccinated women did not develop CIN2+. Among the detected hrHPV types associated with CIN2+ (n = 10, 5.3 %), most were single hrHPV types targeted by the hrHPV vaccine (Supplemental Table 2d).

3.2. Agreement

3.2.1 Agreements between collection devices in hrHPV type-specific detection are presented as PABAK scores (Table 2). For hrHPV16 or 18 types, there was no difference in detection by speculum, U1, or U2 by PABAK scores and a non-significant McNemar's Exact test. The mid-morning and afternoon collections did not differ in hrHPV detection, as compared with a chi-square test, because they were not paired samples. All other hrHPV types showed similar, almost perfect agreement with tight confidence intervals (Supplementary Table 3).

Table 2.

Human Papillomavirus detection agreement by collection techniques for types 16 and 18 among people assigned female at birth, ages 30–65 years, between 2020 and 2022 in Michigan, United States.

	U1 compared to U2	U2 am compared to U2 pm	U1 compared to speculum	U2 compared to speculum
	%, (95 CI)	%, (95 CI)	%, (95 CI)	%, (95 CI)
HPV 16
positive/positive	16	9	18	21
positive/negative	1	14	0	2
negative/positive	0	73	4	5
negative/negative	143	86	144	154
Positive Percent Agreement	100 (81, 100)		82 (62, 93)	81 (62, 92)
Negative Percent Agreement	99 (96, 100)		100 (97,100)	99 (95, 100)
Prevalence Adjusted Kappa	0.99 (0.96, 1.00)		0.95 (0.91, 1.00)	0.92 (0.87, 0.98)
Exact McNemar's p-value	1.00		0.13	0.45
Chi-square p-value		0.54
HPV 18
positive/positive	3	1	4	4
positive/negative	1	3	0	0
negative/positive	0	81	2	4
negative/negative	156	97	160	174
Positive Percent Agreement	100 (44, 100)		67 (30, 90)	50 (22, 79)
Negative Percent Agreement	99 (97, 100)		100 (98, 100)	100 (98, 100)
Prevalence Adjusted Kappa	0.99 (0.96, 1.00)		0.98 (0.94, 1.00)	0.96 (0.91, 1.00)
Exact McNemar's p-value	1.0		0.50	0.13
Chi-square p-value		0.63
Any HPV
positive/positive	40	22	46	53
positive/negative	5	32	0	1
negative/positive	2	60	12	15
negative/negative	113	68	108	113
Positive Percent Agreement	95 (84, 99)		79 (67, 88)	78 (67, 86)
Negative Percent Agreement	96 (91, 98)		100 (97, 100)	99 (95, 100)
Prevalence Adjusted Kappa	0.91 (0.85, 0.98)		0.86 (0.78, 0.93)	0.82 (0.74, 0.91)
Exact McNemar's p-value	0.45		<0.01	<0.01
Chi-square p-value		0.45

HPV means human papillomavirus.

U1 means the first-morning first-void specimen.

U2 means the clinic first-void urine specimen.

U2 am means the clinic first-void urine at a morning appointment.

U2 pm means the clinic first-void urine at an afternoon appointment.

Speculum means the clinician-directed cervical sampling.

PABAK means prevalence-adjusted, bias-adjusted kappa.

PABAK for HPV 16 was above the a priori 90 level, indicating nearly perfect agreement between U1:speculum and U2:speculum. Any HPV type prevalence-adjusted bias-adjusted kappa scores also showed nearly perfect agreement but did show statistical significance with the urine being marginally inferior to the speculum when all types in aggregate were evaluated. All HPV type-specific prevalence-adjusted kappa scores are nearly perfect, with no statistical significance, as shown by McNemar's Exact or chi-square testing (supplemental Table 2).

The agreement for detecting “any hrHPV” type using different collection techniques showed a significant difference between the speculum and either the first-morning first-void specimen (U1) or the clinic-collected FVU specimen (U2) (McNemar's Exact test, p < 0.01), favoring the speculum. Although “any hrHPV” types were detected by the speculum collection more than by either urine collection, the PABAK point estimate score for the comparison was above the 0.81 threshold for almost perfect agreement. Moreover, the U1 and U2 results were almost identical in their agreement (PABAK score = 0.91), with no difference in “any hrHPV” type detections by the time of U2 collection.

Missed hrHPV infections occurred by all techniques, including speculum and urine. Despite guidelines and pictorial instructions, genital cleansing (an incorrect collection technique) may have happened before urination among two women, leading to the discordant hrHPV detection between the speculum and any urine collected specimens.

3.3. Accuracy

3.3.1. The diseased population is defined as those with CIN2+ (n = 10, 5.3 %). The non-diseased population consisted of 85 (45.2 %) women with <CIN2 disease at colposcopy and 93 (49.5 %) with negative/NILM results at routine screening, resulting in a non-diseased group of 178 (94.6 %) of the study population

The accuracy was measured by sensitivity and specificity for hrHPV detection associated with CIN2+ disease, using different sampling techniques. The ratio of the comparative techniques' sensitivity and specificity was the outcome of interest, with overlapping 95 % confidence intervals to indicate non-significance.

HRHPV16 was not a sensitive detector of CIN2+ disease in our population, with sensitivities from the U1 collection being 25 % (0, 55) and the U2 and speculum collection being 30 % (2, 58). However, the sensitivity ratio between U1 and the speculum is 0.83 (0.11, 6.26), and between U2 and the speculum is 1.00 (0.62, 6.20), indicating that both urine and speculum collection techniques are equivalent in detecting CIN2+ caused by hrHPV16 (Table 3).

Table 3.

Accuracy of cervical intraepithelial neoplasia grade 2 or worse associated with types 16 and 18 detection by urine and speculum techniques among people assigned female at birth, ages 30–65 years, between 2020 and 2022 in Michigan, United States.

	U1 (n = 166)	U2 (n = 182)	Speculum (n = 188)
	% (95 CI)	% (95 CI)	% (95 CI)
HPV 16 (disease/nondisease)
positive/positive	2	3	3
positive/negative	16	20	24
negative/positive	6	7	7
negative/negative	142	152	154
Sensitivity	25 (7, 59)	30 (11, 60)	30 (11, 60)
Specificity	90 (84, 94)	88 (83, 92)	87 (81, 91)
False positive rate	10 (6, 16)	12 (8, 17)	13 (9, 19)
False negative rate	75 (41, 93)	70 (40, 89)	70 (40, 89)
Positive predictive value	11 (3,33)	13 (5, 32)	11 (4, 28)
Negative predictive value	96 (91, 98)	96 (91, 98)	96 (91, 98)
Comparisons	U1 compared to speculum	U2 compared to speculum	U1 compared to U2
Sensitivity ratio	0.83 (0.11, 6.26)	1.00 (0.16, 6.20)	0.83 (0.11, 6.26)
Specificity ratio	1.04 (0.76, 1.42)	1.02 (0.75, 1.39)	1.02 (0.74, 1.39)
False positive ratio	0.75 (0.39, 1.46)	0.86 (0.46, 1.62)	0.87 (0.44, 1.74)
False negative ratio	1.07 (0.26, 4.49)	1.00 (0.26, 3.92)	1.07 (0.26, 4.49)
HPV 18 (disease/nondisease	U1 (n = 166)	U2 (n = 182)	Speculum (n = 188)
positive/positive	2	1	2
positive/negative	2	3	6
negative/positive	6	9	8
negative/negative	156	169	172
Sensitivity	25 (7, 59)	10 (2, 40)	20 (6, 51)
Specificity	99 (96, 100)	98 (95, 99)	97 (93, 98)
False positive rate	1 (0, 5)	2 (1, 5)	3 (2, 7)
False negative rate	75 (41, 93)	90 (60, 98)	80 (49, 94)
Positive predictive value	50 (15, 85)	25 (5, 70)	25 (7, 59)
Negative predictive value	96 (92, 98)	95 (91, 97)	96 (92, 98)
Comparisons	U1 compared to speculum	U2 compared to speculum	U1 compared to U2
Sensitivity ratio	1.25 (0.14, 10.94)	0.50 (0.04, 6.44)	2.50 (0.19, 32.80)
Specificity ratio	1.02 (0.75, 1.39)	1.02 (0.75, 1.37)	1.00 (0.74, 1.37)
False positive ratio	0.38 (0.07, 1.89)	0.52 (0.13, 2.10)	0.73 (0.12, 4.40)
False negative ratio	0.94 (0.23, 3.84)	1.13 (0.31, 4.10)	0.83 (0.21, 3.34)
Any HPV (disease/nondisease)	U1 (n = 166)	U2 (n = 182)	Speculum (n = 188)
positive/positive	7	8	10
positive/negative	39	46	59
negative/positive	1	2	0
negative/negative	119	126	119
Sensitivity	88 (53, 98)	80 (49, 94)	100 (72, 100)
Specificity	75 (68, 81)	73 (66, 79)	67 (60, 73)
False positive rate	25 (19, 32)	27 (21, 34)	33 (27, 40)
False negative rate	13 (2, 47)	20 (6, 51)	0 (0, 28)
Positive predictive value	15 (8, 28)	15 (8, 27)	14 (8, 25)
Negative predictive value	99 (95, 100)	98 (95, 100)	100 (97, 100)
Comparisons	U1 compared to speculum	U2 compared to speculum	U1 compared to U2
Sensitivity ratio	0.88 (0.23, 3.34)	0.80 (0.22, 2.87)	1.09 (0.28, 4.33)
Specificity ratio	1.13 (0.81, 1.57)	1.10 (0.79, 1.52)	1.03 (0.74, 1.43)
False positive ratio	0.74 (0.47, 1.18)	0.81 (0.52, 1.25)	0.92 (0.57, 1.49)
False negative ratio	–	–	0.63 (0.05, 8.20)

HPV means human papillomavirus.

U1 means the first-morning first-void specimen.

U2 means the clinic first-void urine specimen.

Speculum means the clinician-directed cervical sampling.

Disease means CIN2+ on colposcopic biopsy.

Nondisease means a negative/normal routine screening result or < CIN2 on colposcopic biopsy.

95 CIs are calculated using the Wilson method for all but the ratio calculations, where the Delta method was used to calculate their 95 CI.

The comparative ratios for the aggregated HPV test, HPV 16 test, and HPV 18 test do not show statistically different test characteristics.

Conversely, the specificity for finding non-disease from a negative hrHPV16 test was 90 % (85, 95) for the U1 collection technique and 87 % (81, 92) for the speculum, resulting in a specificity ratio between U1:speculum of 1.04 (0.75, 1.42), and between U2:speculum of 1.02 (0.75, 1.39), indicating no difference between the urine and speculum collection techniques, with narrow confidence intervals.

Likewise, hrHPV18 was not a sensitive detector of CIN2+ disease in our population, with sensitivities from the U1 collection being 25 % (0, 55) and the speculum collection being 20 % (0, 45), resulting in a sensitivity ratio of 1.25 (0.14, 10.94) showing no difference between U1 and the speculum for detecting CIN2+ caused by hrHPV18, albeit with wider confidence intervals than for hrHPV16. The comparison of U2 to the speculum and U2 to U1 showed similar results.

The sensitivity and specificity ratios for all other 12 hrHPV types are presented in Supplementary Table 4, with comparable results.

The accuracy of urine in detecting CIN2+ caused by “any hrHPV” was not different from that of the speculum specimen (Table 3), with a sensitivity ratio of 0.88 (0.23, 3.34) for U1:speculum and 0.80 (0.22, 2.87) for U2:speculum. The sensitivity ratio was tight when comparing the U1 vs. U2 collections: 1.09 (0.28, 4.33), indicating no difference in the morning or afternoon for a FVU collection. Likewise, the specificity ratios between any two of the collection techniques had overlapping 95 % confidence intervals (1.13 (0.81, 1.57) for U1:speculum; 1.10 (0.79, 1.52) for U2:speculum, and 1.03 (0.74, 1.43) for U1:U2).

4. Discussion

Our study supports and updates US population-based cervical cancer screening research that compares FVU to speculum-collected specimens for potential clinical application in cervical cancer screening, particularly concerning hrHPV16/18-related CIN2+ disease (Sahasrabuddhe et al., 2014; Piyathilake et al., 2016; Rohner et al., 2020). Our study used the Colli-pee collection kit, whereas others have used a standard pot (urine cup); most studies used the FVU for analysis. Our study used the DNA-based HPV MassArray platform. In contrast, others have used Roche, Alinity m, TRUPCR, BD Onclarity, QIAscreen, Anyplex II, Abbott m200, Aptima, Trovagene, and Linear Array platforms for HPV detection.

The transformation of hrHPV to a CIN2+ lesion is time and type-dependent, with hrHPV16 and 18 accounting for the most rapid transformation. Hence, the agreement and accuracy for hrHPV16 and 18 are critical for evaluating specimen techniques. Our large clinical trial suggests that the FVU, collected either at home or in the clinic, exhibits very high agreement and accuracy with the speculum-collected cervical specimen for cervical cancer screening. The almost perfect kappa agreement for both hrHPV16 and 18 detection surpasses the agreement for “any hrHPV” detection, as observed in other studies (Rohner et al., 2020; Daponte et al., 2021; Miazga et al., 2024; Van Keer et al., 2022).

Our agreement results met our a priori clinical threshold for agreement for hrHPV16 and hrHPV18, and our accuracy results show no difference in relative sensitivity/specificity detection of CIN2+ caused by hrHPV16/18. The remaining high-risk hrHPV type-specific agreement and accuracy remained equivalent between urine and speculum collection. Our accuracy measurements had wide confidence intervals due to the small number of women with CIN2+. They cannot provide definitive conclusions regarding clinical accuracy, requiring a study like the Last Mile Initiative, where 500 women with CIN2+ are targeted for enrollment (National Cancer Institute, 2025b). Nevertheless, the analytic cut-off threshold for hrHPV detection in urine is not well established and may differ from that used for speculum-based cervical samples.

HRHPV-vaccinated women did have hrHPV16/18 infections detected in all specimens, but none had yet progressed into CIN2+. Non-vaccine type infections occurred in vaccinated women with CIN2+, indicating that type replacement is occurring. The need for a pan-valent hrHPV vaccine remains.

Generally, women do not prefer speculum exams for cervical cancer screening (Haro et al., 2024; Vinson et al., 2025). Vaginal swabs are currently Food and Drug Administration-approved (National Cancer Institute, 2025a) in the US for primary hrHPV cervical cancer screening and have recently been endorsed by the United States Preventive Services Task Force 2024 guidelines (US Preventive Services Task Force, n.d.) for use every five years for women aged 30 to 65 for routine screening. Urine testing would be more acceptable for populations that restrict exams with vaginal penetration (Guetterman et al., 2025). Our work provides compelling evidence that urine testing is viable for cervical cancer screening.

For a urine collection device to be effective, it must be easy and comfortable to use. Emphasis should be placed on avoiding genital pre-cleaning. The device must capture the FVU and preserve it immediately. The volume of the first voided urine should be at least 10 mL, combined with a preservative, to reach a total volume of 20 mL. It is essential to control the processing time and storage temperature. Optimizing hrHPV extraction and amplification may be necessary for improved sensitivity. For instance, using cell-free hrHPV extraction to detect very short, cell-free fragments of the E1 gene could enhance the sensitivity of a urine-based test (Sahasrabuddhe et al., 2014). These conditions all contribute to the development of a much-needed cervical cancer screening option for women.

5. Strengths and limitations

The strength of this study is the large US-based study of a combined referral and general population to evaluate the FVU for hrHPV detection associated with CIN2+ disease. However, although 188 women were enrolled, the small number of CIN2+ cases brings limitations. A larger number of CIN2+ women among the referral population would provide narrower confidence intervals and greater confidence in generalizability.

Another strength of our study was the rigor of the HPV testing platform, which we have shown to be non-inferior to the Cobas platform (Harper et al., 2025) even though it will never be commercially developed for clinical care.

We provided the raw data. Raw data allow a meta-analysis of all the studies to be completed. No study to date has reached the 500 CIN2+ threshold needed to provide the power to show equivalency between urine and cervical samples. With the aggregation of all studies, a meta-analysis can offer 500 CIN2+ cases, potentially before the clinical study is finished.

A limitation is the evaluation of the single Colli-pee collection device, which does not allow the evaluation of regular urine cups, as has been done in the UK (Cuzick et al., 2017).

6. Conclusion

FVU provides similarly consistent detection of hrHPV16/18, the most critical high-risk hrHPV type, compared to speculum detection. Non-invasive cervical cancer screening techniques have the potential to reach patients who cannot or will not participate in currently available options. Any screening technique developed must be of the same high accuracy for all women.

Patients consented according to the University of Michigan IRB requirements

No material is reproduced from other sources.

CRediT criteria

Conceptualization: MCO, APY, DMH, HMW, AS.

Methodology: MCO, APY, DMH, AS.

Validation: EAB, MLA, EAH, CEK,DMH, AS.

Formal Analysis: DC, AS, DMH.

Investigation: SAK, MO, EC, AL, LM, AM, JS, PR, PZ, JG, JC, KG, JP, RS, NS, EC,DMH.

Resources: DMH, HMW.

Data curation: DC, AS.

Writing- original draft – DMH, MCO.

Writing – review and editing – MCO, AYP, SAK, MO, EC, AL, AS, DC, LM, AM, JS, PR, PZ, JG, JC, KG, JP, EAB, MLA, EAH, CEK, RS, NS, EC, HMW, DMH.

Visualization: DMH.

Supervision: EAH, MLA, CEK.

Project administration: EAH, MLA, CEK.

Funding acquisition: DMH.

The MISSH1 study group has been included alphabetically at the end of the article

Martha L Alves, MSW, MPH mlalves269@gmail.com 0000-0001-6772-6306

Emma A Butcher MPH almane@med.umich.edu 0000-0002-3729-4159

Erin Calhoun, MD, erinmargaretcalhoun@gmail.com, 0009-0003-3752-7156

Elizabeth Campbell, MD elicampb@med.umich.edu 0009-0006-4978-7117

Jane Chargot, MD jchargot@med.umich.edu 0009-0002-4560-5452

Dongru Chen MS chendo@med.umich.edu 0000-0002-6527-0269

Christelle El Khoury, MD christelle.elkhoury@gmail.com 0000-0002-2562-4800

Jonathan Gabison, MD gabisonj@med.umich.edu 0000-0001-9594-4116

Kristina Gallagher, MD krisgall@med.umich.edu 0000-0001-5817-7077

Elizabeth A Haro, MPH elharo@med.umich.edu 0000-0003-3595-5432

Diane M Harper, MD MPH MS, b harperdi@med.umich.edu 0000-0001-7648-883X--

Scott A Kelley, MD scottak@med.umich.edu 0009-0002-8597-4834

Anna Laurie MD astanczy@med.umich.edu 0009-0008-7501-6441

Anna McEvoy, MD mcevoya@med.umich.edu 0000-0002-9500-6282

Leigh Morrison, MD morrisol@med.umich.edu 0000-0002-8493-6687

Marie Claire O'Dwyer, MB, Bch, BAO, MPH marieclo@med.umich.edu 0000-0002-8098-2571

Mutiya Olorunfemi, MD molorunf@med.umich.edu 0009-0007-9612-1702

Julie Prussack, MD jbkap@med.umich.edu 0009-0007-0368-6692

Pamela Rockwell, DO prockwel@med.umich.edu 0000-0002-7343-9132

Natalie Saunders, MD nsaunder@med.umich.edu 0009-0000-4998-3800

Jill Schneiderhan, MD jillsch@med.umich.edu 0000-0001-7417-276X

Ananda Sen, PhD, d anandas@med.umich.edu 0000-0002-9632-6704

Roger Smith, MD rogersmi@med.umich.edu 0000-0002-0757-8177

Heather M Walline, PhD hwalline@med.umich.edu 0000-0003-3790-5738

Alisa P Young, MD alisay@med.umich.edu-0009-0002-1116-955X

Philip Zazove, MD pzaz@med.umich.edu 0000-0003-3483-5607

CRediT authorship contribution statement

Alisa P. Young: Writing – review & editing, Validation, Methodology, Investigation. Marie Claire O'Dwyer: Writing – review & editing, Writing – original draft, Visualization, Supervision, Methodology, Investigation. Roger Smith: Writing – review & editing, Investigation. Dongru Chen: Writing – review & editing, Formal analysis, Data curation. Ananda Sen: Writing – review & editing, Formal analysis, Conceptualization. Heather M. Walline: Writing – review & editing, Validation, Resources, Methodology, Formal analysis, Data curation. Diane M. Harper: Writing – review & editing, Writing – original draft, Validation, Resources, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Ethics approval

HUM00163299.

Funding

UM1TR004404 – NCATS funding.

P30CA046592 – NCI Rogel Cancer Center funding.

PASD-RSG-23-1077156-01-PASD- American Cancer Society Research Scholar Grant.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material 1

Supplementary material 2

Supplementary material 3

Supplementary material 4

Supplementary material 5

Data availability

The data that has been used is confidential.