Human Papillomavirus Intermittence and Risk Factors Associated With First Detections and Redetections in the Ludwig-McGill Cohort Study of Adult Women

Abstract

Background

We assessed the incidence and risk factors for first detection and redetection with the same human papillomavirus (HPV) genotype, and prevalence of cytological lesions during HPV redetections.

Methods

The Ludwig-McGill cohort study followed women aged 18–60 years from São Paulo, Brazil in 1993–1997 for up to 10 years. Women provided cervical samples for cytology testing and HPV DNA testing at each visit. A redetection was defined as a recurring genotype-specific HPV positive result after 1 or more intervening negative visits. Predictors of genotype-specific redetection were assessed using adjusted hazard ratios (aHR) with Cox regression modeling.

Results

In total, 2184 women contributed 2368 incident HPV genotype-specific first detections and 308 genotype-specific redetections over a median follow-up of 6.5 years. The cumulative incidence of redetection with the same genotype was 6.6% at 1 year and 14.8% at 5 years after the loss of positivity of the first detection. Neither age (aHR 0.90; 95% confidence interval [CI], .54–1.47 for ≥45 years vs < 25 years) nor new sexual partner acquisition (aHR 0.98; 95% CI, .70–1.35) were statistically associated with genotype-specific redetection. High-grade squamous intraepithelial lesion prevalence was similar during first HPV detections (2.9%) and redetection (3.2%).

Conclusions

Our findings suggest many HPV redetections were likely reactivations of latent recurring infections.

Incidence of redetection with the same human papillomavirus genotype was found to be substantially higher than incidence of first detection. Redetections were not associated with new sexual partners. This suggests many human papillomavirus detections could be recurring latent infections.

Human papillomavirus (HPV) DNA testing is the primary screening test for cervical cancer in many countries, and is recommended by the World Health Organization for cervical cancer screening worldwide [1]. The emerging model of HPV natural history holds that while initial genital HPV infection is generally acquired via sexual exposure, the course of infection after initial acquisition can be nonlinear and follow different pathways [2]. The loss of HPV DNA detectability in cervical samples may not always reflect true immune clearance. In some cases, it may occur due to immune control of the infection below the limit of detection in a state of viral latency [3]. Consequently, the redetection of HPV DNA after a period of negativity could be due to several reasons, including true new reinfection from a sexual exposure, intermittent detection of a latent infection, or simply transient deposition from a cross-infection at another epithelial site or from a recent sex act.

Many questions remain regarding what is the long-term risk of HPV redetection, whether the risk of HPV redetection changes with age, and whether a redetection of HPV confers the same risk of cervical cancer as the initial infection detection [2]. Most natural history studies have had insufficient sample sizes and follow-up to study redetections of HPV. These questions are of increasing clinical importance, as HPV-based screening guidelines use repeat HPV tests during triage [4], and the screening management of a woman with HPV-positive results may depend on her previous history of HPV positivity [5–7].

The Ludwig-McGill cohort study was a longitudinal study of the natural history of HPV infection and cervical neoplasia, which has contributed much to our knowledge of HPV natural history over the past 30 years [8–12]. We have previously published an analysis of the HPV redetections that were most likely to be true new reinfections from new sexual exposures; that analysis was restricted to redetections which occurred after 3 intervening negative test results [9]. However, these redetections only represented a small fraction of all redetections in the cohort. Our objective with this new analysis was to examine all redetections of the same HPV genotype in women to assess how frequently HPV genotype-specific redetections occur, what are the predictors of HPV redetection in women who have previously tested HPV positive, and whether redetections are associated with the same prevalence of cytological lesions as first HPV detections.

METHODS

Study Design and Participants

The study protocol and methods have been previously described in detail elsewhere [8]. Briefly, the study recruited women between 18 and 60 years old attending a maternal and child health program catering to low-income families in the city of São Paulo, Brazil between 1993 and 1997. Women were followed up every 4 months during the first year, and subsequently twice a year for up to 10 years after enrolment. During the first 4 visits and at every second visit thereafter, women completed an interviewer-administered questionnaire on demographic, socioeconomic, and behavioral risk factors for HPV infection and cervical cancer. Study personnel collected cervical samples for cytology and HPV DNA testing at each visit. All women provided signed consent forms, and ethical approval was obtained from the institutional review boards of McGill University (Montreal, Canada), University of Toronto (Toronto, Canada), Ludwig Institute for Cancer Research (São Paulo, Brazil), and Maternidade Escola Dr. Mario de Moraes Altenfelder Silva Municipal Hospital Clinic (São Paulo, Brazil).

Cervical Smears

Ectocervical and endocervical cells were collected using an Accelon biosampler (Medscand, Inc) and placed in a tube containing Tris-EDTA buffer. Cervical smears were prepared on a glass slide and fixed in 95% ethanol, stained, and read at the São Paulo Branch of the Ludwig Institute's cytopathology laboratory for initial diagnosis based on the Papanicolaou system. The smears were then shipped frozen on dry ice to Montreal and re-read at the Jewish General Hospital based on the Bethesda system [13]. Women who had moderate or worse dysplasia based on initial readings with the Papanicolaou system or who had high-grade squamous intraepithelial lesions (HSIL) on subsequent readings with the Bethesda system were referred for colposcopy and management according to the local prevailing protocol. For this analysis, we used the Montreal cytology results.

HPV Genotyping

HPV DNA was extracted, purified by spin column chromatography, and amplified by polymerase chain reaction (PCR) using the MY09/11 and PGMY protocols [14–16]. Typing of the amplified products was performed by hybridization with individual oligonucleotide probes for 27 HPV genotypes, and restriction fragment length polymorphism for the samples that hybridized with the generic but none of the genotype-specific probes [17]. These combined methods allowed the identification of over 40 genital HPV genotypes, including unknown genotypes. We classified as high-risk oncogenic HPV genotypes 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68, based on both carcinogenicity [18] and for comparability as these genotypes are the targets of many commercial HPV tests. To check the integrity of samples, the assays also included an additional set of primers (GH20 and PC04) to amplify a 268-bp region of the β-globin gene [14]. Samples that were negative for both HPV and β-globin were considered invalid and excluded from analyses.

Statistical Analysis

All analyses were genotype specific, with the unit of observation being the HPV genotype. Hence, women contributed multiple observations to analyses corresponding to individual HPV genotypes. Results were pooled over multiple genotypes.

Descriptive Analyses

To describe detection patterns, we examined women with at least 3 study visits. We considered HPV genotypes to be prevalent if they were detected at the study baseline, and incident if detected at subsequent visits in women who were negative at baseline for that HPV genotype. We considered an HPV detection pattern to be persistent when the same genotype was detected at 2 or more successive visits with valid HPV test results, transient when a genotype was detected only at a single visit, and intermittent when a genotype was detected at 2 or more visits separated by at least 1 intervening negative test for that genotype.

Cumulative Incidence of HPV Detection

To calculate cumulative HPV detection incidence, we analyzed women with at least 2 study visits using stratified Cox proportional hazards models. The models were stratified, with different baseline hazard functions for the first detection, first redetection (second detection), and all subsequent redetections (third or more detections) with the same HPV genotype. The time to first genotype-specific HPV detection was modeled from the baseline study visit in women who were genotype-specific negative at baseline. The time to genotype-specific HPV redetection was modeled from the date that a woman became negative for a genotype she was previously positive for (the clearance date). We used the robust sandwich estimate of Lin and Wei to account for having multiple observations from the same woman [19]. The cumulative incidence of first genotype-specific detection and redetection overall and by age were estimated from the survivor functions of stratified Cox models.

Risk Factors for HPV Detection

For regression models, we considered as a priori predictors demographic and socioeconomic variables (age, menopausal status, income quartile, race, marital status) and behavioral variables (smoking, number of lifetime sex partners, new sex partners, current sex partners, condom use) known to be associated with HPV acquisition. Age and behavioral variables were analyzed as time-varying exposures. We modeled each variable individually in univariate models and all together in a multivariable model, except for marital status, which was not included in the multivariable model due to collinearity with sexual behavior variables. Models had interaction terms to allow predictors to have different effects for first incident detections and for redetections. We used the joint tests for the interaction terms to assess whether hazard ratios (HR) differed for first incident detections compared with redetections.

Values for some predictors were missing at some visits either due to nonresponse or design (questionnaires were given only every second visit after the first year, and questions on income, race, and menopausal status were not repeated in later questionnaires). For sexual behavior variables, we imputed the woman's last nonmissing response backwards to previous visits with missing data. For other variables, we imputed the woman's last nonmissing response forward to subsequent visits with missing data.

Cervical Lesion Prevalence

Finally, we assessed in all women whether the prevalence of low-grade squamous intraepithelial lesions or worse (LSIL), HSIL, and atypical squamous cells of undetermined significance or worse (ASCUS+, which includes LSIL and HSIL), differed between visits with first HPV genotype detections and redetections for high-risk HPV genotypes. We used the 2.5th–97.5th percentiles of 2000 bootstrap resamples of participants to estimate 95% confidence intervals (CIs) for prevalences and prevalence differences. Because the unit of observation was the HPV genotype, numerators and denominators do not represent the number of cytological lesions over the number of women. Rather, they represent the number of genotype-specific observations across visits with prevalent cervical abnormalities over all genotype-specific observations across visits. This method accounts for the cases where a woman could have first detections for some HPV genotypes while simultaneously having redetections with other types on a given visit; in these cases, the lesion would be attributed to all HPV genotypes occurring at that visit.

RESULTS

Descriptive Analyses

There were 2462 eligible women recruited in the Ludwig-McGill cohort study, of which 2184 (88.7%) had at least 2 study visits and 1986 (80.7%) had at least 3 study visits with valid HPV DNA data. Baseline characteristics of women with at least 2 study visits are presented in Supplementary Table 1. The median age at baseline was 32 years (first to third quartile, 26–39 years) and the majority of women were either married (47.9%) or living as married (33.8%). The median follow-up time for women with at least 2 study visits was 6.5 years (first to third quartile, 4.3–7.8 years). In women with at least 3 study visits, 14.9% (60/402) of prevalent genotype-specific detections had an intermittent detection pattern, and 8.4% (196/2346) of incident genotype-specific detections had an intermittent detection pattern, where the same genotype was redetected at a later visit after a negative result (Table 1). Only 29.9% (92/308) of redetections in our analysis occurred after 3 intervening negative results (Supplementary Table 2).

Table 1.

HPV Genotype-Specific Detection Patterns in Women With 3 or More Study Visits, Pooled Over All HPV Genotypes

	Prevalent Detections (n = 402)		Incident Detections (n = 2346)		Total (n = 2748)
Detection Pattern	Patterna	n (%)	Patterna	n (%)	n (%)
Persistent	111/110	137 (34.1)	011	494 (21.1)	631 (23.0)
Transient	100	205 (51.0)	010/001	1656 (70.6)	1861 (67.7)
Intermittent	101	60 (14.9)	0101	196 (8.4)	256 (9.3)

Abbreviation: HPV, human papillomavirus.

Short list of illustrative patterns (1 = HPV positive, 0 = HPV negative) over 3–4 visits that fit into each detection pattern definition. A full list of patterns is provided in Supplementary Table 2.

Cumulative Incidence of HPV Detection

There were 308 redetections (256 second detections and 52 third or more detections) of the same genotype contributing to analyses of cumulative incidence of HPV genotype-specific redetections (Figure 1). The cumulative incidence of first detection with an HPV genotype in women negative for that HPV genotype at baseline was 0.7% (95% CI, .6%–.7%) 1 year after baseline, and 2.3% (95% CI, 2.2%–2.4%) 5 years after baseline, pooled across all HPV genotypes (Figure 1 and Table 2). The cumulative incidence of the first redetection (second detection) of the same genotype was 6.6% (95% CI, 5.5%–7.7%) 1 year after the date of loss of positivity of the first detection, and 14.8% (95% CI, 13.0%–16.6%) 5 years after the date of loss of positivity of the first detection, pooled across all HPV genotypes. The cumulative incidence of additional redetections (third or more detections) of the same genotype was 15.8% (95% CI, 10.1%–21.1%) 1 year after the date of loss of positivity of the previous detection, and 42.0% (95% CI, 30.3%–51.7%) 5 years after the loss of positivity date of the previous detection, pooled across all HPV genotypes.

Figure 1.

Cumulative incidence of genotype-specific incident detection of human papillomavirus (HPV; first incident detection) and of redetection after at least 1 negative visit of the same HPV genotype (second and third or more incident detection), pooled across all HPV genotypes. Time to detection is modeled from the baseline study visit for first HPV detection, and from the first negative visit following the prior detection of that genotype for redetections. Notches represent censored observations, and shaded regions represent 95% confidence intervals.

Table 2.

Cumulative Incidence of Genotype-Specific First HPV Detection From Baseline, and of Same-Genotype Redetection After at Least 1 Negative Visit in Women With 2 or More Study Visits

	Events		1-y Cumulative Incidence				5-y Cumulative Incidence
	First Detectiona	First Redetectionb	First Detectiona		First Redetectionb		First Detectiona		First Redetectionb
HPV Genotype	n	n	%	(95% CI)	%	(95% CI)	%	(95% CI)	%	(95% CI)
All genotypes, pooledc	2368	256	0.7	(.6–.7)	6.6	(5.5–7.7)	2.3	(2.2–2.4)	14.8	(13.0–16.6)
High-risk genotypes, pooledc,d	1145	116	1.0	(.8–1.1)	5.8	(4.3–7.2)	3.3	(3.1–3.6)	14.3	(11.7–16.9)
Low-risk genotypes, pooledc,e	1223	140	0.5	(.4–.6)	7.4	(5.8–9.0)	1.8	(1.7–1.9)	15.3	(12.7–17.8)
HPV6/11	81	5	0.8	(.4–1.2)	2.5	(.0–5.9)	3.2	(2.3–4.0)	7.1	(.0–13.9)
HPV16	236	35	3.1	(2.3–3.8)	7.9	(4.2–11.4)	9.7	(8.3–11.1)	19.3	(13.0–25.2)
HPV18	69	7	0.8	(.4–1.2)	4.2	(.0–8.7)	2.6	(1.9–3.3)	11.5	(.3–19.3)
HPV26	17	2	0.1	(.0–.3)	0.0	(.0–.0)	0.6	(.2–.9)	16.2	(.0–34.6)
HPV31	77	8	1.1	(.6–1.6)	6.9	(.9–12.6)	2.9	(2.1–3.7)	11.3	(2.7–19.2)
HPV32	9	0	0.1	(.0–.2)	0.0	NA	0.4	(.1–.7)	0.0	NA
HPV33	44	1	0.6	(.3–1.0)	2.5	(.0–7.3)	1.7	(1.1–2.2)	2.5	(.0–7.3)
HPV34	2	0	0.0	(.0–.0)	0.0	NA	0.0	(.0–.0)	0.0	NA
HPV35	82	7	0.6	(.3–.1)	7.5	(.9–13.7)	3.3	(2.4–4.1)	15.4	(.7–27.9)
HPV39	46	3	0.3	(.1–.5)	3.2	(.0–9.1)	1.5	(.9–2.0)	13.7	(.0–27.4)
HPV40	38	7	0.4	(.1–.7)	10.8	(.2–20.3)	1.3	(.8–1.8)	23.0	(5.5–37.3)
HPV42	32	3	0.1	(.0–.2)	10.3	(.0–22.8)	1.1	(.6–1.5)	24.0	(.0–47.0)
HPV44	90	11	0.6	(.3–.9)	8.5	(1.7–14.8)	3.2	(2.3–4.0)	22.4	(6.1–35.9)
HPV45	78	5	0.4	(.1–.7)	5.4	(.0–11.2)	2.5	(1.8–3.2)	11.3	(.1–20.5)
HPV51	135	12	1.5	(.9–2.0)	4.8	(1.0–8.5)	5.7	(4.6–6.8)	12.1	(5.0–18.6)
HPV52	100	12	1.0	(.6–1.5)	4.4	(.1–8.5)	3.1	(2.3–3.9)	17.5	(7.5–26.4)
HPV53	146	23	2.3	(1.6–2.9)	7.1	(2.8–11.2)	5.6	(4.5–6.7)	18.2	(10.4–25.2)
HPV54	70	8	0.8	(.4–1.2)	6.9	(.1–13.1)	3.0	(2.2–3.8)	15.3	(4.8–24.7)
HPV56	60	4	0.6	(.2–.9)	5.6	(.0–11.5)	2.4	(1.7–3.1)	7.8	(.1–15.0)
HPV57	6	0	0.1	(.0–.1)	0.0	NA	0.2	(.0–.4)	0.0	NA
HPV58	98	13	1.0	(.6–1.4)	4.7	(.1–9.1)	3.7	(2.8–4.6)	21.5	(9.8–31.6)
HPV59	60	5	0.6	(.4–.9)	5.7	(.0–11.7)	2.2	(1.5–2.9)	10.9	(1.2–19.7)
HPV61	49	6	0.7	(.3–1.0)	8.2	(.2–15.6)	1.8	(1.2–2.4)	12.9	(2.6–22.1)
HPV62	60	12	0.5	(.2–.7)	15.7	(4.3–25.8)	2.3	(1.6–3.0)	33.2	(14.4–47.9)
HPV66	54	3	0.3	(.1–.6)	4.2	(.0–9.8)	1.7	(1.1–2.3)	7.0	(.0–14.4)
HPV67	9	0	0.3	(.1–.6)	0.0	NA	0.3	(.1–.6)	0.0	NA
HPV68	60	4	0.8	(.4–1.2)	5.4	(.0–11.2)	2.4	(1.7–3.1)	8.1	(.0–15.5)
HPV69	6	0	0.1	(.0–.2)	0.0	NA	0.2	(.0–.3)	0.0	NA
HPV70	36	4	0.4	(.1–.7)	7.5	(.0–15.4)	1.4	(.9–2.0)	10.4	(.2–19.5)
HPV71	38	4	0.4	(.2–.7)	7.5	(.0–17.1)	0.9	(.5–1.3)	20.1	(.0–36.9)
HPV72	11	4	0.2	(.0–.5)	24.8	(.0–45.7)	0.4	(.1–.6)	36.4	(.0–59.9)
HPV73	54	1	0.4	(.2–.7)	0.0	(.0–.0)	2.1	(1.4–2.8)	2.4	(.0–6.8)
HPV81	29	6	0.4	(.1–.7)	17.4	(2.2–30.2)	1.0	(.6–1.5)	23.1	(4.2–38.2)
HPV82	24	0	0.3	(.1–.6)	0.0	NA	0.9	(.5–1.4)	0.0	NA
HPV83	53	1	0.3	(.1–.6)	2.7	(.0–7.9)	1.8	(1.2–2.4)	2.7	(.0–7.9)
HPV84	112	16	0.8	(.4–1.2)	9.7	(3.1–15.9)	3.6	(2.7–4.5)	19.8	(9.6–28.8)
HPV89	24	2	0.1	(.0–.2)	9.3	(.0–20.9)	0.9	(.5–1.4)	9.3	(.0–20.9)

Abbreviations: CI, confidence interval; HPV, human papillomavirus; NA, not available (could not be calculated).

First incident detection of a given HPV genotype from the baseline study visit.

Second detection of the same HPV genotype from the first negative visit following first positivity with that genotype.

Results from all individual HPV genotypes pooled together in analysis.

HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68.

All other types excluding HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68.

The joint test for the effect of HPV genotype suggested there were significant overall differences in incidence of first detection with individual HPV genotypes (P < .0001), with HPV16 being the genotype with the highest rate of first incident detection (Table 2). There were also significant overall differences in the incidence of redetection with individual HPV genotypes (P < .0001), but while HPV16 had a higher redetection incidence than the pooled average (19.3% vs 14.8% at 5 years), there were many other genotypes with higher redetection incidences (Table 2). The cumulative incidence of redetection was similar when comparing pooled high-risk and low-risk HPV genotypes as categories (Table 2 and Supplementary Figure 1).

Risk Factors for HPV Detection

While the rate of first incident detections decreased with age, age was not a significant predictor of the rate of redetection (Table 3 and Figure 2). The joint test for interaction was highly significant (P < .0001), suggesting a more protective association between age and first incident detections than for redetections. Being a current smoker was associated with a 1.32 (95% CI, .99–1.76) times higher genotype-specific redetection rate after adjustment for other variables. Increasing number of lifetime sex partners was associated with a significantly increased incidence of both first incident HPV detection and same-genotype redetections. Having a sex partner in the last interval was not a significant predictor of either first incident detections or redetections; in most cases these were the woman's ongoing steady sex partner. Conversely, having a new sex partner in the last interval was associated with a higher rate of first incident detections (HR, 2.04, 95% CI, 1.74–2.40), but not with same-genotype redetections (HR, 0.98; 95% CI, .70–1.35); the joint test for interaction was significant (P < .0001), suggesting the effect of having a new partner is different for first incident detections than for redetections.

Table 3.

Hazard Ratios of HPV Genotype-Specific First Incident Detection and Redetection (Second or More Detections) by Women's Characteristics

	First Incident Detection (n = 2368)						Redetections (n = 308)
Variablea	Events	PY	HR, crude	(95% CI)	HR,b adj	(95% CI)	Events	PY	HR, crude	(95% CI)	HR,b adj	(95% CI)	InteractioncP Value
Age, y													<.0001
<25	379	36 534	1.00	(ref)	1.00	(ref)	35	663	1.00	(ref)	1.00	(ref)
25–34	970	159 095	0.53	(.44–.62)	0.57	(.49–.67)	112	3203	0.81	(.56–1.18)	0.82	(.56–1.19)
35–44	658	163 179	0.34	(.28–.41)	0.38	(.31–.46)	115	2270	1.20	(.82–1.77)	1.18	(.79–1.77)
≥45	361	92 166	0.32	(.25–.40)	0.36	(.28–.45)	46	1265	0.87	(.54–1.39)	0.90	(.54–1.47)
Menopausal status													.51
Premenopausal	2272	430 449	1.00	(ref)	1.00	(ref)	297	7136	1.00	(ref)	1.00	(ref)
Postmenopausal	96	20 608	0.89	(.66–1.19)	1.30	(.96–1.76)	11	265	0.98	(.44–2.19)	0.98	(.44–2.22)
Marital status
Married	750	216 068	1.00	(ref)	…	…	69	2204	1.00	(ref)	…	…
Living as married	711	130 276	1.58	(1.36–1.81)	…	…	104	2365	1.41	(1.01–1.97)	…	…
Widowed	93	14 398	1.86	(1.34–2.57)	…	…	9	242	1.11	(.49–2.49)	…	…
Separated	284	35 762	2.27	(1.98–2.61)	…	…	38	818	1.44	(.97–2.15)	…	…
Single	459	41 933	3.08	(2.68–3.68)	…	…	67	1351	1.59	(1.11–2.29)	…	…
Income quartile													.48
1 (lowest)	607	105 665	1.22	(1.02–1.46)	1.12	(.95–1.33)	64	1830	0.83	(.56–1.23)	0.84	(.57–1.24)
2	587	108 853	1.17	(.97–1.41)	1.18	(.99–1.41)	69	1703	0.96	(.67–1.38)	1.02	(.72–1.45)
3	557	111 106	1.10	(.91–1.33)	1.14	(.95–1.36)	90	1769	1.19	(.84–1.69)	1.20	(.84–1.70)
4 (highest)	573	117 208	1.00	(ref)	1.00	(ref)	79	1992	1.00	(ref)	1.00	(ref)
Race													.42
White	1468	288 767	1.00	(ref)	1.00	(ref)	180	4435	1.00	(ref)	1.00	(ref)
Non-white	900	162 206	1.09	(.96–1.24)	0.96	(.85–1.09)	128	2966	1.08	(.83–1.40)	1.08	(.83–1.41)
Smoking status													.28
Never	1144	221 811	1.00	(ref)	1.00	(ref)	123	3519	1.00	(ref)	1.00	(ref)
Former	865	78 204	0.89	(.74–1.06)	0.85	(.72–1.00)	134	1272	1.17	(.82–1.68)	1.11	(.76–1.60)
Current	359	150 958	1.11	(.97–1.28)	1.06	(.93–1.22)	51	2611	1.44	(1.09–1.91)	1.32	(.99–1.76)
Condom use last interval													<.0001
Yes	583	81 935	1.00	(ref)	1.00	(ref)	80	1907	1.00	(ref)	1.00	(ref)
No	1714	356 502	0.67	(.59–.77)	0.96	(.85–1.10)	207	5072	0.99	(.75–1.30)	1.00	(.76–1.33)
Lifetime number of sex partners													<.0001
0–1	596	181 521	1.00	(ref)	1.00	(ref)	58	1782	1.00	(ref)	1.00	(ref)
2–3	902	149 693	1.83	(1.58–2.12)	1.65	(1.42–1.91)	95	2458	1.17	(.82–1.67)	1.16	(.82–1.65)
4+	798	106 794	2.23	(1.91–2.61)	1.78	(1.51–2.09)	134	2733	1.55	(1.11–2.16)	1.50	(1.06–2.10)
Any sex partner in last interval													.88
No	203	39 373	1.00	(ref)	1.00	(ref)	27	657	1.00	(ref)	1.00	(ref)
Yes	2093	399 010	1.06	(.89–1.27)	0.89	(.73–1.08)	260	6320	0.94	(.61–1.45)	0.92	(.59–1.45)
New sex partner in last interval													<.0001
No	1877	405 523	1.00	(ref)	1.00	(ref)	242	6001	1.00	(ref)	1.00	(ref)
Yes	416	32 593	2.86	(2.46–3.33)	2.04	(1.74–2.40)	45	971	1.08	(.78–1.50)	0.98	(.70–1.35)

Abbreviations: CI, confidence interval; HPV, human papillomavirus; HR, hazard ratio; PY, person-years at risk; ref, reference.

All variables considered as time-varying predictors in analyses, except for income and race which were only measured at study baseline, and menopausal status which was only measured twice.

Adjusted for all variables in table, except for marital status which was excluded from the multivariable model due to collinearity with sexual behavior variables.

Joint test for interaction effect between the predictor variable and order of detection (first or redetection) in the multivariable adjusted model using the robust sandwich variance estimate. This tests whether the HRs are equivalent for first incident detections and redetections.

Figure 2.

Cumulative incidence of human papillomavirus (HPV) genotype-specific first detection (A) and redetection (B) by age at time of detection, pooled across all HPV genotypes. Time to detection is modeled from the baseline study visit for first HPV detection, and from the first negative visit following the prior detection of that genotype for redetection. Age is modeled as a time-varying exposure. Notches represent censored observations, and shaded regions represent 95% confidence intervals.

Cervical Lesion Prevalence

There were 646 ASCUS+ cytology results during follow-up, including 258 LSILs and 77 HSILs. Table 4 presents the prevalence of ASCUS+, LSIL, and HSIL across high-risk HPV genotype observations, that is, the probability that a woman who is negative or positive for a given high-risk HPV genotype on a given visit had a concurrent ASCUS+, LSIL, or HSIL cytology result. The prevalence of HSIL was similar across visits with first detections (2.9%; 95% CI, 1.8%–3.7%) and redetections (3.1%; 95% CI, 1.3%–4.9%), with a prevalence difference of −0.3% (95% CI, −2.2% to 1.6%). Similar results for first and redetections were also observed for LSIL and ASCUS+.

Table 4.

Cross-sectional Prevalence of Cytological Results by Genotype-Specific HPV Positivity, Pooled Over All High-risk HPV Genotypes and Visits Over all Women

High-risk HPV Status at Visit, Genotype Specifica	ASCUS+ Prevalence			LSIL Prevalence			HSIL Prevalence
	n/Visitsb	%	(95% CI)	n/Visitsb	(%)	(95% CI)	n/Visitsb	(%)	(95% CI)
Negative	7842/308 004	2.5	(2.3 to 2.8)	3095/308 004	1.0	(.9 to .1)	928/308 004	0.3	(.2 to .4)
Positive, first detection	399/2264	17.6	(15.4 to 19.6)	202/2264	8.9	(7.5 to 10.2)	65/2264	2.9	(1.9 to 3.7)
Positive, redetections	40/250	16.0	(10.4 to 21.1)	18/250	7.2	(3.1 to 10.6)	8/250	3.2	(1.3 to 4.9)
Difference, first detection – redetections		1.6	(−3.1 to 6.7)		1.7	(−1.6 to 5.5)		−0.3	(−2.2 to 1.6)

Abbreviations: ASCUS+, atypical squamous cells of undetermined significance or higher; CI, confidence interval; HPV, human papillomavirus; HSIL, high grade squamous intraepithelial lesion; LSIL, low grade squamous intraepithelial lesion.

High-risk HPV: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68.

Numerator is number of genotype-specific visits with given cytological results over all women, denominator is number of genotype-specific visits over all women.

DISCUSSION

In this analysis of the Ludwig-McGill cohort over a median follow-up of 6.5 years, many women had intermittent HPV detection patterns. The risk of redetection of a given HPV genotype was 14.8% by 5 years after the date a woman became negative for that genotype, pooled across HPV genotypes. While first HPV detections were associated with the acquisition of new sex partners, same-genotype HPV redetections were not. We observed a similar prevalence of HSIL at visits with first detections and redetections with the same HPV genotype.

While having new sexual partners was associated with an increased risk of first HPV detection, it was not associated with redetections of the same HPV genotype in our analysis. The effect of new sexual partner acquisition varies across studies, with some finding that having new sexual partners increases the risk of redetection [9, 20] while others do not [21]. Our results should be interpreted in light of the previous analysis by Trottier et al of the Ludwig-McGill cohort study [9], which had found that some redetections were reinfections associated with the acquisition of new sexual partners. There were fundamental differences between analyses that should be noted. The Trottier et al study had research questions focused on HPV reinfections, and restricted analyses to redetections occurring after 3 consecutive negative visits to exclude potential intermittent detection. In contrast, we included all redetections with at least 1 intervening negative result. This restriction accounts for part of the difference in results. The other major difference was that Trottier et al performed a woman-level analysis rather than an HPV genotype-level analysis. Woman-level analyses are closer to the clinical results that a woman would receive using a screening test targeting multiple HPV genotypes at once, whereas HPV genotype-level analyses focus on understanding the biological course of individual infections. Trottier et al's woman-level analysis looked at the time to first redetection with any HPV genotype a woman was previously positive for, so was restricted to 19 first redetections and excluded 73 other same-genotype redetections after 3 negative intervening results. These first redetections across multiple genotypes were more strongly associated with new sexual partner acquisition. This effect of new sexual partners on HPV redetection was strongly diluted in our current analysis due to the inclusion of many more genotype-specific redetections than in Trottier et al. The results from both analyses suggest that while there are redetections that are true new infections caused by sexual transmission, the majority of genotype-specific HPV redetections in this cohort are not associated with new partner acquisition and may instead represent reactivations/redetections of previously existing infections.

Our results suggest that many HPV redetections are likely to be reactivations of latent infections as well as new reinfections from sexual exposure. Previous studies have found that many HPV infections have intermittent detection patterns [21–25]. These tend to have lower viral loads than persistent or transient detections [9, 24], suggesting that some redetections may be infections with viral shedding fluctuating below the limit of detection. We used definitions of HPV positivity patterns similar to those used in previous studies to assess comparability with previously published data. We found that 8.4% of genotype-specific HPV-positive observations in the Ludwig-McGill cohort study had intermittent positivity patterns. This was similar to a cohort study from Costa Rica, which found that 8% of women ever testing HPV positive over 7 years had intermittent detection patterns [23], but lower than studies from the United States, which found that 13%–26% of HPV detections had intermittent detection patterns [24, 26]. However, these above observed patterns are based on crude counts; the proportion of observations with intermittent detection patterns may vary across studies with different follow-up due to right-censoring of observations. Survival analysis methods provide a better estimate of the cumulative probability of redetection over time by accounting for censoring. Using a Cox model, we estimated that 5 years after loss of positivity of a first HPV detection there was a 14.8% cumulative probability of redetection of the same HPV genotype. These results are similar to another study which found that 18% of HPV16 infections became redetectable by 8.5 years after the date of loss of positivity [20]. Redetection of the same HPV genotype is therefore fairly common and can occur several years after testing negative.

Previously, the Guanacaste cohort study had found that women with HPV detected only once and women with HPV redetections both had a similar 7%–9% risk of cervical intraepithelial neoplasia grade 2+ (CIN2+) [23]. Results from Kaiser Permanente Northern California found that the risk of CIN3+ in women with intermittent detection patterns depended mostly on whether the woman had a current HPV-positive result [5]. We found that the prevalence of cytological lesions did not substantially differ between first and subsequent redetection episodes with the same HPV genotype. However, our study used cytology results rather than histology and different statistical methods, and so is not directly comparable to these previous studies. Also, we assessed the prevalence of lesions rather than the incidence, and women could have simultaneous first detections and redetections with different genotypes on the same visit. The cytology prevalence results should therefore be interpreted as associations rather than causal effects, as we could not attribute individual lesions to specific HPV types or detection events.

Increasing age was associated with a lower risk of first HPV detection but not with HPV redetections. This was expected for first detections, as age is correlated with sexual activity and may also be a proxy of the prevalence of HPV infection in women's sexual partners. The lack of association between age and redetections conversely suggests that redetection is less likely to be attributable to sexual transmission. These results are not consistent with the hypothesis that the risk of reactivation of latent infections increases with age. Increasing age is associated with a decline in immune responses to pathogens [27]. However, we had little follow-up time in participants over 60 years old, so we would not be able to detect effects of immunosenescence on HPV reactivation at much older ages.

A potential limitation of our study was that because women were recruited from a maternal and child health program catering to low-income families, participants are not representative of the general population. The fraction of all HPV detections attributable to sexual transmission may be lower in this population, as most women were married or living as married with a steady partner, and few reported new sexual partners over the course of the study. The fraction of detections attributable to sexual transmission is likely to be higher in populations with more sexual partner turnover [28, 29]. Because we had missing data for some variables, we imputed questionnaire responses backwards or forwards; this likely led to some misclassification of exposures. Also, we do not have data from before women were recruited into the study, and most women had had previous sex partners. It is consequently likely that many of the observed first detections were potentially themselves redetections of previous infections, misclassified as first detections. The hazard ratios in our study therefore likely underestimate the role of new sexual partner acquisition on the risk of incident first acquisition of HPV infection.

In summary, we found that redetection of the same HPV genotype is fairly common in women and may in some cases occur many years after a woman becomes HPV DNA negative. These results may help in destigmatizing a positive HPV test result, as they suggest that many HPV detections may be a reactivated past infection, rather than a new infection from recent sexual behaviors or partner infidelity. The results also suggest that episodes of HPV redetection are associated with a similar prevalence of underlying cervical lesions as first detections.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. Ludwig-McGill cohort study team members: affiliated with the Ludwig Institute for Cancer Research in Sao Paulo, Brazil are Maria Luiza Baggio, Lenice Galan, João Simão Sobrinho, José Carlos Mann Prado, Lara Termini, Maria Cecília Costa, Romulo Miyamura, Andrea Trevisan, Patricia Thomann, João Candeias, Laura Sichero, Paula Rahal, Antonio Ruiz, Jane Kaiano, Monica Santos, Patricia Savio, Paulo Maciag, Tatiana Rabachini, Silvaneide Ferreira, and Luisa Villa (coprincipal investigator): affiliated with McGill University in Montreal, Canada are Mariam El-Zein, Marie-Claude Rousseau, Salaheddin Mahmud, Nicolas Schlecht, Helen Trottier, Harriet Richardson, Alex Ferenczy, Thomas Rohan, Myriam Chevarie-Davis, Karolina Louvanto, Joseph Tota, Eileen Shaw, Agnihotram Ramanakumar, Eliane Duarte, Sophie Kulaga, Juliette Robitaille, and Eduardo Franco (principal investigator).

Disclaimer . The funders of the study had no involvement in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Financial support. This work was supported by the Ludwig Institute for Cancer Research (intramural grant to L. L. V. and E. L. F.); the US National Cancer Institute (grant number CA70269 to E. L. F.); the Canadian Institutes of Health Research (CIHR; grant numbers MA-13647, MOP-49396, and CRN-83320 to E. L. F.); the Fonds de la Recherche du Québec en Santé (salary award to H. T.); and the Canadian Institutes of Health Research (new investigator salary award to H. T.).

Availability of data, software, and research materials. Participants of the Ludwig-McGill cohort study did not consent to have their data made publicly available, and confidentiality precludes the publishing of their data. To access the data for research purposes, please contact Eduardo Franco (eduardo.franco@mcgill.ca) or Luisa Villa (l.villa@hc.fm.usp.br). The program code for the current results is available at the McGill University Dataverse repository: https://doi.org/10.5683/SP3/EIF6A7.