Anal Squamous Intraepithelial Lesions (SILs) in Human Immunodeficiency Virus–Positive Men Who Have Sex With Men: Incidence and Risk Factors of SIL and of Progression and Clearance of Low-Grade SILs

Abstract

Background

Human immunodeficiency virus (HIV)–positive men who have sex with men (MSM) are at risk of anal squamous cell carcinoma. Data are limited on the natural history of the precursor to this carcinoma, anal squamous intraepithelial lesions (SILs).

Methods

HIV-positive MSM were screened for histopathological SILs by means of high-resolution anoscopy (HRA). For participants without SILs at baseline, we estimated the cumulative incidence and risk factors for SILs. For those with low-grade SILs (LSILs) at baseline, the risk of progression to high-grade SILs (HSILs) and the clearance rate were estimated at the lesion level.

Results

Of 807 men without SILs at baseline, 107 underwent follow-up HRA between 1 to 4.5 years later. At the second visit 18 men (16.8%) showed LSIL, and 25 (23.4%) HSIL. Age was associated with incident LSILs (adjusted odds ratio [aOR], 2.10 per 10-year increase in age; P = .01). Of 393 men with LSILs at baseline, 114 underwent follow-up HRA 0.5 to 2.5 years later. Of the 177 LSILs found at baseline, 87 (49.2%) had cleared at the second visit, and 29 (16.4%) had progressed to HSILs.

Conclusion

Incident LSILs and HSILs were common during follow-up among HIV-positive MSM without dysplasia at baseline. Among men with LSILs at baseline, nearly half of these lesions cleared, and a small portion progressed.

(See the Editorial Commentary by Barroso on pages 7–8.)

Human immunodeficiency virus (HIV)–positive men who have sex with men (MSM) are at higher risk of anal squamous cell carcinoma than the general population [1, 2]. The precursor of anal squamous cell carcinoma, high-grade squamous intraepithelial lesions (HSILs), can be detected by means of high-resolution anoscopy (HRA)–guided biopsies. SILs are histopathologically graded as anal intraepithelial neoplasia (AIN) 1, 2, or 3, but are increasingly categorized as low-grade SILs (LSILs) (AIN1) or HSILs (AIN2 and AIN3), according to the Lower Anogenital Squamous Terminology (LAST) Project criteria [3]. The prevalence of HSILs is higher in HIV-positive than in HIV-negative MSM [4–8], but no additional risk factors have been consistently identified [9–20]. Although data on the natural history of LSILs are limited, the risk of progression from LSILs to HSILs has been estimated at between 7.2% at 12 months and 38.7% at 36 months [10, 12, 21, 22]. The rate of spontaneous regression of LSILs is still unclear.

In the current study, we investigated the incidence of SILs among men without evidence of anal dysplasia at baseline, and the risk of progression to HSILs and clearance of LSILs among those with LSILs at baseline, in a cohort of HIV-positive MSM screened by means of HRA in Amsterdam, the Netherlands. We studied the association of several demographic, behavioral, and HIV-related variables with the incidence, progression and clearance of intra-anal lesions.

METHODS

Participants and Setting

Data collection has been described in detail elsewhere [9]. In short, HIV-positive MSM were offered AIN screening at 1 of 3 HIV outpatient clinics in Amsterdam that started an HRA program in 2008 (clinic A), 2011 (clinic B), and 2009 (clinic C). HRA was performed by individually trained anoscopists. The screening consisted of a digital rectal examination followed by intra-anal and perianal inspection using a colposcope (Zeiss OPMI pico surgical microscope) after repeated application of acetic acid (3%–5% solution) and staining with Lugol iodine when indicated.

Biopsy specimens were obtained from lesions suspected of being SILs and graded by pathologists who specialized in SILs. The location of the biopsy specimen in the anal canal was determined using the hours on a clock in lithotomy position, whereby 12 o’clock is the anterior position, 3 o’clock the left-hand, 6 o’clock the posterior, and 9 o’clock the right-hand position. P16 staining was used when necessary to distinguish between LSILs and HSILs, as recommended by the College of American Pathologists [3].

Participants with no dysplasia were invited for a follow-up HRA visit 2 years after the initial HRA. Participants with LSILs were invited for follow-up HRA after a year and participants with HSILs were treated and invited for follow-up HRA every 3–6 months until resolution of HSILs, after which they were invited for annual HRA visits.

In this analysis, participants were included if they had had ≥2 HRA visits and either no dysplasia or an intra-anal LSIL as the maximum (highest-grade) diagnosis at baseline. Participants were excluded from the analyses if they had a life expectancy of <12 months, if they had intra-anal HSILs or squamous cell carcinoma at baseline, or if they were treated for LSILs at baseline. Moreover, participants without dysplasia and those with LSILs at baseline were excluded if they did not return for a follow-up visit between 1–4.5 years (in those with no dysplasia), or 0.5–2.5 years (in those with LSILs) after the baseline visit.

Behavioral data (ie, smoking status) were self-reported by the patient. HIV-related data (ie, the use and duration of combination antiretroviral therapy [cART], duration of viral suppression, CD4 cell count at initial HRA, nadir CD4 cell count, plasma viral load at initial HRA, and history of an AIDS-defining illness) were obtained from the HIV Monitoring Foundation [23]. The study (reference W15_047 15.0058) was approved by the Medical Ethics Review Committee of the Academic Medical Center, Amsterdam.

Statistical Analysis

Baseline characteristics of the MSM with no dysplasia at baseline or a maximum diagnosis of LSIL were explored using descriptive statistics. Baseline characteristics were compared between participants with or without a follow-up visit between 1 and 4.5 years for those without dysplasia at baseline, and between 0.5 and 2.5 years for those with LSILs at baseline.

For the participants with no dysplasia at baseline, we calculated the overall cumulative incidence of SILs at the second HRA visit, and the incidences of LSILs and HSILs separately. We also estimated the overall incidence rate (IR) of SILs, as well as the IRs of LSILs and HSILs. Incident SILs were defined as the first diagnosis of SILs. Person-time was calculated from the date of the initial HRA until the date of the follow-up HRA for those participants without an incident event and until the estimated date of the lesion for those with SILs. It was assumed that the lesion occurred at the midpoint between the 2 visits. Logistic regression was used to assess determinants of incident SILs at the second HRA visit. Analyses were also performed to assess determinants of the LSIL and HSIL outcomes separately.

We performed 2 sets of analyses on participants with LSILs at baseline: an analysis at the participant level and one at the lesion level. At the participant level, the highest-grade lesion at the second HRA was defined as the overall diagnosis, meaning a participant was considered to have cleared LSILs only if not a single SIL was found at the second visit. We estimated the clearance rate in the period until the second HRA visit. Person-time was calculated from the date of initial HRA to the date of follow-up HRA for those participants without a clearance event and until the estimated date of clearance of the lesion for those with cleared LSILs. It was assumed that clearance occurred at the midpoint between the 2 visits. Logistic regression was used to assess determinants of clearance. No progression rate was estimated, because it is not possible at the participant level to distinguish between HSILs resulting from progression of LSILs found at baseline and truly incident HSILs.

At the lesion level, we estimated the progression from LSILs to HSILs and the clearance of LSILs, whereby 1 individual could be included in the analysis with ≥1 LSIL. The lesions observed at the first visit were linked to observations at the second visit based on location, with a 2-hour (clock face) range to allow for locating errors or imprecision. When a biopsy specimen could be linked to multiple locations at the second visit, the highest-grade diagnosis was used. Progression was defined as a lesion progressing from LSIL to HSIL, and clearance as the absence of abnormalities at the location where the LSIL was found at the first visit. LSILs biopsied at baseline at a location that was not deemed necessary to be biopsied at the second visit were assumed to have cleared. Multilevel logistic regression (correcting for correlation within the participant) was used to determine the determinants of progression and clearance at the lesion level.

The univariable logistic regression analysis included 1 demographic variable (age), 1 behavioral variable (smoking status), 7 HIV-related variables (current cART use, years on cART, years of viral suppression, nadir CD4 cell count, CD4 cell count at first HRA, viral load at first HRA, and history of an AIDS-defining illness), and the clinic where the HRA was performed. For CD4 cell count and HIV load, the measurements closest to the date of the initial HRA were used. Years living with viral suppression was defined as having a viral load persistently <200 copies/mL in tests from August 1999 onward, allowing for 1 blip in viral load <400 copies/mL between measurements where the viral load was <200 copies/mL. For samples tested before August 1999, the detection limit of the assay that was used for that sample was used as the cutoff for viral suppression [24].

All variables associated in univariable logistic regression analysis (at P < .2; Wald test) were included in the multivariable analysis, but the following a priori selected parameters were forced into the model: age at time of HRA, years of viral suppression, CD4 cell count, and nadir CD4 cell count. Multicollinearity was tested using the variance inflation factor. If multicollinearity was detected, 1 collinear variable was dropped from the model. The effect of this exclusion was checked in a sensitivity analysis, where the dropped collinear variable now replaced a previously included variable. We conducted all regression analyses in 2 ways: first, ignoring the fact that participants were sampled from 3 clinics (using logistic regression at the participant level and multilevel logistic regression with participants as a clustering variable at the lesion level), and second, acknowledging the 2-stage sampling process by using random-effects models, with the clinic as the clustering variable (multilevel logistic regression). Both models were compared to find the model with the best fit. In multivariable analysis, variables were considered significantly associated with the outcome if the P value was <.05. Statistical analyses were performed using Stata software (version 13.1; StataCorp).

RESULTS

Baseline Characteristics

A total of 1678 HIV-positive MSM underwent initial HRA between 12 February 2008 and 24 November 2015. Their baseline characteristics have been reported elsewhere [9]. Of the 807 men without anal dysplasia at baseline, 594 (73.6%) did not undergo follow-up HRA, 86 (10.7%) underwent follow-up HRA within 1 year after the initial HRA, and 20 (2.5%) underwent follow-up HRA >4.5 years after the initial HRA. A total of 107 MSM (13.3%) had a second HRA visit between 1 and 4.5 years after the first visit and were included in the analysis (Figure 1). Men included in the analysis were older (P = .006), were more frequently born in the Netherlands (P = .006), and had a lower median nadir CD4 cell count (P = .02) than the excluded men. Men from clinic A were more likely to be included in the analysis (P < .001) (Table 1).

Of the 392 men who had a maximum diagnosis of (intra-anal) LSIL, 199 (50.8%) did not have a second HRA visit, 39 (9.9%) had a second HRA visit within 6 months of initial HRA, 12 (3.1%) had a second HRA visit >2.5 years after the initial HRA, and 28 (7.1%) received treatment for LSILs at the initial HRA visit. A total of 114 men (29.1%) underwent follow-up HRA 0.5–2.5 years after the initial HRA and were therefore included in the analysis (Figure 1). Compared with the men who did not have a follow-up HRA visit between 0.5–2.5 years after baseline, the included men more often had a CD4 cell count >500/μL (P = .052) and an undetectable viral load (P = .04) (Table 1).

Participants Without Dysplasia at Baseline: Analysis on the Participant Level

The median time between the first and second HRA visits was 14.9 months (interquartile range, 12.9–23.9 months). At the second HRA visit, 64 of the 107 men (59.8%) remained free of abnormalities, 18 (16.8%) had an incident LSIL, 19 (17.8%) had AIN2, and 6 (5.6%) had AIN3 (Figure 1). The IR for SILs (LSILs + HSILs) was 30.0 per 100 person-years (95% confidence interval [CI], 22.3–40.5). The IR of LSILs was 12.6 per 100 person-years (95% CI, 7.9–19.9), and the IR of HSILs was 15.9 per 100 person-years (95% CI, 10.7–23.5).

Older age was significantly associated with incident SILs in both univariable (odds ratio [OR], 1.59 per 10-year increase in age; P = .03) and multivariable (adjusted OR [aOR], 1.64 per 10-year increase in age; P = .02) analysis (Table 2); no other variables were associated with incident SILs.

In addition, risk factors for LSILs and HSILs were assessed separately. For incident LSILs, results were comparable to those of the overall model: in multivariable analysis, older age was significantly associated with incident LSILs (aOR, 2.10 per 10-year increase in age; P = .01) (Supplementary Table 1). Viral loads were significantly associated with incident HSILs in univariable analysis (OR, 4.44; P = .04), but because viral load is strongly correlated with years living with viral suppression, it was not included into the multivariable model. In multivariable analysis, no factors were significantly associated with incident HSILs (Table 3).

Participants with a Maximum Diagnosis of LSIL at Baseline: Analysis at the Participant Level

The median time between the first and the second HRA visit was 12.9 months (interquartile range, 12.3–13.9 months). Of the 114 men with LSILs at baseline, 32 (28.1%) had HSILs at the second HRA visit, and LSILs had cleared in 47 (41.2%). The clearance rate was estimated at 45.1 per 100 person-years (95% CI, 33.9–60.1). None of the variables were was associated with clearance in the multivariable model, except for the clinic where the HRA was performed (Supplementary Table 2). Based on the large standard errors and 95% CIs, and based on the biologically implausibility that the clinic would be a determinant of clearance, we ran the model without the clinic as a variable. This model yielded similar results to the model including clinic (data not shown).

Participants With a Maximum Diagnosis of LSIL at Baseline: Analysis at the Lesion Level

At baseline, a total of 177 LSILs were identified among 114 MSM. Of these lesions, 29 lesions (16.4%) had progressed to HSILs by the second visit, and 87 (49.2%) had cleared. At the lesion level, lower CD4 cell counts (OR, 0.76 per 100-cell-increase; P = .02) were associated with progression in the univariable multilevel logistic regression analysis (Table 4). In the multivariable analysis, lower CD4 cell counts (aOR, 0.74 per 100-cell increase; P = .03) were significantly associated with progression. Clearance of lesions was associated with previous smoking (OR, 4.06; P = .04) in the univariable multilevel analysis. In the multivariable analysis, previous smoking (aOR, 4.11; P = .04) was still significantly associated with clearance (Table 5).

DISCUSSION

Participants Without Dysplasia at Baseline

In this study, we found that among men without SIL at baseline, 17% showed LSIL and 23% showed HSIL over 1-4.5 years of follow-up. Other studies reported cumulative HSIL incidences ranging from 6% in 2 years to 32%–40% in 3 years [10, 12, 21, 22]. We found that higher age was a risk factor for incident LSILs, but we found no risk factors for incident HSILs. Concerning HSIL prevalence, 1 previous study found no risk factors among HIV-positive MSM [11]. Other prevalence studies have reported various demographic, behavioral, and HIV-related risk factors that increased the risk of anal HSILs, including the presence of ≥1 (high-risk) human papillomavirus (HPV) type [10, 13–15, 19, 20], current use of cART [14], older age [12, 19], lower CD4 cell count (before starting cART) [12, 15, 19], AIDS-defining illness [19], and smoking status [10, 15–17]. Studies also reported various determinants that decreased the risk of anal HSILs, including current cART use [20], duration of cART use [9, 12, 18], living with viral suppression for >1 year [9], and older age [9]. In conclusion, risk factors were not collected consistently across studies, and none have been identified consistently in prevalence studies; the largest study to date only found weak associations [9].

Participants with a Maximum Diagnosis of LSIL at Baseline

In participants with LSILs at baseline, 28% showed HSILs during follow-up, and 41% had cleared LSILs. At the lesion level we found that 16% of the lesions had progressed to HSILs by the second HRA visit, whereas almost 50% had regressed spontaneously. To the best of our knowledge, this is the first study to assess progression of LSILs at the lesion level. Studies assessing progression at the person level found cumulative incidences of HSILs ranging from 7.2% at 12 months to 38.7% at 36 months [10, 12, 21, 22]. However, most of these studies assessed progression by combining participants without dysplasia and with a maximum diagnosis of LSILs at baseline, resulting in progression rates that include both incident and progressed lesions. Moreover, in analysis at the person level it is not possible to distinguish whether the HSIL found at a follow-up visit is progression of an LSIL found at an earlier visit or an incident lesion at another location. Therefore, we consider that our cumulative incidence at the lesion level gives a more accurate estimate of progression. On the other hand, follow-up at this level might give an exaggerated effect on the natural clearance rate. Performing a biopsy for diagnostic purposes may induce clearance. The clearance rate found in our study might therefore not portray the natural outcome of LSILs.

Higher CD4 cell counts were negatively associated with progression. Other studies assessed determinants of progression at the person level; these included the presence of (multiple) high-risk HPV types [10, 21], current use of cART, and having a stable sexual partner [21]. One study found no determinants of progression to HSILs [22]. Of the 177 lesions found at the first visit in our study, nearly 50% had spontaneously regressed by the second visit. At the participant level, this meant that all LSILs had cleared at the second visit in 41.2% of participants, with a clearance rate of 45.1 per 100 person-years. Previous smoking was associated with more clearance of LSILs. Because smoking is one of the risk factors for anal cancer [25, 26], this outcome was unexpected and does not fit with other evidence.

Strength and Limitations

To the best of our knowledge ours is the first study assessing progression and clearance of LSILs at the lesion level, including evaluation of possible determinants. Moreover, ours is among the few prospective longitudinal studies on the natural history of anal LSILs.

A limitation of our study is the relatively small sample size, partly due to the fact that many men had not undergone follow-up HRA. This, in turn, was partly because of the later start of performing HRA in 2 of the clinics (clinic B started in 2011, clinic C started in 2009). Therefore, at the end of data collection in clinic B and C, fewer men had had the opportunity to be screened for a second time and thus could not be included in this analysis. Moreover, some LSILs were treated after the initial HRA, further reducing the sample size, and limiting the statistical power of the analysis. The fact that a large proportion of men did not have a second HRA visit may have also resulted in selection bias, because the men included in the analyses differed in some aspects from the men who were not included (Table 1).

As a second limitation, to perform an analysis at the lesion level, we linked locations at the first and second visits, allowing for an uncertainty margin of ±2 hours of rotation of the anal canal. In some instances a lesion found at the first visit could be linked to multiple lesions at the second visit. For these lesions, we assumed the highest-grade diagnosis as the “true” diagnosis. However, this may have been incorrect and may have resulted in an overestimation of the risk of progression and an underestimation of the clearance rate. Moreover, progression rates could have been influenced by interobserver variability between pathologists.

As a third limitation, there were significant differences between the 3 clinics in the detection of progressed LSILs at the second HRA visit, with clinic B showing the highest cumulative progression risk. Various anoscopists performed HRA at the 3 clinics, and correctly identifying and biopsying abnormalities with HRA requires a long learning curve [27–29]. Differences in skill and learning curves between clinics may account for the observed differences in LSIL progression risk between clinics. Finally, findings in recent years have suggested that, in the cervical transformation zone, LSILs might not always precede HSILs, and HSILs may also develop directly after HPV infection [30, 31]. Because the anal and cervical transformation zones differ [32], the pathway of anal HSIL development is uncertain; it may involve progression from LSILs to HSILs, direct development of HSILs, or both. Follow-up studies, including HPV genotyping and molecular analysis of the host cell (epi)genome, in the respective LSILs and succeeding HSILs, suggestive of progression from LSILs to HSILs, could help determine the pathways in anal HSIL development.

In conclusion, incident LSILs and HSILs were common among HIV-positive MSM without dysplasia at baseline, and older men were at higher risk for incident LSILs. Among men with LSILs at baseline, nearly half of these lesions regressed spontaneously, and only a small proportion progressed to HSILs. Higher CD4 cell counts were negatively associated with progression; the odds of clearance were higher among participants who had smoked previously. There is debate concerning the clinical significance of LSILs, and treatment is in general not recommended. However, at the participant level, >1 in 4 participants with LSILs at baseline had HSILs identified at the second visit. Given the high incidence of HSILs in both participants without dysplasia at baseline and in those with LSILs, follow-up HRA seems warranted.

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

Acknowledgements. We thank Ramon van der Zee, Hans-Erik Nobel, Anders Boyd, Liza Coyer, and Renske Steenbergen for their help during this study, and Stichting HIV Monitoring for sharing data on human immunodeficiency virus–related parameters.

Financial support. This study was supported by Aidsfonds (grant 2014029; file P11119) and the DDL Diagnostic Laboratory.

Potential conflicts of interest. H. J. C. d. V. reports a grant from Medicine, outside the scope of the current work. J. M. P. received research funding from the Netherlands Organisation for Health Research and Development for human papillomavirus (HPV) vaccine studies, outside the scope of the current work. The institution of M. F. S. v. d. L. received study funding from Sanofi Pasteur MSD, research funding from Janssen Infectious Diseases and Vaccines, and an in-kind contribution for an HPV study from Stichting Pathologie Onderzoek en Ontwikkeling. M. F. S. v. d. L. was a coinvestigator in a Merck-funded, investigator-initiated study and an investigator on a Sanofi Pasteur MSD–sponsored trial, and he served on a vaccine advisory board for GSK.

Presented in part: Netherlands Epidemiological Conference, Groningen, the Netherlands, 13–14 June 2019.

References