Anal carcinoma: incidence and effect of cumulative infections

Introduction: Human papilloma virus (HPV) causes anal condyloma that is a risk factor for anal carcinoma. The incidence and mechanism of invasive anal carcinoma in patients with anal condyloma are prospectively determined.

Patients and methods: From 1993 to 2002, 228 consecutive patients (164 HIV positive) with anal canal condylomas were included in the study, after curing of their lesions. They were asked to attend follow-up visits at 3- or 6-month intervals. We checked for anal co-infection with syphilis, gonococci, viruses (Epstein-Barr virus, cytomegalovirus, herpes simplex, HPV types), and quantified Langerhans' cells (LC) in anal mucosa at baseline and during follow up. We cured and analysed relapsed condylomas during follow up (3-112 months; median 26). Serum HIV loads and CD4 T-lymphocyte counts were determined at each visit and the densities of LC in consecutive specimens from patients with cancers were compared with that for a matched control group (n = 23).

Results: Analysis of 199 patients showed high-grade dysplasia (HGD) in 13.6% of patients, more in HIV-positive (16%) than in HIV-negative (6%) patients at baseline. During follow up, 3.5% (7/199; six HIV positive) patients developed invasive carcinoma after 13-108 months and 112 (56%) patients relapsed condylomas. HIV and anal co-infection were identified as independent risk factors (P < 0.01) for HGD and cancer: odd ratio (95% confidence interval) of 9.4 (2.4-37.4) and 3.67 (0.95-14.2), respectively. LC densities in anal mucosa were lower in patients with invasive carcinoma than in controls.

Conclusion: The risk of invasive carcinoma in HPV-infected patients is increased by HIV and anal co-infection. Decreases in LC numbers in anal mucosa may favour this outcome.

Introduction

Anal cancer is rare in the general population. However, its incidence is increasing due to the increase in frequency of sexually transmitted diseases (STD) [^1-6]. Human papilloma virus (HPV) and HIV have been identified as possible risk factors for anal invasive carcinoma [^7-9].

Immediately following the infection of epithelial cells by HPV, antigen presenting cells - called dendritic or Langerhans' cells [^10-15] - increase in number within the mucosa, and the numbers of lymphocytes increase in the lamina propria to clear infection. If HPV is not cleared anal condyloma may develop. Condylomas are commonly considered as benign lesions [⁹], but as oncogenic types of HPV (HPV16, HPV18, HPV31) have been detected within carcinoma cells, the question of progression of condyloma to invasive carcinoma might be raised [^16-18]. HIV infection is thought to increase the risk of anal carcinoma by 30 times. Although a higher prevalence of oncogenic HPV has been demonstrated for HIV-infected patients, the mechanism by which HIV increases the risk of carcinoma is unknown. The survival rate of AIDS patients has increased following the introduction of antiretroviral therapy (HAART), and would be expected to be accompanied by an increase in the incidence of anal carcinoma. However, the risk of anal carcinoma has not yet been prospectively evaluated in patients with anal canal condylomas. The only prospective cohort study reported to date concerns women with genital HPV infection and showed a frequency of 0.16 cases of intraepithelial anal neoplasia per 100 person-years [¹⁹]. Among HPV-infected patients identification of those with an increased risk of anal carcinoma is an important aim for clinicians. In this prospective study, we report the incidence of high-grade dysplasia (HGD) or cancer in consecutive patients with anal canal condylomas. This risk was considered with respect to HIV, HPV types, number of condyloma relapses and anal co-infection plus HPV, together with systemic and anal mucosal immunity at baseline and during follow up. We show that the cumulative risk of anal carcinoma in HPV-infected patients remains low, but is increased by HIV and anal co-infections probably due to a chronic decrease in Langerhans' cells (LC) in the tissue.

Methods

Patients and study design

All patients referred to the department of coloproctology from June 1993 to June 2002 were recruited for a prospective study. Patients with invasive anal cancer at baseline were excluded. Patients were interviewed with a standardized questionnaire and underwent proctological examination at baseline, 1 month later and every 3-6 months thereafter as previously described [¹⁸]. Briefly, patients included in the study were those who presented condylomas in the anal canal (regardless of extended margin) that had been completely cured. Careful examination including anoscopy was performed under local or general anaesthesia at each visit and lesions, if any, were cured again and specimens were examined histologically; the patient was considered to be in remission if examination was normal. Time since inclusion in the follow-up program was recorded. If HGD was detected, the patient was asked to attend follow-up visits at 3-monthly intervals. If invasive carcinoma was detected, he or she was considered to the end point and received appropriate treatment for cancer. In the remaining cases, patients attended 6-monthly follow-up visits. If any of the programmed visits was missed, the patient was contacted by mail. He or she was considered lost to follow up if two visits were missed and no information was available at the time of analysis. Of the 228 patients referred, 136 HIV-positive and 63 HIV-negative patients were included and 29 patients were excluded due to evidence of an invasive tumour (three HIV-positive patients), failure to assess all tests and/or cure all lesions (12 patients) or because the patient was lost to follow up (14 patients). The study was approved by the local ethics committee (comité consultatif de protection des personnes dans la recherche biomédicale).

Materials

All anal samples were conserved in formalin for pathological examination. In a preliminary study, we used formalin-fixed and frozen biopsy samples for the detection of LC. Because similar results were obtained with frozen and formalin-fixed materials, we subsequently used only formalin-fixed material for the detection of LC. We used frozen tissue samples for the detection of viruses [cytomegalovirus (CMV), Epstein-Barr virus (EBV) and herpes simplex virus (HSV)]. These analyses made it possible to determine the number of current anal infections for each individual. All patients underwent testing for HIV-1 (antibody was detected by conventional enzyme immunoassay screening and confirmed by Western blot) and serum CD4 T-lymphocyte (CD4 T-Ly) counts were determined. If the HIV test was positive, HIV-1 RNA load was measured in plasma by either Amplicor Monitor assay or bDNA Quantiplex assay. This procedure was repeated at each visit unless the anal mucosa was normal.

Histological analysis

Condyloma, dysplasia and carcinoma were diagnosed on 4-μm paraffin-embedded sections, stained with HES according to standard histological criteria described elsewhere [^18-20]. The periodic acid-Schiff, Zielh and Grocott methods were used to detect microbiological agents. High-grade squamous intraepithelial lesions (`grade III anal intraepithelial neoplasia' or HGD) were defined as previously described [¹⁸] and were considered as the only precancerous lesion.

Immunohistochemistry

Antigen-presenting cells

LC were counted using the three-step peroxidase technique and a mAb (Immunotech; Marseilles France) directed against CD1a epitopes that best characterized LC. We analysed 5-12 serial sections for each patient with negative controls in which primary antiserum was replaced with phosphate-buffered saline, and positive controls in which representative sections of skin were stained using a similar procedure [¹⁸]. Briefly, morphometric analysis of LC was determined in normal anal mucosa with reference to ultrastructural characteristics showing an indented nucleus and Birbeck granules (tennis-racket shaped organelles) and without tonofilaments, desmosomes and melanosomes. The normal mean density of LC was found to be in the range 15-19/mm² anal mucosa. The minimum normal threshold for LC density is defined as 15 LC/mm² anal mucosa in the present study, quantified by using an ocular grid to enable counting of immunostained cells in 10 consecutive fields, at a magnification of ×400 (i.e., for a 2.8-mm long anal epithelium sample in each case); results are expressed as LC/mm². We analysed the distribution and number of anal LC in patients with invasive carcinoma (n = 7 in this cohort study), using samples taken during relapses in the follow-up period leading up to the diagnosis of invasive carcinoma, which were obtained from pathology files. Similarly as a control, samples from 23 patients matched for sex, HIV status and duration of follow up were selected at random from individuals without carcinoma and analysed: a total of 117 samples for 30 individuals corresponding to 2-7 (median, 3) periods. Mean value for each individual was calculated and results are expressed as the difference between two groups (controls versus patients with invasive carcinoma; see Table 1).

Detection of CMV

A mouse mAb directed against CMV clone E13 (Biosys, Compiègne, France) was used and the specificity was checked [^10,18].

In situ hybridization (ISH) and in situ PCR for the detection of viruses

Screening and typing were performed according to our previously published protocol [^10,18].

HPV

For ISH, biotinylated and fluorescein isothiocyanate (FITC)-labelled genomic DNA probes were labelled by nick translation. First a mixture of various genomic HPV DNAs was used to screen for the presence of HPV. If positive, the patient was declared HPV positive. This mixture has been checked in the laboratory to detect specifically types 6, 11, 15, 18, 33, 31, 16, 35, 39, 40, 42, 45, 52, 53, 54, 59. Afterward, in HPV-positive cases specific probes for the types of HPV were used. The 20-mers MY11 and MY09 (Perkin Elmer, Norwalk, Connecticut, USA) were used as primers for PCR-ISH only for samples that appeared to be HPV negative by ISH alone and HPV was considered to be undetectable if both ISH and PCR-ISH were negative [^10,18].

EBV and HSV

A biotinylated oligonucleotide probe (Argène) complementary to the two nuclear EBER RNAs of EBV and a biotinylated genomic DNA probe (Enzo Diagnostic, Farmingdale, New York, USA) corresponding to a mixture of two clones, HSV1 and HSV2 DNA, were used.

Gonorrheae identification

Neisseria gonorrheae was detected in rectal secretion as a Gram-negative diplopcoccus and was identified in a 24-h culture by using specific antibody.

Statistical analysis

The characteristics of the patients were recorded, using counts and percentages for categorical variables, and median and range or mean ± SD, as appropriate, for continuous variables. All statistical analyses were performed on actuarial data files in 2002. The following subgroups were identified: individuals with and without HGD or invasive carcinoma (cumulative status), HIV-positive and -negative, with and without condyloma relapse, with and without anal co-infection. Categorical variables were compared between groups using the χ² test or Fisher's exact test, as appropriate. Continuous data were compared using Student's t test or the Mann-Whitney U test, as appropriate. Differences between groups were considered to be significant if P < 0.05. All significant characteristics were considered to be univariate risk factors and crude relative risks (RR) were estimated for each factor. To identify independent risk factors for anal carcinoma, we included all risk factors identified in univariate analysis with P = 0.1 in a multiple logistic regression model. The final model was constructed using a stepwise selection procedure. This enabled us to identify risk factors for anal HGD and cancer. In analyses of the effect of additional anal infection, the individual was considered to have additional infection if at least one more microbiological test apart HPV was positive. As this was found to be the case only in HIV-positive individuals, we considered three classes in the multiple logistic regression tests: HIV-negative, HIV-positive patients with additional anal infection and HIV-positive without additional anal infection. To take into account the data obtained during follow up, microbiological test results were corrected and the results obtained with samples taken during relapses were considered when establishing relative risk for HGD. For all analyses concerning patients' characteristics including HGD at baseline, we used the first HIV load determination, and for those events occurring during the follow up the last HIV load or mean of these-if several within 3 months prior to the event. Adjusted odds ratios (AOR) were determined from the estimated regression coefficients.

Results

Characteristics at enrolment

The main characteristics of 199 patients who had undergone all of the examinations required and were included in the final analysis are shown in Table 1. At baseline, 67% of patients had condyloma extension to the margin (versus 51% during the relapses); none of the patients had invasive carcinoma and 26 (13%) displayed HGD. The frequency of HGD did not differ significantly between HIV-positive and HIV-negative patients (16.4% versus 7.6%). The times required to obtain condyloma-free anal mucosa at baseline were 9 ± 4 weeks for HIV-positive and 7.1 ± 5 weeks for HIV-negative patients (P = 0.04).

Table 1

Outcomes during follow up

The median (range) duration of follow up was 23 (3-108) months. Patients with HGD or cancer were followed for a longer period (35.1 ± 10 versus 21.2 ± 7 months; P = 0.001) than the other patients, whereas mean time to the condyloma relapse was not significantly higher in these groups (9 ± 3 versus 10 ± 5 months, not significant), corresponding to half the follow-up period.

HGD and cancer

During the follow up, HGD was detected by histological examination in 38 (19%) patients (32 in HIV-positive and six in HIV-negative patients) in at least one examination. Seven (3.5%; six HIV-positive men aged 26-53 years, one HIV-negative woman aged 62 years) of these patients developed invasive carcinoma in the anal canal, with extension to the margin (two patients). These seven patients all showed HGD before the development of carcinoma-six at baseline and the seventh 12 months after entry in the study. The invasive carcinomas occurred 13-108 months after entry to the study. Considering HGD status at baseline, six of 26 (23%) patients with, and one of 173 (0.5%) without developed invasive carcinoma (P < 0.01). Two of the six HIV-positive patients with carcinoma had missed two or more visits before the diagnosis of carcinoma and had the most advanced tumours. Considering all patients with HGD or invasive carcinoma together, there was no significant difference between patients with and without HGD and cancer in terms of age, sex, body mass index, anal intercourse or drug abuse (Table 1). In contrast, current anal co-infections and oncogenic HPV types were more frequent in patients with HGD or cancer, and these patients had fewer LC in the anal mucosa than the controls (Table 1).

Effect of condyloma relapse

During follow up, 112 of the 199 patients experienced at least one relapse and 34 patients presented at least three relapses. These were more frequent in patients with HGD or cancer (Table 1). Seventeen of the patients with HGD or cancer (45%) had three or more relapses, versus 17 (10%) without HGD or cancer (P < 0.001). The interval between relapses was shorter and the mean number of relapses higher in individuals with HGD or cancer than in individuals with neither of these conditions.

Effect of anal co-infection

Any anal infection other than HPV was considered as co-infection. This was observed in 45 patients (30 at baseline, 15 during follow up): HSV (n = 32), CMV (n = 17) EBV (n = 9) and N. gonorrheae (n = 1). In these patients, the overall frequency of condyloma relapse was twice that of patients without additional anal co-infection (82% versus 45.7%; P < 0.001) and the cumulative frequency of HGD and cancer was three times higher than that in the other patients (37.7% versus 11.5%; P = 0.0001). Additional anal co-infection was identified as an independent risk factor for cancer and HGD (Table 2).

Table 2

Oncogenic HPV types.

Oncogenic HPV were type 18 (n = 8), 16 (n = 18), 31 (n = 2) and 33 (n = 4) in HIV-positive patients and type 16 (n = 6) or 18 (n = 1) in HIV-negative patients. Considering all patients together, these oncogenic HPVs, were detected in one-third of patients with HGD or cancer (Table 3).

Table 3

Effect of HIV

A positive HIV test at baseline was identified as an independent risk factor for cancer and HGD (Fig. 1). The risk was higher in HIV-infected individuals with anal co-infection. Serum HIV load affected the frequency of relapses and the likelihood of HGD and cancer whereas the use of one or more antiretroviral drugs in combination had no significant effect on the incidence of anal carcinoma. Briefly, patients with more than 200 HIV RNA copies/ml serum, close to the time of condyloma relapse (within 3 months before relapse), were more likely to go on to develop HGD or cancer (Table 3). Invasive cancer was found only in patients (6/44) with HIV loads above 1000 HIV RNA copies/ml as compared with none of 119 patients with viral loads lower than this.

Fig. 1

Effect of anal dendritic cells

The density of LC in anal mucosa at baseline was significantly higher in patients without HGD or cancer (Table 1). Low density of LC, defined as fewer than 15 LC/mm² [^{10, 11}] was a risk factor for HGD and cancer (Table 2). We determined LC density at each relapse before the occurrence of invasive carcinoma in seven patients and showed LC diminution over the time whereas in the controls (19 HIV-positive men, two HIV-negative men, two HIV-negative women), LC density remained at a plateau above the cut-off point. Individual values of LC density in anal mucosa decreased during follow up (r, 0.35; P < 0.003) in patients with cancer, reaching zero at the time of cancer diagnosis (Fig. 2). Percentages of LC diminution over the normal cut-off value (15 LC/mm²) were significantly higher in patients with (47 analyses from seven patients) than in those without cancer (79 analyses from 23 patients): 50 ± 32% versus 6 ± 6; P = 0.002.

Fig. 2

Discussion

In HPV-infected patients, the incidence of HGD was 15.5% and that of invasive carcinoma was 3.5%, corresponding to 0.1 per 1000 person-years of observation, respectively. HIV infection and anal co-infection are independent risk factors for these outcomes. A significant decrease in the number of antigen presenting cells in anal mucosa over a prolonged period before the occurrence of carcinoma may be involved in the mechanism of carcinogenesis. Thus, careful surveillance should be carried out in HPV-infected patients. Our results suggest also that LC may protect anal mucosa against carcinoma. HPV causes benign lesions, known as condyloma. Since HPV infection is highly prevalent in the general population the problem is how to identify those individuals with the highest risk of anal carcinoma. The relative risk of developing invasive carcinoma has been estimated to be between two and eight times higher in individuals with benign condyloma. This phenotype is easy to detect. However, anal carcinoma is four times more frequent in the canal than in the margin and can be detected only by anoscopy. Epidemiological series have estimated the risk of cancer only with respect to anal condyloma in the margin [^1-6]. Piketty et al. [²¹] performed smear cytology without anoscopy, to detect HPV and HGD without histological methods. They reported a frequency of 18% for HGD, with no invasive carcinoma, in HIV-infected patients; this is lower than the frequency currently detected by using anoscopy and careful proctological examination. Conley et al. [¹⁹] showed a higher risk of developing vulvovaginal or anal HGD in patients bearing condyloma or being HIV positive. These data are supported by our present results and others from the literature [^{19, 22}] indicating HIV infection as the main additional risk factor for developing anal carcinoma (Fig. 1). The invasive carcinoma in our series occurred despite careful screening; without the follow-up program this incidence may have been higher, or the cancers may have been discovered at more advanced stages. We, therefore suggest that standard follow-up screening should include anoscopy and biopsies to detect HGD in the anal canal. Although anal intercourse was not identified as an independent risk factor for carcinoma in our series and in others [^5,18,20], most individuals exposed to STD had experienced more than one genital or anal infection during their lifetime [⁵]. They clearly have a significantly elevated risk of anal cancer ([⁵], Table 3).

Furthermore, considering additional anal infection and a high rate of anal condyloma relapse together, we postulated that anal tissue immunity might be altered in patients with invasive carcinoma. Intraepithelial LC present antigens to intraepithelial lymphocytes [^23-26] in anal mucosa and their numbers may increase in patients with HPV infection [^10,13,14,27]. LC stimulate CD4 and CD8 T-Ly in the lamina propria [^14,15,27,28]. LC density was low in all individuals with carcinoma. The number of LC decreased during follow up, reaching zero in the seven individuals who developed carcinoma, whereas they remained at a plateau in matched individuals without cancer. The reason for this chronic failure in LC stimulation is unclear. Several lines of evidence suggest that HIV targets LC and integrates into these cells [^{13,14,18,28-30}], impeding their stimulation. Nevertheless, these longitudinal data suggest that immune cell alteration may precede carcinogenesis. If this is indeed the case, the best means of protecting HPV-infected individuals against carcinoma would not be HPV-vaccine therapy but the administration of drugs with specific antiviral or LC-stimulating effects. In HIV patients with anal carcinoma, HIV loads exceeded 1 × 10³ HIV RNA copies/ml serum at the time of cancer diagnosis and it is thought that HIV itself favours mucosal HPV activity and HPV-induced damage [^31-34] as assessed by a high risk of condyloma relapse and HGD occurrence. The natural history of HPV infection appears to be different in HIV-infected than in uninfected subjects. HIV-positive individuals were more likely to develop incident anal carcinoma than HIV-negative individuals in this and other studies [^19,24,35]. This suggests that HIV infection itself confers an additional risk of developing intraepithelial lesions. We therefore suggest that serum HIV quantification is a valuable tool for selecting among HPV infected individuals those with anal dysplasia. This study has several limitations. Data were gathered between 1993 and 2002, corresponding to a period of change in HAART. Most patients took one or two antiviral drugs until December 1995, and the first protease inhibitors were administered in 1996. This may account for the lack of effect of combinations of drugs on the incidence of anal carcinoma. In some patients, HIV load varied from one visit to another, and taking a value for HIV load determined 3 months before the relapse as a possible predictive factor is open to criticism. Although we could not exclude the possibility that HGD occurred in patients without condyloma relapse, this seems unlikely because of the long follow-up period and anoscopy performed in all individuals. In summary, we found invasive carcinoma in the anal canal of patients with only small numbers of LC in the anal mucosa. As this is routinely the case in HIV-infected patients with anal condyloma, detectable HIV load in the serum and anal co-infection, surveillance at 3- to 6-month intervals seems to be necessary for as long as HIV RNA is detectable in the serum. We advise anoscopy for the detection of condylomas and recommend instrumental or surgical resection of all lesions. Even if cytology is performed, tissue samples should be examined histologically, with particular attention focused on HGD. Treatments stimulating LC should be evaluated after condyloma cure.

Acknowledgements

We thank T. Aparicio and D. Chosidow for their contribution of information from clinical records.

Sponsorship: Supported by the ANRS (The French National AIDS Research Agency).