Methods

We recruited MSM at Sexual Health clinics in Central and North Manchester. The study opened in
March 2013 and closed in March 2015 receiving ethical approval from the NRES Committee North
West – Greater Manchester North. Following informed consent, every participant underwent an
endoanal sample using a FLOQswab put into SurePath fixative (Becton Dickinson & Co.).

Sampling involved a FLOQswab premoistened with SurePath which was inserted beyond the
internal sphincter to 2–3 in, which was then withdrawn using a circular motion to sample all sides
of the anal canal [8]. Digital rectal examination was then performed using lubricant/lidocaine gel,
followed by the insertion of a disposable anoscope (Pelican Feminine Healthcare, Cardiff, UK).
Prior to anoscopic inspection, a swab soaked in 5% acetic acid was placed in the anal canal for
several minutes before being removed. The anus was then examined at
10 magnification with higher magnification used if required. Any
aceto white lesions were biopsied and classified by a specialist pathologist. Lugol’siodine was only
used in cases of uncertainty to determine the need for and the site of a biopsy. Anal liquid-based
cytology was processed using standard procedures. Slides were read blinded to other results, and
classified as negative, low-grade abnormality or high-grade abnormality. Biopsies were classified
according to UK practice as AIN1, 2 or 3; AIN1 and AIN2/3, which approximate to low grade
squamous intraepithelial lesions and high grade squamous intraepithelial lesions (HSIL) according
to LAST Guidelines [9]. All AIN2þ lesions (HSIL) were tested for p16 expression, in accordance with
LAST Guidelines, using CINtec PLUS (Roche mtm, Heidelberg, Germany) with positivity being ‘block
positive’ as defined by LAST.

Human papilloma virus testing

After cytological processing, the original residue SurePath samples were sent to the Virology
Laboratory at Manchester Royal Infirmary. One millilitre of SurePath was tested according to the
manufacturers’ instructions, and screened for high-risk HPV (HR-HPV) types using the Roche Cobas
4800 (Roche MD, Pleasanton, California, USA) which also provides a readout for HPV types 16 and
18. All HR-HPV positive samples were further genotyped for 16 high-risk and eight low-risk types
using the Greiner BioOne Papillocheck assay.

From the residue SurePath, a 500 ml aliquot was pretreated using proteinase K and DNA was
extracted from 125 ml using the automated Nuclesens easyMAG system (BioMerieux, Marcy
L’Etoile, France) rather than the manual check extraction kit recommended by the manufacturers.
Amplification and detection was carried out according to the manufacturers’ instructions. Cases of
AIN3 or worse (AIN3þ) were referred for surgical management, which generally involved excision,
whereas AIN grades 1 and 2 were placed on surveillance. Participants were routinely recalled at 6
months for a second examination in order to check for any lesions missed at the initial visit, as a
means of determining the sensitivity of the initial screening visit, rather than evaluating natural
history. Participants recruited October to December 2014 were recalled after 3 months. At the
initial visit, each participant was asked to complete a questionnaire which addressed awareness of
HPVand risk of anal cancer, and at the time of follow-up a further questionnaire was offered
requesting experience of anal cancer and willingness to comply with screening in the future.

Statistical analysis
In addition to descriptive statistics outlining the sample characteristics, an exploratory analysis
was performed to investigate the ability of liquid-based cytology to predict underlying AIN. A
binary indicator variable identifying biopsy results as AIN grade 2 or above (AIN2þ) was modelled
using a standard logistic regression. The primary predictor of interest, liquid-based cytology, was
dichotomized into low grade or worse vs. negative.

The model was then adjusted for covariates chosen a priori to be age, smoking status (never,
previous, and current), alcohol (never, previous, and current), HIV status, and HPV status (positive
or negative). Results are reported in the form of an odds ratio (OR).

Results

Screening data

We recruited 284 MSM, of whom 203 were HIV-positive and 81 HIV-negative. Demographic data
are shown in Table 1. The median ages of the groups were 42 years for HIVþve MSM and 38 for
HIV–ve MSM. The duration of anoreceptive intercourse was greater than 10 years in almost 90%
of HIVþve and 70% of HIV–ve MSM. Amongst the MSM, 281 out of 284 HPV tests yielded valid
results. There was a very high prevalence of HR HPV amongst MSM at 85.1% (239/281) including
88% (176/200) and 77.8% (63/81) for HIV-positive MSM and HIV-negative MSM, respectively. The
results of genotyping all of the HPV-positive samples is shown in Fig. 1 for each of the two groups.
The most prevalent type in both HIV-positive and negative MSM was HPV 16 with types 16 or 18
being detected in 45 and 35% of HIV- positive and negative MSM, respectively. HPV genotyping
also revealed that 74% of infections contained multiple high-risk types. The prevalence of types
16/18 detected using the full genotyping Greiner BioOne Papillocheck assay was lower compared
with the Cobas (Roche) assay, which may have been due to higher sensitivity of the latter assay.

Table 2 shows the screening results at baseline for cytology, HR-HPV and histology. One striking
finding was that amongst all MSM, 120 of 279 (43%) of those with adequate cytology had
abnormal results; 92 of 199 (46.2%) and 28 of 80 (35%) for HIV-positive and negative, respectively.
Amongst all MSM, high-grade and low-grade changes were seen in 10.56 and 31.34% of cases,
respectively. Despite the high prevalence of cytological abnormality amongst MSM, more than
one- third of the cases of AIN grade 2 or worse (AIN2þ) were associated with negative cytology,
including five of 17 AIN3þ. Amongst the MSM, the sensitivity of LBC to detect AIN2þ and AIN3þ
was 62.9% [95% confidence interval (CI) 50.5, 74.1] and 76.5% (95% CI 50.1, 93.2), respectively,
and the specificity was 60.0% (95% CI 51.0, 68.5) and 54.6% (95% CI 47.1, 62.0), respectively. This
low sensitivity prompted a repeat reading of all slides blinded to theoriginal read. This resultedin
260 of 284 identical reads.

Twelve out of 19 that were upgraded, had originally been reported as negative but only four of
these were associated with underlying AIN2þ. For HR-HPV, the sensitivity was 94.1% (95% CI 85.6,
98.4) and 94.4% (95% CI 72.7, 99.9) for AIN2þ and AIN3þ, respectively, but the specificity very low
at 16.9% (95% CI 10.9, 24.5) and 13.9 (95% CI 9.19, 19.8), respectively. Co-testing with both HPV
and cytology would have identified all but two AIN2þ, but it would have resulted in only 21 of 203
HIVþve and 14 of 81 HIV–ve MSM having combined negative results. Almost 90% would therefore
have required anoscopy referral, based on a single positive result. The relatively poor clinical utility
of LBC and HPV suggested by the descriptive statistics are reflected in the sensitivity and specificity
though the study was not formally powered to assess these with respect to the detection of
AIN2þ. The histopathology data for biopsies are also shown in Table 2. These represent the most
severe grade of lesion for any individual. Overall 204 of 284 (71.8%) of MSM underwent a biopsy
because of an anoscopic abnormality. Three microinvasive cancers were detected amongst HIV-
positive MSM. The prevalence of AIN grade 3þ was 6.9% (14/203) and 3.7% (3/81) amongst HIV-
positive and HIV-negative MSM, respectively. The corresponding prevalence for AIN2þ was 26.6%
(54/ 203) and 20.99% (17/81) and for AIN 1 45.81% (93/203) and 37.04% (30/81), respectively.
Immunohistochemistry for p16 was performed on 69 of 71 AIN2þ lesions. This showed 64.1%
(34/53) of AIN2 stained block positive, compared with 87.5% (14/16) AIN3þ.

Table 3 correlates the baseline screening HPV and cytology results with AIN newly diagnosed at
the first or second visit. The second visit when all screening tests were repeated was undertaken
as a means of identifying disease not detected at visit 1. A further 59 AIN lesions of any grade were
detected, of which only one was AIN grade 3. This indicated a high sensitivity in the detection of
underlying AIN3þ with 17 of 18 cases being identified at the initial screening visit. In general the
prevalence of types 16/18 amongst MSM with AIN reflected that in the entire screened cohort.
Around one-third of the HPV-positive AIN3þ and AIN2þ were positive for only non 16/18 HR-HPV
genotypes, suggesting that using types 16/18 as a triage for referral would be too insensitive.
Lesions that were newly detected at the second visit did not differ significantly from those
detected at the first visit, in terms of baseline HPV and cytology. It seems likely that most of these
lesions were missed at the initial anoscopy, including the single AIN3.

Some of this may have been related to anoscopy experience; however, all of the HRA was
performed by a single practitioner (A.S.) who had been trained in specialist centres and had
performed 50 anoscopies under direct supervision prior to the start of the study. The prevalence
of AIN2þ detected over the 4–6 month periods of the trial (23, 24, 22 and 28%) does not support a
significant learning curve during the study.

The exploratory analysis of LBC’s ability to predict AIN2þ in MSM indicated a significant (P ¼ 0.009)
increase in the odds of AIN2þ being present when LBC was borderline or greater. An OR of 2.37
(95% CI 1.24, 4.52), indicated that a LBC result of borderline or worse was more than almost twice
as likely to have underlying AIN2þ compared with those with negative cytology.

Regarding the covariates in the model none were significant; however, HPV-positive patients were
also twice as likely to have an AIN2þ biopsy result (OR ¼ 2.35 95% CI 0.73, 7.61 P ¼ 0.153).

Acceptability of screening

Using a scale of 0–10 (0 ¼ no pain, 10 ¼ worst pain/ discomfort ever) participants were asked to
report the level of pain they recalled feeling during the test and also following the test. The
average levels of pain reported during the test were 3.7 (range 0–9, HIVþ3.7, HIV–3.7), and
following, 3.8 (range 0–10, HIVþ3.5, HIV–4.6), with pain lasting a mean of 3 days. Pain and
discomfort were generally not ranked high, during or following the tests. The majority of
participants reported little or no psychological distress whilst waiting to receive their results. On a
scale of 0–10 (0 ¼ no distress, 10 ¼ extreme distress), the mean score was 2.5 with only four
individuals (4 HIVþ and 0 HIV–) reporting 8 or higher. Using a similar scale to measure distress
after receiving results, the mean score of distress was 2.3 with only five individuals reporting a
score of 8 or higher, all of whom were HIVþ and had received a positive result.

A survey of willingness to attend for anal screening as well as knowledge of HPV and anal cancer
was conducted amongst the participants. These data were obtained at the first visit, regarding
knowledge of HPV and anal cancer (Table 4). At the second visit, the acceptability of the
experience of being screened at baseline was asked about (Table 4). Initial questionnaires were
completed by 281 participants, and 167 completed the questionnaire at follow-up. Amongst all
those completing the knowledge questionnaire, over half had heard of HPV (63.0%, 177/ 281),
two-thirds had heard of anal cancer (76.2%, 214/ 281) and almost half were aware of a link
between the two. Overall 40.2% (113/281) knew they belonged to a group at increased risk of anal
cancer. Almost half of HIVþ MSM were aware of this and awareness was lower amongst HIV-MSM
(27.4%, 23/281).

If offered, 92.4% said they would attend anal screening in the future. Almost 70% of participants
stated they would agree to attend for annual screening, most of the remainder would agree to 2–3
yearly screening and only one would not agree to be screened again (Table 4). Amongst those who
completed the ‘willingness to be screened’ questionnaire, around 80% felt ‘very positive’ about
being screened in ANALOGY. Only 11 participants reported having experienced anal symptoms
prior to the study and nine had previously sought help from a health professional about them.

Discussion

The ANALOGY study has demonstrated that anal screening is both feasible and acceptable
amongst MSM who have anoreceptive sex. The prevalence of high- grade AIN in HIV-positive and
HIV-negative MSM was very similar to that reported in the meta-analysis of Machalek et al. [4]
which was based on eight studies, all of which were controlled by anoscopy. It should be noted of
course, that the high prevalence seen in ANALOGY reflects a lack of prior screening, with the
accumulation of AIN over a number of years. The three screen detected cancers provide evidence
of the ability to detect asymptomatic early cancer as well as precancerous lesions.

The advent of liquid-based cytology has improved anal cytology because faecal residue is filtered
out, but despite this, the high proportion of AIN grade 2 or worse, not detected by cytology,
suggests insufficient sensitivity to be of value as a stand-alone screening test. Although the
standard logistic regression modelling showed that abnormal cytology increased the odds of
underlying AIN2þ, the negative predictive value in particular was not high. This may be due to
inadequate sampling although very few slides were classified as unsatisfactory.

The sensitivity of cytology in this study (62.9% for CIN2þ; 76% for CIN3þ) was at the lower range of
that reported (69–93%) in a systematic review [10]. In a similar sized study of HIV positive and
negative MSM from the U.K. backed up with HRA, the sensitivity of anal cytology to detect AIN2/3
was 83% (52/63) [11]. It is worth noting that the repeat reading revealed a relatively small increase
in detection of underlying AIN2þ which suggests to us that the low sensitivity is not simply due to
inexperience. Sampling may be an issue because unlike cervical screening, the anal transformation
zone cannot be visualized directly unless done under anoscopic control. Although HPV testing is
clearly highly sensitive, it lacks specificity due to its very high prevalence (over 80%) in MSM.
Limiting HPV testing to types 16/18 would add specificity; however, only (59%) of AIN 3þ
would have been detected. This is reflected in the standard logistic regression modelling which
demonstrated a nonsignificant increased odds ratio for HPV-positive results. It therefore appears
from this study that co-testing with HPV and cytology would lack the necessary specificity. HRA
therefore emerges from our study as the screening modality with most clinical utility, particularly
for a selected and sufficiently high-risk group.

It is clear from the ANALOGY study that anal screening of a high-risk group is both feasible and
acceptable, with participants reporting that screening was well tolerated and that it would be
highly acceptable to attend in the future, and to be re-screened at regular intervals. Survey
participants had of course been enrolled in the study and were therefore likely to be motivated,
and positively biased towards the acceptability of screening. Before the study their knowledge
about HPV, anal cancer and risk appeared to be limited. If screening were to be introduced,
information addressing these knowledge gaps would be needed.

There are however a number of important considerations with respect to clinical and cost
effectiveness. The first is to consider which group could be regarded as being at sufficient risk to
warrant screening. The ANALOGY data confirm other data which show that HIV-positive MSM
have a high disease prevalence, and because they comprise only 5% of MSM and carry a far
greater risk of anal cancer, that this would represent the most cost-effective group to screen. A
national policy of screening and surveillance for all MSM would be very expensive and consume a
great deal of resource.

For screening to be considered an effective preventive strategy, a key issue is the successful
management of detected disease, both low and high grade. A follow-up study from the United
States of treated AIN3 has indicated the limitation of treatment in terms of failing to prevent
cancer [12]. Radical excision of AIN3 is frequently considered excessive, whereas more
conservative excision or ablative treatment, or indeed non- surgical treatment such as imiquimod,
carry a high risk of recurrence. Two important trials of treatment for high- grade AIN are in the
early stages; ANCHOR [13] in the United States will randomize different types of therapy, with
cancer incidence as the primary endpoint and LOPAC [14] in the United Kingdom will compare
laser ablation of AIN2þ with surveillance again with cancer as the primary endpoint. For the time
being treatment of AIN grade 3 is generally recommended because of the risk of progression to
cancer. The risk of progression at 5 years has been estimated in two recent follow-up studies of
HIVþve men to be 3.2 and 1.7% for AIN3 and high- grade anal cytology, respectively [6,15]. Over
20 years the risk of cancer could therefore be around 10%, and anal cancer detected as a result of
surveillance would probably be diagnosed at an earlier stage than that presenting with symptoms.
The case for surveillance of low-grade AIN is not clear cut. These lesions probably carry little risk of
cancer, but their considerable prevalence means that alarge programme of surveillance over many
years would risk accumulating a very large number of individuals under follow-up, and for some at
least, ongoing anxiety could affect their quality of life. Consideration would need to be given to
discharging men with AIN grade 1 to a routine interval examination, rather than undergo annual
surveillance.

Cost effectiveness is a crucial consideration in the provision of screening and the absence of any
economic data is a limitation of our study. An early cost-effectiveness study from the United States
had suggested that biennial screening using cytology to screen MSM would be associated with an
incremental cost-effectiveness ratio of $13 000 per quality adjusted life years saved [16]. A cost-
effectiveness analysis from the United Kingdom published in 2010 concluded that many of the
criteria for assessing the need for a screening programme for HIV-positive MSM were not met
[17], and the authors concluded that it was unlikely that screening for anal cancer would be cost
effective. In a more recent Canadian study of 400 HIV-positive MSM, using a decision analytical
model, the investigators concluded that despite higher unit costs, the limitations of both HPV
testing and cytology meant that anoscopy would be the most cost effective strategy [18]. To our
knowledge there are no data on interval cancers in the context of anal screening so defining cost
effective intervals for screening is difficult. A 5-year incidence of anal cancer of 2–3% in the
presence of AIN3 suggests that a randomized trial of screening to acquire reliable evidence would
need to be very large and require many years of follow-up.

It seems unlikely that a randomized trial of anal screening to evaluate the effectiveness and cost
effectiveness of anal screening and with an endpoint of mortality or even cancer incidence could
be successfully achieved. Our data and those of other studies indicate that amongst HIV- positive
MSM, screening would detect a sufficient amount of AIN grade 3 to be considered clinically
beneficial, but in so doing, the extent to which cancer would be prevented and lives saved remains
uncertain.

Clinical algorithms would need to be designed to reduce the risk of developing a large cohort of
men with low- grade change undergoing excessive surveillance. The implementation of a policy of
screening HIV-positive MSM by means of HRA would necessitate new services to be developed
which would be challenging in terms of training and healthcare costs, but we have probably
reached a point where this cannot be ignored.

Acknowledgements

H.K., J.P.: Contributed to the conception and design of

the study, interpretation of data, drafting of manuscript.

A.M.S., L.S.: Contributed to design of the study,

acquisition of data, drafting of manuscript. A.S.:

Acquisition of data, drafting of manuscript. M.G.:

Analysis of data, revision of manuscript. L.N., M.D.,

R.F.T.M., J.M.: Acquisition of data, revision of manu-

script. E.C.: Interpretation of data, drafting of manu-

script. All authors gave final approval of the manuscript.

Other contributors to the ANALOGY study: Sukthankar

A, Clinical Specialist, Manchester Centre for Sexual

Health, Central Manchester University Hospitals NHS

Foundation Trust. Higgins S, Clinical Specialist, Depart-

ment of Sexual Health and HIV, North Manchester

General Hospital, Pennine Acute Hospitals NHS Trust.

Burnett J, Research Nurse, Institute of Cancer Sciences,

University of Manchester.

Source of funding: the study was funded by the NHS

Cancer Screening Programme (now operated by PHE).

Conflicts of interest

There are no conflicts of interest.