Rate of cervical cancer, severe intraepithelial neoplasia, and adenocarcinoma in situ in primary HPV DNA screening with cytology triage: randomised study within organised screening programme.

Abstract

To assess the performance and impact of primary human papillomavirus (HPV) DNA screening with cytology triage compared with conventional cytology on cervical cancer and severe pre-cancerous lesions.
Randomised trial.
Population based screening programme for cervical cancer in southern Finland in 2003-5.
58 076 women, aged 30-60, invited to the routine population based screening programme for cervical cancer.
Primary HPV DNA test (hybrid capture II) with cytology triage if the result was positive or conventional cytological screening (reference).
Rate of cervical cancer, cervical intraepithelial neoplasia (CIN) grade III, and adenocarcinoma in situ (as a composite outcome referred to as CIN III+) during 2003-7 through record linkage between files from the screening registry and the national cancer registry.
In the HPV and conventional arms there were 95 600 and 95 700 woman years of follow-up and 76 and 53 cases of CIN III+, respectively (of which six and eight were cervical cancers). The relative rate of CIN III+ in the HPV arm versus the conventional arm was 1.44 (95% confidence interval 1.01 to 2.05) among all women invited for screening and 1.77 (1.16 to 2.74) among those who attended. Among women with a normal or negative test result, the relative rate of subsequent CIN III+ was 0.28 (0.04 to 1.17). The rate of cervical cancer between arms was 0.75 (0.25 to 2.16) among women invited for screening and 1.98 (0.52 to 9.38) among those who attended.
When incorporated into a well established organised screening programme, primary HPV screening with cytology triage was more sensitive than conventional cytology in detecting CIN III+ lesions. The number of cases of cervical cancer was small, but considering the high probability of progression of CIN III the findings are of importance regarding cancer prevention.
Current Controlled Trials ISRCTN23885553.

Full text

Page 1

RESEARCH
Rateofcervicalcancer,severeintraepithelialneoplasia,and
adenocarcinoma in situ in primary HPV DNA screening with
cytology triage: randomised study within organised
screening programme
Ahti Anttila, research director of the mass screening registry,1Laura Kotaniemi-Talonen, researcher,1Maarit
Leinonen, researcher,1Matti Hakama, professor,1Pekka Laurila, chief physician,2Jussi Tarkkanen, senior
consultant in pathology,2Nea Malila, director of the mass screening registry, professor,1,3Pekka Nieminen,
chief physician4,5
ABSTRACT
Objective To assess the performance and impact of
primary human papillomavirus (HPV) DNA screening with
cytology triage compared with conventional cytology on
cervical cancer and severe pre-cancerous lesions.
Design Randomised trial.
Setting Population based screening programme for
cervical cancer in southern Finland in 2003-5.
Participants 58076 women, aged 30-60, invited to the
routine population based screening programme for
cervical cancer.
Interventions Primary HPV DNA test (hybrid capture II)
with cytology triage if the result was positive or
conventional cytological screening (reference).
Main outcome measures Rate of cervical cancer, cervical
intraepithelial neoplasia (CIN) grade III, and
adenocarcinomainsitu(asacompositeoutcomereferred
to as CIN III+) during 2003-7 through record linkage
betweenfilesfromthescreeningregistryandthenational
cancer registry.
Results In the HPV and conventional arms there were
95600 and 95700 woman years of follow-up and 76 and
53 cases of CIN III+, respectively (of which six and eight
werecervicalcancers).TherelativerateofCINIII+intheHPV
arm versus the conventional arm was 1.44 (95%
confidence interval 1.01 to 2.05) among all women invited
for screening and 1.77 (1.16 to 2.74) among those who
attended. Among women with a normal or negative test
result, the relative rate of subsequent CIN III+ was 0.28
(0.04to1.17).Therateofcervicalcancerbetweenarmswas
0.75(0.25to2.16)amongwomeninvitedforscreeningand
1.98 (0.52 to 9.38) among those who attended.
Conclusions When incorporated into a well established
organised screening programme, primary HPV screening
withcytologytriagewasmoresensitivethanconventional
cytologyindetectingCINIII+lesions.Thenumberofcases
of cervical cancer was small, but considering the high
probability of progression of CIN III the findings are of
importance regarding cancer prevention.
Trial registration Current Controlled Trials
ISRCTN23885553.
INTRODUCTION
The incidence of invasive cancer is the most informa-
tive standard in the evaluation of cervical cancer
screening programmes and screening methods. Data
on cancer incidenceandmortality,however, are avail-
able for only some of the conventional cervical smear
based screening programmes.12Alternative technolo-
giessuchasliquidbasedcytologyorhumanpapilloma-
virus (HPV) screening or vaccination have been
proposed as possible means to improve prevention of
cervical cancer.23Most studies have relied on surro-
gate end points such as rates of cervical intraepithelial
neoplasia (CIN) or sometimes only cytological
findings.4These end points are highly variable
between programmes and age groups and do not
necessarily precisely represent the real effect on inci-
dence and mortality.25Longitudinal information on
the most severe surrogate endpoint markers6-9or on
the incidence of cervical cancer and mortality10is
rarely available for the newer methods.
We evaluated the impact of primary HPV DNA
screening with cytology triage on cervical cancer,
severe cervical intraepithelial neoplasia, or adenocar-
cinoma in situ (as a composite outcome referred to as
CINIII+).Weusedarandomiseddesignincorporated
in the routine population based organised screening
programme for cervical cancer in Finland. The refer-
ence test was conventional cytology. The evaluation
was based on the total number of cases of CIN III+
detected within five years after the index invitation.
METHODS
Recruitment
The study was based on follow-up during 2003-7 in
women randomised (1:1) to primary HPV DNA
screening with cytology triage or conventional
1Mass Screening Registry of the
Finnish Cancer Registry, Pieni
Roobertinkatu 9, FIN-00130
Helsinki, Finland
2HUSLAB Kätilöopisto Pathology
Laboratory, Helsinki
3School of Public Health,
University of Tampere, Tampere
4Department of Obstetrics and
Gynaecology, Helsinki University
Central Hospital, Helsinki
5Pathology Laboratory of the
Finnish Cancer Society, Helsinki
Correspondence to: A Anttila
ahti.anttila@cancer.fi
Cite this as: BMJ 2010;340:c1804
doi:10.1136/bmj.c1804
BMJ | ONLINE FIRST | bmj.com page 1 of 8

Page 2

cytological screening within the population based
screening programme for cervical cancer in southern
Finland between 1 January 2003 and 31 December
2005. We used data from eight municipalities and
two screening laboratories. In total 58282 women
wereinvitedtoparticipateinthescreeningprogramme
(fig 1). The screening invitations at the recruitment
phaseincludedthefiveyearlyinvitationstothescreen-
ing programme. Information on study design, screen-
ing methods used, and cross sectional screening
findings have been previously published.11-14
Randomisation and exclusions
Of the 58282 women, 29144 (50%) were individually
randomised to screening with the HPV DNA test and
29138 to conventional cytological screening as the
control arm. Randomisation was done two to
12 months before invitation. After randomisation,
206 women were excluded, 107 from the HPV arm
and99from theconventional arm(fig1).We included
data from 58076 women in the analysis.
Conventional cytology screening protocol
Thenationalscreeningprogrammeforcervicalcancer
in Finland was started in the early 1960s, and women
aged 30-60 are invited by letter to attend screening
every five years. Trained nurses or midwives usually
take screening samples in primary healthcare facilities
in the municipality. In the trial, cytological sampling
was the same in both arms: a VCE smear (vaginal,
cervical and endocervical samples) was taken with a
cytobrush from the junction and endocervix, and
with two spatulas from the ectocervix and vaginal for-
nices,andspreadonthesameobjectiveglass.Thefixa-
tive was 95% ethylalcohol and Papanicolaou staining
was used. A cytotechnician reviewed all the screening
smears and referred any abnormal findings to a cyto-
pathologist. The cytopathologist also re-evaluated 10-
20%ofthesmearswithnormalresults.Thecytopathol-
ogistwasresponsibleforthefinaldiagnosisandrecom-
mendations.
In conventional practice, a referral for colposcopy
and biopsyis basedon a cytology result equal to Papa-
nicolaou class III-V, roughly corresponding to low
grade squamous epithelial lesions or a worse (LSIL+)
in the Bethesda 2001 reporting system. A borderline
cytology (Papanicolaou class II, corresponding to aty-
pical squamous cells of undetermined significance
(ASC-US) or reactive cytological abnormality, or
both, in the Bethesda 2001) usually leads to a recom-
mendation for intensified screening, meaning a
repeated screening invitation after 12-24 months.
Women with negative results on colposcopy and his-
tologyarealsorecommendedforintensifiedscreening.
If women have borderline results on cytology two or
three times during intensified screening, they are
referred for colposcopy.15The threshold for treatment
inthecurrentprogrammeisaCINIlesion.InFinland,
CIN I lesions in women aged 30 or more are treated,
butwomenagedunder30tendtobefollowedupwith-
out treatment.
HPV screening protocol
In the HPV screening protocol women underwent an
HPV DNA test instead of conventional cytology as the
primaryscreeningtest.Aftertheconventionalcytologi-
cal smear was prepared (see above), the rest of the cel-
lular material from the endocervical sample was used
for an HPV test.11-14The transport medium containing
the screening sample was processed with the supplies
andreagentsofthehybridcaptureIIassay(QiagenFin-
land, Helsinki). Results of the detection assay were
expressed as a ratio of relative light units (RLU ratio)
with 1.00 (equivalent to HPV DNA concentration of
1 pg/ml) as the cut off for a positive result. The test cut
off and internal quality assurance procedures were
according to the manufacturer’s instructions.
In women with a negative HPV result, we did not
evaluatecytologyandthenextinvitationforscreening
was scheduled after five years as in the cytology based
programme. If the HPV result was positive, cytology
wasevaluatedwithatriagemethodtodecideaboutthe
management. The cytotechnician and thereafter the
cytopathologist carried out the cytology triage as in
the conventional arm. The technician analysing the
smear and the physicians at the clinical phase were
aware of the result of HPV DNA test. The treatment
threshold was the same as for conventional screening.
In the HPV screening protocol, any decision for
immediate referral was based on a clearly positive
cytological finding. In cases of borderline cytology,
Randomisation to conventional screening
(control arm) (n=29 138)
Excluded from analysis (n=99):
Died before invitation (n=37)
Emigration before invitation (n=34)
Earlier cervical cancer (n=28), diagnosed at
any time since 1953
Randomisation to HPV screening
(experimental arm) (n=29 144)
Excluded from analysis (n=107):
Died before invitation (n=31)
Emigration before invitation (n=41)
Earlier cervical cancer (n=35), diagnosed at
any time since 1953
Women eligible for routine screening in 2003-5 (n=58 282)
Randomisation
Excluded: n=0 (all women who received 5 yearly invitation were eligible for randomisation)
Enrolment
Allocation
Valid invitations (n=29 039)
19 221 attended and 9818 did not attend
(19 218 (99.98% of attenders) underwent
conventional cytology and 3 women (0.02%)
underwent HPV test as primary screening test)
Valid invitations (n=29 037)
19 449 attended and 9588 did not attend
(17 795 (91.5% of attenders) underwent HPV
test and 1654 women (8.5%) underwent
conventional cytology as primary screening test)
Analysis
53 cases of ≥CIN III observed after invitation,
33 among attendees and 20 among
non-attendees. Among attendees all cases
identified after cytology as primary screening
test. None lost to follow-up (all invitees under
register based follow-up until death, emigration,
diagnosis of cervical cancer, or end of follow-up)
(205 emigrated (censored at emigration) and
202 died from causes other than cervical cancer
during follow-up)
76 cases of ≥CIN III observed after invitation,
59 among attendees and 17 among
non-attendees. Among attendees 56 cases
identified after HPV test and 3 cases identified
after cytology as primary screening test. None
lost to follow-up (all invitees under register
based follow-up until death, emigration,
diagnosis of cervical cancer, or end of follow-up)
(222 emigrated (censored at emigration) and
213 died from causes other than cervical cancer
during follow-up)
Follow-up
Fig 1 | Enrolment, random allocation, completeness of follow-up, and analysis
RESEARCH
page 2 of 8
BMJ | ONLINE FIRST | bmj.com

Page 3

intensified screening was recommended. Intensified
screening was also recommended if the HPV result
was positive but the result of cytological triage was
negative. During the intensified screening in the HPV
arm, women were referred for colposcopy after
repeated borderline findings at cytological triage or
after three consecutive positive HPV test results even
if cytology was normal.
A small group of women, about 8.5%, in the HPV
arm underwent a conventional cervical smear test
because an HPV test was not taken, mainly for techni-
calreasons(theproperbrushortubewasmissingorthe
sample was not taken because of error). Few women
refused the HPV test. The HPV test was not available
to women in the conventional screening arm, even
though in a few cases (<0.1% of visits) it was carried
out in error.
Follow-up data and linkage
Thescreeningdatabasewaslinkedwiththepopulation
register and cancer register to obtain information on
vital status, emigration, cervical cancer, cervical intra-
epithelial neoplasia grade III (CIN III), and adenocar-
cinoma in situ. All the outcomes in our study were
based on data from the cancer registry. Within the stu-
died municipalities, all screening invitations and visits
were recorded. Many women were also screened out-
side the screening programme. Each year about the
same number of cervical smears are taken opportunis-
tically as in the organised programme, but these data
are not registered.1516
The follow-up started at the time of the first valid
invitation of the current randomised study period
(2003-5) and closed at emigration, death, diagnosis of
CIN III/adenocarcinoma in situ, or cervical cancer or
at 31 December 2007, whichever was earlier. The
screening data did not include the exact date of invita-
tion, so we used the month and year of the screening
visit to approximate invitation time among attendees.
Amongnon-attendeesweestimatedthemonthofinvi-
tationasthemedianmonthofvisitsintheirmunicipal-
ity during the invitational year.
Sample size
The a priori planned outcome included the incidence
of invasive cervical cancer and CIN III/adenocarci-
nomainsituat5,10,and15yearsthroughrecordlink-
age between screening and cancer registers. We
needed about a million woman years at risk for an
80% power to detect a 50% decrease in the subsequent
riskofcervicalcancer.12ConsideringdataonCINIII+
the statistical power to detect a 1.5-fold difference
among all invitees was about 70% (relative rate 1.5, α
0.05, one sided) and for detecting a twofold difference
amongthosewithpositiveresultswas95%(2.0,α0.05,
one sided) in current phase of follow-up.
Definitions of summary screening findings
We used the following terms: invitee—a woman who
gotaninvitation(apersonalletter)toattendscreening;
attendee—awomanwhotookpartinscreening(thatis,
a cervical sample was taken); and non-attendee—a
woman who was invited for screening but did not
attend.
In attendees, the main groups consisted of women
with positive or negative results on primary screening
tests. An RLU ratio <1.00 on an HPV test and Papani-
colaou class I on a cervical smear test were classed as
negativeresults.AnRLUratio≥1.00onanHPVtestor
anabnormalresultofthecervicalsmearwasclassedas
a positive result. Women with a positive result were
further grouped into those with a positive screening
episode and those who were recommended for inten-
sified screening.
Women with a positive screening episode were
referred for colposcopy and thereafter any histologi-
cally confirmed cervical intraepithelial neoplasia or
cervical cancer was detected at the index screening
round. Women were recommended for intensified
screening if the results of colposcopy or histology
were negative (that is, no cervical intraepithelial neo-
plasia diagnosed at the index episode among recom-
mended women) or if the result of the cervical smear
wasclassII(inconventionalscreening)orthecytology
triage result was normal or gave a borderline result (in
HPV screening).
Statistical methods
We compared patterns of CIN III and respective sen-
sitivities between arms with Poisson regression; the
numeratorwasthenumberofdetectedlesionsofinter-
est, and the denominator was the woman years at risk
in all women who were invited or attended. We calcu-
lated relative risk estimates (relative rate) for the HPV
screeningarmusingtheconventionalscreeningarmas
the reference. The 95% confidence intervals for the
relativeriskestimatesandpotentialeffectmodification
with age were assessed with likelihood ratio statistics.
The analyses were based on the intention to screen
principle throughout, both among invitees as well as
in more detailed subgroups of attendees.
The invitees were further grouped into attendees
and non-attendees and the attendees into those with
positive and negative test results during the first or
“index” screening round (see the definitions above).
Women remained in the same category throughout
the current follow-up period.
RESULTS
In the 58076 women with analysed data, there were
191218 woman years at risk. The average follow-up
was 3.3 years with a maximum of five years. In the
HPV screening arm, 29037 women were invited and
19449 attended. In the conventional arm 29039
women were invited and 19221 attended. The
woman years at risk among attendees comprised
67.0% of the person time in the HPV screening arm
and 66.3% in the conventional arm (fig 1 and table 1).
There were no marked differences in attendance or
follow-up time between study arms. There were more
women with a positive screening episode (that is, with
RESEARCH
BMJ | ONLINE FIRST | bmj.compage 3 of 8

Page 4

any diagnosis of cervical intraepithelial neoplasia or
cervical cancer) in the HPV screening arm than in the
conventional screening arm (110 v 72) and more
recommended for intensified screening at the index
screen (1244 v 1053).
Therewere76casesofCINIII+inwomeninvitedin
the HPV arm and 53 cases in those invited in the con-
ventionalarm(table 2).Figure2showsthecumulative
number of cases of CIN ΙΙΙ+ over follow-up. Figure 3
shows the corresponding cumulative incidence. Com-
pared with the conventional arm, the relative rate of
CIN III+ in the HPV arm was 1.44 (95% confidence
interval 1.01 to 2.05; table 2) in all women invited. In
attendeestherewere59casesofCINIII+lesionsinthe
HPVarmand33intheconventionalarm(1.77,1.16to
2.74). The relative rate of cervical cancer in the HPV
Table 1 |Number of women and woman years at risk with percentage distribution in cervical screening programme
HPV screening Conventional screening
Women
Woman
years
% of women
years
Women years
per womanWomen
Woman
years
% of women
years
Women years
per woman
Invitees29 037 95 553100.03.3 29 039 95 666100.03.3
Attendees19 44964 02567.0 3.319 22163 39666.3 3.3
Non-attendees958831 528 33.03.39818 32 27033.73.3
Screening group (among attendees):
Screening test positive1354 45447.13.4 1125 3766 5.93.3
Screening episode positive 1103700.6 3.4722350.4 3.3
Recommendationforintensifiedscreening*1244†41746.5 3.410533531 5.63.4
Screening test negative18 095 59 48092.9 3.3 18 096 59 63094.1 3.3
By age group (years):
Age 30-39:
Invitees7939 25 961100.0 3.37894 25 852100.0 3.3
Attendees4630 15 15258.4 3.34571 15 00058.0 3.3
Age 40-49:
Invitees8893 29 337100.0 3.38849 29 292100.0 3.3
Attendees5894 19 48566.4 3.35728 18 95164.7 3.3
Age 50-64:
Invitees 12 205 40 257100.0 3.3 12 29640 522100.0 3.3
Attendees8925 29 38973.03.3 892229 44572.7 3.3
*In women recommended for intensified screening, four in HPV group and 29 in conventional group had inadequate screening test.
†Out of 1244 women who were recommended for intensified screening in HPV screening arm, HPV test was done and was positive for 1151 women: 353 of them were Papanicolaou class II+ at
cytology triage, 794 were class I, and 4 had inadequate smear. HPV test was not done for 93 women, and recommendation was based on Papanicolaou class II+ result among them.
Months since index invitation
No of cases
Did not attend
0
10 203040 50 60
Months since index invitation
Positive screening episode
0
1020 3040 50 60
0
5
10
15
20
25
30
0
5
10
15
20
25
30
No of cases
Recommended for intensified screening Negative screening test
HPV screening
Conventional
Fig 2 | Cumulative number of cases of CIN III+ by months since invitation
RESEARCH
page 4 of 8
BMJ | ONLINE FIRST | bmj.com

Page 5

arm compared with the conventional arm was 0.75
(0.25to2.16)ininviteesand1.98(0.52to9.38)inatten-
dees, based on six and eight cases in invitees and six
and three cases in attendees.
Table 3 gives information on CIN III+ by main
groups at index screening. There was an increase in
the detection of cases of CIN III+ in the HPV arm in
both groups of women with positive screening test
results—that is, those with a positive episode and those
with a recommendation for intensified screening. In
women who had a negative screening test result the
relative rate between arms was 0.28 (0.04 to 1.17), sug-
gestingalowerrateofCINIII+forHPVscreeningthan
for conventional screening, though not significant.
In 1244 women recommended for intensified
screening in the HPV arm, 794 women had a positive
HPVtestresultbutanegativeresultoncytologytriage
(PapanicolaouclassI);353womenhadapositiveHPV
test result and at least borderline result on cytological
triage; four women remained with an inadequate
screening result because of inadequate cytological
triage; and for 93 women cytology was done as the
primary screening test. In the intensified screening in
these subgroups CIN III+ was subsequently detected
in 11, 15, zero, and one, respectively. These corre-
spond to 10 cases detected in the whole conventional
screening arm, all based on a borderline cytological
finding.
Further analyses by age groups indicated that there
was no marked variation in the patterns of CIN III+
between screening arms over age groups (table 4).
DISCUSSION
In a well established routinely organised screening pro-
grammeHPVtestingwithcytologytriagewasmoresen-
sitive than conventional cytology in the detection of
cervical intraepithelial neoplasia grade III and above
(CIN III+). We looked at primary HPV DNA test with
cytologytriagecomparedwiththeconventionalcervical
smear test. The study was based on a maximum follow-
up of five years since the index invitation.
Our earlier cross sectional studies on data from the
screening register also suggested higher sensitivity for
primary HPV screening with cytology triage com-
pared with conventional cytology when any cervical
intraepithelial neoplasia at the recruitment was used
as a surrogate indicator.1113The specificity of the two
testing protocols was similar. Our study adds longitu-
dinal information based on cancer registry files and
confirms a higher sensitivity of HPV screening com-
pared with conventional screening, yielding more
cases of CIN III+ within the recruitment period—that
is, the “index” episode—and also in the intensified
screening after a positive HPV test result. At the same
time the incidence of CIN III+ among women with
negative results seemed lower in the HPV arm than
in the conventional arm.
Strengths and limitations
The random allocation of the two screening methods
within the routine screening programme took care of
the comparability problems between arms. The aver-
age attendance rate in Finland is 72%, but in southern
Finland, where we conducted our study, it has been
constantly lower at 66%.17This relatively low atten-
dance rate is still a problem, and several earlier cohort
studies have shown more cervical cancer among
women who do not attend screening.2More effort is
needed to improve attendance. A few more cases of
cervical cancer were subsequently detected in non-
attendees in the conventional arm than in the HPV
arm, even though the corresponding number of
Table 2 |Numbers of cases of cervical cancer, adenocarcinoma in situ, and CIN III and relative
rate (95% confidence interval) by study arm for all women who were invited for cervical
screening and those who attended
Study group
No of cases
RR (95% CI) for comparison
between arms
HPV
screening
Conventional
screening
Cervical cancer, adenocarcinoma in situ, or CIN III
Invitees 76 53 1.44 (1.01 to 2.05)
Attendees 5933 1.77 (1.16 to 2.74)
Cervical cancer
Invitees68 0.75 (0.25 to 2.16)
Attendees63 1.98 (0.52 to 9.38)
Adenocarcinoma in situ
Invitees75 1.40 (0.44 to 4.73)
Attendees53 1.65 (0.40 to 8.04)
CIN III
Invitees 6340 1.58 (1.07 to 2.36)
Attendees 4827 1.76 (1.11 to 2.86)
Table 3 |Number of cases of cervical cancer, adenocarcinoma in situ, or CIN III with relative
rate (95% confidence interval) by study arm and screening result among women who
attended cervical screening programme
Study group
No of cases
RR (95% CI)
for comparison
between armsHPVscreening
Conventional
screening
Cervical cancer, adenocarcinoma in situ, or CIN III
Screening test positive57 26 2.17 (1.38 to 3.51)
Screening episode positive 30161.86 (1.03 to 3.49)
Recommendation for intensified screening 27102.67 (1.34 to 5.80)
Screening test negative270.28 (0.04 to 1.17)
Cervical cancer
Screening test positive53 1.65 (0.40 to 8.04)
Screening episode positive32 1.49 (0.25 to 11.3)
Recommendation for intensified screening211.98 (0.19 to 42.6)
Screening test negative10 NA
Adenocarcinoma in situ
Screening test positive52 2.48 (0.53 to 17.3)
Screening episode positive010.00 (NA)
Recommendation for intensified screening51 4.95 (0.80 to 94.8)
Screening test negative010.00 (NA)
CIN III
Screening test positive4721 2.22 (1.34 to 3.78)
Screening episode positive27 13 2.06 (1.08 to 4.12)
Recommendation for intensified screening20 8 2.48 (1.13 to 5.97)
Screening test negative16 0.17 (0.01 to 0.97)
NA=not available.
RESEARCH
BMJ | ONLINE FIRST | bmj.compage 5 of 8