Eduardo L. Franco
Recent advances in the molecular epidemiology
of cervical cancer
McGill University, Montreal, Canada
Cervical cancer is one of the m ost com m om neoplastic diseases affecting w om en, w ith a com bined w orldw ide inci-dence that is second only to cancer of the breast. In developing countries, how ever, cervical cancer ranks as the m ost im -portant of all fem ale neoplas.m s, w hereas in w estern devel-oped nations it ranks as the tenth m ost com m om m alignant disease [parkin et aI, 1993]
In N orth A m erica, the overall 5-year survival rate for w om en w ith cervical cancer is 65% , low er than that for breast cancer [H olleb et aI, 1991]. Taken together, the proper diag-nosis, treatm ent, and follow -up of the patients w ith cervical cancer or its precursor lesions place an im portant burden on the public health system , particularly in developing countries.
Epidem iologic studies conducted during the past three decades have consistently indicated that cervical cancer risk is strongly influenced by m easures of sexual activity, both of the w om an and her partner [Franco, 1991]. O ther risk factors include tobacco sm oking, parity, use of oral contraceptives, conditions leading to im m unodepression, diet low in beta caro-tene and vitam ins C and A , and infrequent pap sm ear testing [Brinton, 1992; M unoz and Bosch, 1992]. W hile there is s'till som e debate in the literature as to the actual im portance of the latter variables, no one questions the validity of the find-ings for sexual activity. A s a result. In the past 20 years labo-ratory scientists and epidem iologists have tried to identify the sexually-transm itted agent or agents that act as the interm edi-ate cause of cervical cancer.
M ost of the research conducted during the 60s and 70s attem pted to search for an etiologic role for the herpes sim -plex viruses (H SV ). A ltough H SV has proven to be carcino-genic in vitro and in vivo clinical studies eventually dem on-strated that only a fraction of cervical carcinom as contained traces (viral D N A ) ofH SV infection and epidem iologic stud-ies failed to dem onstrate an association betw een H SV and cervical cancer [review ed in Franco, 1991]. H ow ever, the m ost
im portant reason for the attention to be diverted from the H SV w as the em ergence in the 1980s of papilom m aviruses (H PV ) as the m ost likely cause 9f cervical cancer [Zurhausen, 1991]. Clinical and sub-clinical H PV infections are the m ost com m om sexually-transm itted diseases today. There are over 70 H PV types defined on the basis of D N A hom ology [V onkrogh, 1991]. Tw o m ajor groups are defined according to their epithelial affinity: the ones infecting the dry skin nd those infecting the m oist m ucosal areas of the body. G enital types are tipicaly divided into three groups based on the fre-quency of association w ith m alignant tum ors, and thus, the presum ed oncogenic potential. the low risk group includes types 6, 11, and 42-44, w hich are com m on low -grade SIL (LG -SIL), less so in high-grade SIL (H G -SIL), and practi-cally nonexistent in cancer specim ens. The interm ediate risk group is com prissed of types 31, 33, 35, 51 and 52, w hose com bined frequencies of association increase w ithin the SIL spectrum , but decrease in carcinom as. The high risk group includes H PV types 16, 18,45 and 56, w hich are strongly associated w ith carcinom as and exhibit diverse behavior w ith respect to H G SIL. The latter group seem s, in fact, to be com -posed of tw o different subgroups: high-risklH PV 16, w hich seem s to be equally predictive of H G -SIL and cancer, and high risklH PV 18/45/56, w hich are strongly associated w ith cancers but som ew hat uncom m on in H G -SIL [Lorincz et aI,
1992].
Interestingly, som e inconsistences em erged in early m olecular epidem iology studies of H PV and cervical cancer that used first generation D N A hybridization m ethods to de-tect the virus. Contrary to expectations, these studies found that cervical H PV infection w as not associated w ith sexual activity variables [V illa and Franco, 1989; Reeves et aI, 1989; . K jaer et aI, 1990], a paradoxical finding considering that cer-vical cancer risk in strongly associated w ith sexual behavior. A s Show n subsequently by us and others [Franco, 1991,1992;
S chiffm an and S chatzkin, 1994], by m easurem ents errors of H P V infection status w ere the cause for these incoherent find-ings. T hecniques to detect the presence of H P V in cervical cells have involved considerably, from sim ple scoring of
cy-tologic signs of H P V to im m unocythochem ical staining, nucleic acid hybridization m ethods, and m ore recently, the polym erase chain reaction (P C R ). M odern P C R protocols based on the so-cal1ed consensus prim ers (B ecause they am -plify defined regions in genes that are highly conserved across
H P V types) have sensiivity and specifity for epidem iologic studies [G regoire et aI, 1989; M anos et aI, 1990; V anderbrule et aI, 1990].U se of P C R has m inim ized the problem of m isclassification of viral status in epidem iologic studies.
. A consensus panel convened by the W orld H ealth
O rganizations's International A gency for R esearch on C an-cer (IA R C ) has recently concluded that there is now com pel-ling evidence, both fom the biologic and from the
epidem iologic standpoints, to consider H P V infecton as cause of cervical cancer [F ranco, 1992]. A variety of case-control and cohort studies conducted in the last five years have con-sistently show n that H P V infection is the strongest risk factor
for cervical cancer, w ith relative risks (R R ) in the 20-70 range. S uch a m agnitude is higher than that for the association be-tw een sm oking and lung cancer and is second only to that for the association betw een the chronic carrier state of hapatitis B infection and liver cancer, causal relations in cancer that are no longer challenged [F ranco et ai, 1994].
C ervical H P V infection detected by D N A hybridiza-tion techniches is found i 15% -40% of asym ptom atic w om en
of reproductive age [F ranco, 1991]. Interestingly, how ever, w hen additional cervical specim ens are taken from these w om en in fol1ow -up surveys the m ajoity of the infections are harbour to the sam e H P V type in subsequent specim ens [M oscicki et ai, 1992,1993; H ildesheim et ai, 1994; F ranco
et ai, 1994]. In additon, prospective epidem iologic studies have indicated that the risk of subsequent cervical neoplastia seem s to be proportional to the num ber of specim ens testing posi-tive for H P V [K outski et ai, 1992]. T hese findings suggest
that only persistent infections of the cervical epithelium m ay be the ones eventual1y triggering carcinogenic developm ent. C onsidering that there is now an ongoing debate concerning w hether H P V testing should be added to existing cervical cancer screening program s [N uovo and N uovo, 1991; R eid and L orencz, 1991; Johnson, 1995], it is im perative that
is-sues related to viral persistencec be adressed by epidem iologic studies.
T he difficulty how ever, resides in defining persistence . V iral typing plays an useful role in this regard, by providing a first screen of m ultiple specim ens from the sam e w om an, thus indicating w ich cases can be safely ruled out as being
tran-sient infections because different H P V types w ere found in consecutive specim ens. B ut, how w ould interpret cases w ith concordance of types during fol1ow -up? F or instance, w ould tw o H P V -16 (a com m on genital type) positive specim ens taken
a few m onths apart from the sam e w om an indicate persistent infection, or could they represent tw o consecutive transient infections w ith H P V -16? W e have recently poposed a solu-tion to this problem [F ranco, et ai, 1994] by dem onstrating
the aplicability of D N A sequencing techniques to detect m o-lecular variants of the H P V in the epidem iologic studies of the natural history of cervical neoplastia. T hese variants are identifyed on the basis of m utational patterns in the D N A , thus allow ing interpretation of persistence on firm er grounds.
O ncoing cohort studies incorporating detection of m olecular variants of H P V should eventually answ er questions related to the applicability of H P V testing as a screening tool for cervical cancer.
