,---------------------

I I
Epidemiology series
EPIDE:'I.UOLOGY SERIES

2002 02
Descriptive studies: what they can and cannot do

David A Grimes, Kenneth F Schulz

Descriptive studies often represent the first scientific toe in the water in new areas of Inquiry. A fundamental element
of descriptive reporting is a clear, specific, and measurable definition of the disease or condition in question. Uke
newspapers, good descriptive reporting answers the five basic W questions: who, what, why, when, where ••• and a
sixth: so what? Case reports, case-series reports, cross-sectional studies, and surveillance studies deal with
individuals, whereas ecological correlational studies examine populations. The case report Is the least-publishable unit
in medical literature. Case-series reports aggregate individual cases in one publtcation. Clustering of unusual cases in
a short period often heralds a new epidemic, as happened with AIDS. Cross-sectional (prevalence) studies describe
the health of populations. Surveillance can be thought of as watchfulness over a community; feedback to those who
need to know is an integral component of surveillance. Ecological correlational studies look for associations between
exposures and outcomes in populations-eg, per capita Cigarette sales and ·rates of coronary artery disease-rather
than in individuals. Three important uses of descriptive studies include trend analysis, healtfl.care planning, and
hypothesIs generation. A frequent error In reports of descriptive studies is overstepping the data: studies without a
comparison group allow no inferences to be drawn about associations, causal or otherwise. Hypotheses about
causation from descriptive studies are often tested in rigorous analytical studies.

Descriptive studies have several important roles in all-terrain vehicle ll or snowmobile,l~ especially when
medical research. They are often the first foray into a new drunk; can be lethal.
disease or area of inquiry-the first scientific "toe in the What is the condition or disease being studied?
water". I They document the health of populations and Development of a clear, specific, and measurable case
often prompt more rigorous studies. Since descriptive definition is an essential step in descripth'e epidemiolog~t.
studies are often reported/ clinicians need to know their \'(!ithout such a description, the reader cannot interpret
uses, strengths, and weaknesses. the report. Some conditions, such as fractures, can be
A descriptive study is "concerned with and designed overt. Other diagnoses might be challenging: mUltiple
only to describe the existing distribution of variables, sclerosis, systemic lupus erythematosus, and pelvic
without regard to causal or other hypotheses."3 The key inflammatory disease (salpingitis), for example. By use of
qualifier about causal hypotheses is sometimes forgonen the consensus or Delphi panel B approach rather than
by investigators, resulting in erroneous conclusions. Here, evidence, some organisations have promulgated case
we provide an overview of the advantages and definitions that have subsequently been sho\\n to be
disadvantages of descriptive studies, provide examples of invaJid. H For instance, evidence indicates that vaginal
several types of descriptive study, examine their clinical discharge and a raised erythrocyte sedimentation rate
uses, and show how they can be misinterpreted. predict salpingitis, I S yet these predictors are not included
in widely-used diagnostic criteria. It
The descriptive triad-or pentad? Generally, stringent criteria for case definitions are
Five "W" questions desirable. Adminedly, if only the more se\'ere cases of
Traditional descriptive epidemiology has focused on three disease are targeted, milder or earlier cases will be missed.
key features: person, place, and time, ~ or agent, host, and Although this approach ine\itably leads to some loss of
environment. 5 An alternative approach is that of infonnation, the trade-off is bener specificity; severe cases
newspaper coverage. Good descriptive research, like good of a disease are less likely to be confused with other
newspaper reporting, should answer five basic "\V" conditions than are mild cases. An example would be the
questions-who, what, why, when, and where-and an stringent case definition used for toxic shock syndrome,
implicit sixth question, so what? \vhich requires involvement of multiple organ systems. I ';
Who has the disease in question? Age and sex are l\10re recently, expanding the case definition of AIDS has
universally described, but other characteristics might be yielded a sudden surge in "new" cases. 1:
important too, including race, occupation, or recreational \'<'hy did the condition or disease arise? Descripth'e
activities. The risk of venous thromboembolism, for studies often provide clues about cause that can be
example, increases exponentially with age. 6 Only 1% of pursued with more sophisticated research designs (panel).
breast cancers arise in men, but Klinefelter's syndrome or When is the condition common or rare? Time prO\ides
a family history of breast cancer increase their risk."";' Race important clues about health events. The prototype might
affects the risk of leiomyom as of the uterus. 9 Commercial be the outbreak of gastroenteritis soon after ingestion of
fishing remains a risky business,tC> and having fun with an staphylococcal toxin. Some temporal relations can be
long-eg, vaginal adenosis and clear cell carcinoma of the
Lancet 2002; 359: 145-49 vagina appeared years after intrauterine e"-lJosure to
diethylstilboestrol. IS Furthennore, cen;cal and other
Family Health Intemational, PO Box 13950, Research Triangfe epithelial cancers develop decades after infection with
Park, NC 27709, USA (0 A Grimes MO, K F Schulz PhD) human papillomavirus, and births and deaths from
Correspondence to: Dr David A Grimes pneumonia and influenza have regular seasonal patterns,
(e-mail: [email protected]) as might spenn counts.l9

THE I...A.!"lCET· Vol 359 • January 12, 2002· www.thelancet.com 145

/-
EPIDEMIOLOGY SERIES
•
Cross-sectional (prevalence) studies
Examples of early leads from descriptive studies Prevalence studies describe the health of populations. For
Clinical observation Underlying association
example, in the USA, periodic surveys of the health status
Hepatocellular adenoma in young Exposure to high-dose
of the population are done by the federal government-eg,
women oral contraceptives
the Health Interview Survey and the Health and Nutrition
Blindness in newborn infants High ambient oxygen
Examination Survey. Analogous to the decennial census,
concentrations in
these studies provide a snapshot of the population at a
incubators
particular time.
Kaposi's sarcoma in young men Infection with HIV-l
Prevalence studies can be done in smaller populations
Angiosarcoma of the liver in Industrial exposure to
as well. For example, the results of a survey done in a
employees vinyl chloride
Puerto Rican pharmaceutical factory indicated an
Cataracts, heart defects, Maternal infection with
exceptionally high prevalence of gynaecomastia among
and deafness in newborns rubella during pregnancy
employees (figure). This finding led to the hypothesis that
exposure to ambient oestrogen dust in the plant might be
the cause; serum concentrations of oestrogen lent support
Where does or does not the disease or condition arise? to the hypothesis. After improvements in dust control in
Geography has had a huge effect on health. Living close to the factory, the epidemic disappeared. 34 Similar
rodents and insects (and thus their parasites) has shaped prevalence studies have linked gynaecomastia with feeding
both medical and political history.20 Living where drinking ofrefugees 35 and tainted food. 36
water has high fluoride protects against dental caries,21 Although generally distinguished from cohort and case-
whereas residing downwind from a lead smelter is less control studies, the cross-sectional study can be thought
salutary.22 Latitude plays a part in both multiple sclerosis 23 of as the case-control analogue of a population cohort
and vitamin D deficiency; 24 sunlight might decrease25 or study.37 Since both exposure and outcome are ascertained
increase 26 cancer risk. at the same time (the defining feature of a cross-sectional
So What? The implicit "W" relates to the public study), costs are small and loss to follow-up is not a
health effect. In view of the proliferation of descriptive problem. However, because exposure and outcome are
reports/ what is their import? Is the condition a current identified at one time point, the temporal sequence is
and timely one? Is it serious? Are large numbers involved? often impossible to work out. An exception would be
Are its societal implications broad? Has it been studied long-standing exposures, such as sex or blood type, which
beforer Although many descriptive reports herald unquestionably preceded the outcome. For exposures that
new illnesses or monitor health, the net effect of vary, information of aetiological relevance from the past
others might be only thicker curricula vitae at the expense might be more useful than current information.); ,
of thinner forests.
Surveillance
Types of descriptive studies Surveillance is another important type of descriptive
Descriptive studies consist of two major groups: those study. Surveillance can be thought of as watchfulness over
that deal with individuals and those that relate to a community. A more formal definition is "the ongoing
populations. Studies that involve individuals are the case systematic collection, analysis, and interpretation of
report, the case-series report, cross-sectional studies, and health data essential to the planning, implementation, and
surveillance, whereas ecological correlational studies evaluation of public health practice, closely integrated
examine populations. 4 with the timely dissemination of these data to those who
need to know. ms The key feature here is feedback, as in a
Case report servomechanism. Prevention and control of the problem
The case report is the least publishable unit in the medical are fundamental parts of the feedback loop.
literature. Often, an observant clinician reports an Surveillance can be either active or passive. Passive
unusual disease or association, which prompts further surveillance relies on data generally gathered through
investigations with more rigorous study designs (panel). traditional channels, such as death certificates. By
For example, a clinician, among others, reported benign contrast, active surveillance searches for cases. The
hepatocellular adenomas, a rare tumour, in women who reporting of abortion-related deaths provides an example.
had taken oral conrraceptives. 2s A large case-control study Since 1972, the US Centers for Disease Control and
pursued this lead and confinned a strong association Prevention has been doing active surveillance of these
between long-term use of high-dose pills and this rare, but deaths in the USA, using multiple overlapping sources-
sometimes deadly, tumour. 29 Not all case reports deal with ie, state maternal mortality study committees, professional
serious health threats, however; some simply enliven the organisations, newspapers, and colleagues in the specialty.
generally drab medicalliterature. 30-32 By comparison with official statistics, active surveillance
identifies about twice as many deaths.3~ Similarly,
Case-series report underreporting of maternal deaths remains an
A case-series aggregates individual cases in one report. international problem. 4 0--43
Sometimes, the appearance of several similar cases in a Epidemiological surveillance has made important
short period heralds an epidemic. For example, a cluster contributions to health, but none more Impressive
of homosexual men in Los Angeles with a similar clinical than smallpox eradication. Surveillance and containment
syndrome alerted the medical community to the AIDS were responsible for the elimination of smallpox
epidemic in North America. 33 Whereas a report of a single from the world, an extraordinary public-health
unusual case might not trigger further investigation, a achievement. 44 Whereas mass immunisation of the world's
case-series of several unusual cases (in excess of what population had failed, the approach of identification
might be expected) adds to the concern. A convenient of cases through surveillance and then immunisation
feature of case-series reports is that they can constitute the of susceptible persons in the surrounding communities
case group for a case-control study, which can then stopped transmission. Without a non-human vector,
explore hunches about causes of disease. the virus died out.

146 THE LANCET· Vol 359 • January 12, 2002 ·www.thelancet.com

 EPIDE~UOLOGY SERIES

Clues about cause

A third use of descriptive studies
is to develop hypotheses
abou t caus e (panel). Obs erva nt
clinicians note d an
association betw een high conc entr
atio ns of m•.-ygen in
incu bato rs and blin dnes s in babi
es; this finding led to
analytical studies, then a rand omi
sed cont rolle d trial,
confirming the association. 61 Une
xpec tedl y high rates of
canc er amo ng wom en who had
pain ted radi um dials in
watches aler ted investigators to
the dan ger of this
occupational exposure. 6!

Adv ant age s and disa dva nta ges

Descriptive stud ies have both stren
gths and weaknesses.
Ofte n, the data are already available
and thus ine~"'Pensive
Gynaecomastla, a condition asso and efficient to use. Furt hen nore
ciated with occupational , few ethical difficulties
exposure to oestrogen dust, feed exist. However, descripti\;e stud
ing of refugees, and Ingestion ies have imPOrtant
of taint ed food limitations. Tem pora l associati
ons betw een puta tive
causes and effects mig ht be unclear.
A dang erou s pitfall is
Ecological correlational studies that the investigators mig ht draw
causal inferences whe n
Cor rela tion al stud ies look for non e is possible.~'
asso ciati ons betw een
expo sure s and outc ome s in
pop ulat ions rath er Overstepping the data
than in individuals. 4 Bec ause
muc h data mig ht A com mon mistake in inference
alre ady hav e been cone cted , is post hoc ergo prop ter
corr elati onal stud ies hoc reasoning (after the thin g, ther
can be a con ven ient initial sear efore
The mea sure of asso ciat ion betw
ch for hyp othe ses. thing), an exam ple of a false caus M on acco unt of the
een expo sure and e. In othe r words, a
outc ome is the corr elat ion tem pora l association is incorrect1y
coefficient r, whi ch infe rred to be a causal
indi cate s how line ar is the rela one. In one egregious exam ple, seve
tion betw een expo sure n wom en in Pasa dena ,
and outc ome . For exam ple, deat California, crea ted cont rove rsy arou
h rate s from coro nary nd the wod d in the
arte ry disease corr elat e with late 1980s. Seen in one phy sicia n's
per capi ta sales of office, the wom en had
cigarettes. 45 By con tras t, access developed functional ovarian cyst
to safe legal abo rtio n s while taking the new
is inversely· corr elat ed with mul tiph asic oral cont race ptiv e
mat erna l mortality.46-49 piUs.~ Base d on this
The rang e of pote ntia l asso ciati unco ntro lIed observation, a case-seri
ons to be expl ored is es repo rt war ned that
near ly limitless. so phas ic pills mig ht pose a thre at
to pati ent heal th and
Cor rela tion al stud ies have imp orta safety. The med ia prin ted the
nt limi tatio ns-i e, stor y, and unk now n
the inab ility to link expo sure to num bers of wom en arou nd the wor
outc ome in individuals ld stop ped taking thei r
and to cont rol for con foun ding pills,os because they did not und
(a mix ing or blur ring erst and the difference
of effects). An exam ple of the latte between functional cysts and ovar
r is the obse rvat ion ian cancer. Sinc e the
that deat h rate s from coro nary repo rt had no com pari son gro
arte ry disease also up- eg, wom en usin g
corr elat e with the num ber of mon oph asic pills or thos e using non
colo ur television sets e-th e auth ors could
per capi ta. 4 Eve n telev ision 's hars not draw any conc lusio ns abo ut caus
hest criti cs are unlikely e of disease.
to argu e that it clogs coro nary In the wake of this repo rt, dam
vessels, an exam ple age- cont rol efforts
of "ecological faIIacy".37 Alth start ed quickly. \Vithin 2 years, a
oug h a link betw een pub licat ion showed no
television violence and violence temp oral aSSOCIation betw een
in scho ols seem s mor e the mar keti ng of
plau sible , whe ther this asso ciat mul tiph asic pills and the num ber
ion is inde ed caus al of wom en adm itted to
is difficult to esta blish . 51 hospital for trea tme nt of beni gn ovar
5 years elapsed before cohorr~
ian cysts.~ Howe\"er,
and case-control studies"'~
Use s of des crip tive stu die s confirmed no association betw een
mul tiph asic pillS and
Trend analysis ovarian cysts. By this time, the pubJ
ic-health dam age had
Descriptive studies have several usef been done.6"I
ul roles. Being able to Ano ther sad exam ple in which
mon itor the heal th of popUlations misi nter pret atio n of
is imp orta nt to heal th- descriptive stud ies hurt pub lic heal
care administrators. Tre nd analysis th is rout ine elec tron ic
is often prov ided by fetal mon itori ng in labour. A qua
ongo ing surveillance. Exa mpl es ner of a cent ury ago,
incl ude the emerging tem pora l associations betw een
epidemic of syphilis in the Rus sian the intro duct ion of
Fed erat ion, n.s) and the elec tron ic fetal mon itori ng and falli
inte rnat iona l epid emi c of mul tiple ng perinatal mor talit y
birth s, prem atur ity, rates led to the conc lusio n that cont
and low birthweight caus ed by inuo us fetal hea n rate
assisted repr odu ctiv e mon itori ng was a goo d thing.:~ Mor
technologies. 54-58 Bot h epidemics eove r, auth oriti es of
raise trou blin g societal the day pred icted a 50% redu ctio
issues. n in perinatal mor bidi ty
and mortali ty from its use.: o
Plan ning Bas ed on this rosy assessment
from prom inen t
A seco nd use is heal th-c are plan obst etric ians , this expensive and intru
ning . For exam ple, the sive technology took
intr odu ctio n of lapa rosc opy , coup obstetrics by stor m. How ever
led with bad pres s , the initial upb eat
abo ut oral cont race ptiv es and assessment did not survive scientifi
intra uter ine devices, c scrutiny. Years later,
tripl ed US rate s of tuba l steri a meta-analysis of the rand omi
lisation in the 1970 s. 59 sed controlIed trials
Hos pita ls and amb ulat ory surg ery showed that , by com pari son with
cent res had a surg e rout ine inte nnit tent
in dem and for oper atio ns, yet ausc ulta tion , rout ine elec tron ic fetal
less need for hosp ital mon itori ng confers
beds . Similarly, the intr odu ctio no lasting benefit to infants,
n of highly active whereas it significantly
anti retro vira l ther apy for pati ents increases operative deliveries; thus
with AID S decr ease d han ning women.: l
bed occupancy.60 Based on objective reviews, both
the Can adia n Tas k
Forc e on the Periodic Hea lth Exam
ination:~ and the US

THE lAN CET · Vol 359 • January

12, 2002 • w\\'\v.thelancet.com
1·17 .
J,
----------------------"-,-,.._---------- --~~-,
-
--------------
~~

----------------
EPID EMIO LOG Y SERI ES
--.===~~~
for AIDS amon g adolescents
expa nded SUlVeillance case definition
e73 have given rout ine W'k0-' Rep 1992; 41: 1-19.
Preventive Services Tas k Forc and adult s. MMW R Morb Morta l
a D reco mm enda tion (fair RE. Aver ette HE. Adenosis and
electronic fetal mon itori ng 18 Yon JLJr , Lutz MH, Girta nner
en: a review. Gynecol
nst its rout ine use) . Des pite this advice, abou t adeno carci noma of the vagin a in youn g wom
evidence agai
the USA incl ude electronic Onco11974; 2: 508- 17.
thre e-fo urth s of all birth s in the limitations of 19 Levine RJ, Math ew RM, Chen
ault CB, et al. Differences in the quali
ty
ng. 73
Fail ure to appr ecia te er and winte r. N Engl J Med
fetal monitori ng harm and of seme n in outdo or workers durin g summ
caus ed lasti
descriptive stud ies has 1990; 323: 12-16 . .
ars. es. New York: Anch or Books, 1998
squa nder ed billi ons of doll 20 McNeill WHo Plagues and peopl
itudin al study of denta l carie s prevalence and incid ence
21 Parko A. Long
olivine diabase (low fluoride) areas
in the rapakivi (high fluoride) and
Con clus ion , tentative appr oach to ofLa itila, Finla nd. Proc Finn Dent Soc 1990; 86: 103- 06.
Descriptive studies are ofte n the first 22 Diaz-Barriga F, Batre s L, Cald eron J, et al. The El Paso smelt er 20
se stud ies generally child ren. Environ Res 1997;
a new even t or cond ition . The years later: resid ual impa ct on Mexi can
dise ase or assess the heal th
emp hasi se features of a new 74: 11-16 .
geographic distri butio n of multi ple
mun ities . Hea lth adm inist rato rs use 23 Ebers GC, Sadovnick AD. The
statu s of com ds and plan for gy 1993; 12: 1-5.
itor tren sclerosis: a review. Neuroepidemwlo
descriptive stud ies to mon er M, et al. Prevalence of vitam in
D
By cont rast, epid emio logi sts and clinicians 24 Chap uy Me, Prezi osi P, Maam
lation . OstMIPoros 1m 1997; 7:
resources. ch for clues of insuf ficien cy in an adult nonn at popu
rts to sear
generally use descriptive repo es. In this role, 439-4 3. lity.
gene ratio n of hypo thes s of sun expo sure on cance r morta
caus e of dise ase- ie, 25 Ainsleigh HG. Beneficial effect
n a spri ngb oard into mor e 132-4 0.
descriptive stud ies are ofte Prro Med 1993 ; 22:
and
groups. Com mon pitfalls nan MA. Skin cance r, mela noma ,
26 Fears TR, Scott o J, Schn eiden
rigorous stud ies with com pari son nce of a clear, sunlight. AmJ Public Healt h 1976 ; 66: 461- 64.
ude an abse
of descriptive repo rts incl nitio n, and 27 Wingo P A, Higgins JE, Rubi n
GL, Zahn iser SC, eds. An
case defi ve healt h. Gene va: WHO , 1994.
specific, and repr oduc ible ies with out a epidemiologic appro ach to reproducti
the data . Stud relationship to oral
inte rpre tatio ns that overstep abou t caus e of 28 Sche nken JR. Hepa tocel lular adeno
ma:
lusions 559.
com pari son grou p do not allow conc contraceptives? JAMA . 1976; 236:
et al. Epidemiology ofhep atoce lIula
r
disease. 29 Rooks JB, Ory HW, Ishak KG, use. JA.M A 1979; 242:
the role of oral contr acept ive
aden oma:
ents
d L Sack ett for their helpful comm 644- 48.
We thank Willard Cates and Davi our unfor eseen hazar d of jogging. N Engl
J
Muc h of this material stems from 30 Waug h RJ. Penile frostbite, an
on an earlier version of this report. se.
datio n Facu lty Deve lopm ent Cour 178.
15 years of teaching the Berlex Foun Med 1977; 296:
J Med 1977; 297: 1127.
31 Levit F. Jogge r's nipples. N Engl
ride DQ, Lehm an LP, Mang iardi JR. Break -danc ing neck.
32 McB
References N Engl J Med 1985; 312: 186. l
a: Los Angeles. MMW R Morb Morta
Hulley SB, Cum ming s SR, Browner
WS, Grad y D, Hear st N, 33 Anon. Pneu mocy stis pneu moni
RE, eds. Desig ning clinic al research: an epidemiologic W'kry Rep 1981; 30: 250- 52.
New man
ncott Williams and Wilkins, 200 I. a RO, de Mora les AV. The
appro ach, 2nd edn. Baltimore: Lippi 34 Harri ngton JM, Stein GF, River
of form ulatin g oral contraceptives: a SUlVey of
SW. Clini cal resea rch in general medical occup ation al hazar ds
2 Fletc her RH, Fletc her Envir on Healt h 1978; 33: 12-15 .
. N Engl J Med 1979; 301: 180--83. plant empl oyees . Arch
journals: a 30-ye ar persp ective
ssen ME, Duga n JB, Farer LS. Epid emic of
gy, 2nd edn. New York: Oxford 35 Satti n RW, Roisin A, Kafri
3 LastJ M, ed. A dictionary of epide miolo
g mega l Haiti an enu·a nts. Public Health Rep 1984 ;
gyne coma stia amon
University Press , 1988. , 99: 504- 10.
miology in medicine. Boston: Little
4 Henn ekens CH, Buring JE. Epide 36 Fara GM, Del COIVO G, Bernu
zzi S, et al. Epide mic of breas t
Brow n and Com pany , 1987. Lancet 1979; 2: 295-9 7.
Foun datio ns of epidemiology, 2nd
edn. enlar geme nt in an Italia n school.
5 Lilienfeld AM, Lilienfeld DE. man KJ. Mod em epide miolo gy. Boston: Iitde , Brow n and
Univ ersity Press , 1980 . 37 Roth
New York: Oxfo rd
Gold berg RJ, et al. A popu lation -base
d Com pany , 1986.
6 Ande rson FA Jr, Whee ler HB, prehe nsive plan for epidemiologic
case- fatali ty rates of dcep vein 38 Cent ers for Disease Cont rol. Com
perspective of the hospi tal incid ence and
Atlan ta, GA: Cent ers for Disease Cont rol, 1986.
The Worc ester DVT Study. surve illanc e.
throm bosis and pulm onary embolism. JC, Roch at RW, Patte rson JE, Dolm an A.
933- 38. 39 Cates W Jr, Smit h statistics data for moni torin g
Arch Intern Med 1991; 151: Assessment of sUlVe iUanc e and vital
r-de Jong P. Review article: -1975 . AmJ Epide mioll 978;
7 Sasco AJ, Lowenfels AB, Paske abort ion mortality, Unite d State s, 1972
er-a meta-analysis of published case-
epidemiology of male breas t canc 108: 200- 06.
ted aetiological factors.
control studi es and discussion of selec 40 Kao S, Chen LM, Shi L, Wein
rich MC. Unde rrepo rting and
er 1993 ; 53: 538- 49. lity in Taiw an. Acta Obstct Gynecol
ImJ Canc
men. Epidemiol Rro 1993; 15: 220-
31. misclassification of mate rnal morta
8 Thom as DB. Breast cance r in Scand 1997; 76: 629- 36.
elma n D, Barbi eri RL, et al. Variation in the n J, Dekk er G, Van Dong en P,
9 MarshaIl LM, Spieg 41 Schu itema ker N, Van Roos male
g prem enop ausal wom en by age rrepo rting of mate rnal morta lity
incid ence of uterin e leiomyoma amon Van Geijn H, Grav enho rst JE. Unde
col19 97; 90: 967-7 3. 97; 90: 7&-82.
and race. Obstet Gyne E. in The Neth erlan ds. Obstet Gyne col19
a M, Mrozinski W, Wojciechowski s P, Hatto n F. Reasons for the
10 Jarem in B, Kotu lak E, Stam awsk 42 Bouv ier-C olle MH, Varn oux N, Coste
nsible for occupational hazard in lity in Franc e, as indic ated by a
Deat h at sea: certa in factors respo h unde rrepo rting of mate rnal morta
nnen. Int J Occup Med Environ Healt of childbearing age. 1m J Epidemiol
Polish seam en and deep-sea fishe survey of all death s amon g wom en
1997; 10: 405- 16. le 1991 ; 20: 717- 21.
WB, Desh mukh N. All-terrain vehic J, Blasini I, Adam sons K. Misre porti
ng
11 Kran e BD, Ricci MA, Sweeney a center. Am Surg 1988; 54: 43 Com as A, Nava rro A, Cond e Med P R 1990; 82:
a review at a rural level II traum lity in Puert o Rico. Bol Asoc
injuries: of mate rnal morta
471- 74. ern 343- 46.
ational injuries and deaths in north n in west and centr al Africa revisi
ted.
12 Gabe rt T, Stuel and DT. Recre ties from snow mobiles over 44 Foeg e WHo SmaJlpox eradi catio
of injuri es and fatali 1998 ; 76: 233- 35.
Wisconsin: analysis Bull World Health Organ
3 years. WIS Med J 1993; 92: 671-7
5. ing and geographic variation in
45 Fried man GD. Cigarette smok
of the Delphi panel for consensus the Unite d States. J Chronic DIS
13 Zinn J, Zalokowski A. The use coron ary heart disease morta lity in
atory perfo nnan ce.
devel opme nt on indicators of labor 1967; 20: 769- 79. 38.
Clin Lab Mana ge Rro 1999; 13:.3
86-408. is enough. Lancet 1995; 345: 137-
46 Anon . Abortion: one Rom ania F. The publi c healt h
ce MR, Sweet RL. Criteria nescu
14 Hage r WD, Esch enbac h DA, Spen 47 Steph enson P, Wagn er M, Bade a M, Serba
ng of salpin gitis. Obstet Gyneco11983; 61: abortion: lessons from Romania.
for diagn osis and gradi conse quen ces of restricted induc ed
h 1992 ; 82: 132& -31.
113- 14. Am J Public Healt
CA, Reynolds GH, Thom pson SE. e AL. The effect ofleg al abort ion
on the
15 Hadg u A, West rom L, Brooks 48 Seward PN, BalIard CA, Ulen
c inflam mato ry disease: a multivariate analysis., a large counr y hospi tal. AmJ Obstet Gynecol
Predi cting acute pelvi rate of septi c abort ion at
86; 155: 954-6 0.
Am J Obstet Gyne col19 1973; 115: 335- 38. ts
Casto n SA, Todd JK. Toxic shock apeu tic abort ion in California: effec
16 Wiesenthal AM, Ressm an M, 49 Stew art GK, Gold stein PJ. Ther
syndromes by strict and lity. Obstet Gyne coll9 71; 37:
syndr ome, I: clinical exclusion of other of septic abort ion and mate rnal morta
l985; 122: 847-5 6.
screening definitions. AmJ Epid emwl 510- 14.
on system for HIV infection and
17 Anon . 1993 revised classificati

~
ry 12, 2002 · www .thela ncet. com
THE LAN CET · Vol 359 • Janua
148
J
, ______-'"C-____________________________________________________________________________________________________________
EPIDE~UOLOGY SERIES _

50 Noller KL, Resseguie W, Voss V. The effect of"changes in 61 Silverman WA. Memories of the 1953-54 Oxygen Trial and its
atmospheric pressure on the occurrence afthe spontaneous onset of aftermath: the failure of success. COl1rrol Clill ToOls 1991; 12: 355-58.
labor in term pregnancies. Am J Obsrer Gy,1<!cc/1996; 174: 1192-99. 62 Stebbings JH, Lucas HF, Stehne)o AF. !\ionality from cancers of
51 Centenvall BS. Television and \'iolence: the scale afme problem and major sites in female radium dial workers.AmJ llld -'ltd 1984; 5:
where to go from here. JAMA 1992; 267: 3059-63. 435-59.
52 Borisenko KK, Tichonova U, Renton ML Syphilis and other sexually 63 Copi 1M, Cohen C. IntrOduction to logic, 10th 00n. Upper Saddle
transmitted infections in the Russian Federation. 1m] STDAIDS River, NJ: Prentice Hall, 1998.
1999; 10: 665-68. 64 Caillouette JC, Koehler AI- Phasic conrracepti\-e pills and functional
53 Tichonova L, Borisenko K, \"lard H, Meheus A, Gromyko A, ovarian CYSts. AmJ ObSUl 0'11«« 1987; 156: 1538-12.
Remon A. Epidemics of syphilis in the Russian Federation: trends, 65 Maninez F_ Responsibility of health prmiders and the media in
origins, and priorities for cancro!. LoT/eet 1997; 350: 21{}-13. response to scientific information_ Eur J Colllraupr R.'Proi H~-allh Car..'
54 Tough SC, Greene CA, Svenson LW, BelikJ. Effects of in virro 1997; 2: 25-30.
fenilization on low binh weight, preterm delivery, and multiple birth. 66 Grimes DA, Hughes JM. Use of multiphasic oral conrracepth-es and
J Pedialr2000; 136: 618-22. hospitalizations of women with functional o\-arian cySts in the United
55 Bider D, Livshitz A, Tur Kaspa I, Shulman A, Levron J, Dor J. States. Obstet 0'/Iecol1989; 73: 1037-39.
Incidence and perinatal outcome of multiple pregnancies after 67 Lanes SF, Binnann B, Walker AM, Singer S. Oral contracepti\'e
intracytoplasmic sperm injection compared to standard. in vitro type and functional ovarian cySts. Am J ObSTa 0'1/.:.:<>11992; 166:
fertilization. J Assist Reprod Genet 1999; 16: 221-26. 956--61.
56 Steegers~Theunissen RP, Zwenbroek WM, Huisjes AJ, Kanhai HH, 68 Holt VL, Daling JR, M.cKnight B, Moore D, Sterg2chis A, Weiss NS.
Bruinse HW, Merkus HM. Multiple binh prevalence in The Functional ovarian C);;ts in relation to the use of monophasic and
Netherlands: impact of maternal age and assisted reproductive triphasic oral conrracepti\·es. Obstel G)'1I~o,;.()11992; i9: 529-33.
techniques. J Reprad Med 1998; 43: 173-79. 69 Jones EF, Beniger JR, Westoff CF. Pill and leD discontinuation in
57 Dunn A, Macfarlane A. Recent trends in the incidence of multiple the United States, 1970--1975: the influence of the media. Fam Pfazm
binhs and associated monality in England and Wales. Arch Dis Child Perspecc 1980; 12: 293-300.
Fetal Noonatal Ed 1996; 75: FIO-I9. 70 Quilligan El, Paul RH. Fetal monitoring: is it worth it? Obsut G;7lcyd
58 Anon. Pregnancies and binhs resulting from in vitro fertilization: 1975; 45: 96--100.
French national registry, analysiS of data 1986 to 1990. FIVNAT 71 Thacker SB, Stroup OF, Peterson HB. EfficaC}' and safety of
(French In Vitro National). Fertil Steri11995; 64: 746--56. inrrapanum electronic fetal monitoring: an update. Olnra 0'1uu;l
59 Peterson HB, GreenspanJR, DeStefano F, Ory HW, Layde PM. The 1995; 86: 613-20.
impact of laparoscopy on tubal sterilization in United States hospitals, 72 Canadian Task Force on the Periodic Health Examination. The
1970 and 1975 to 1978. AmJ Obster GY1/eco[19SI; 140: SII-14. Canadian guide to clinical pre\'entive care. OttO\\-a: .\linister of Supply
60 Anon. Update: trends in AIDS incidence, deaths, and prevalence- and Services Canada, 1994.
United States, 1996. MMW2? Morb Monal Wkry Rep 1997; 46: 73 US Preventive Services Task Force. Guide to clinical pre\'emh-e
165-73. services, 2nd edn. Baltimore, ,\ill: Williams and Wilkins, 1996.

THE lANCET· Vol 359 • January 12, 2002 • www.thelancet.com 149 (