How To Do It Organise a multicentre trial CharlesWarlow Multicentre and single centre clinical trials share methodological problems which are well known (box). Multicentre trials do, however, have several distinct advantages.They provide: (o) larger sample size so that the result is more precise, appropriate subgroup analysis is more feasible, and there is a lower risk of an apparentIy "negative" result when the treatrnent is, in truth, effective; (b) quicker results before peoplelose scientÍfic and cornmercialinterest in the treatrnentand before it is modjfied or the theoretical indications for it are changed; (c) wider disseminationof the results and, possibly, more widesrread helief in their \'alidit\': (d) standardiseddefinitions of diseaseand measurements of outcomeamongcentresand amongcountries in intemationaltrials; (e)a wider rangeof clinical and methodologicalskiIIs to solveprotocol problems; (f) usually a wider range of patients,facilitating the generalisationof results, which can be broadlyapplied to futUrepatients in othercentresand other countries; (g)large negativetrials which are more likely to be publishedthan small negativetrials. This is important otherwisesmaIIpositive trials, which aremore likely to besubmitted for publicationand probablymore likely to be acceptedrather iban small negative trials, will tend to dominatethe scientÍficliterature; (h) less suspicion and rivalry among centres and countries without necessarily suppressing healthy competition; (1)lessscientÍficisolation; (J)better nationaland intemational collaboration; (k) facilitation of further multicentre trials of potentially important treatrnents,provided that the initial trial is not too demanding. The difficulties of multicentre trials comparedwith singlecentre trials mainly concemthe coordinationof many people in severalcentres and even countries (see box on next page). There are usually severa! possible solutions, the best depending on circumstances,geography,number of centres,budget,and so on. What follows is not meantto be anossifiedblueprint but somesuggestions.Suggestions,moreover, which have not beentestedin randomisedtrials but which, at least,arebasedon someexperience. Department of Clinical Neurosciences, Westem General Hospital, Edinbürgh EH4 2XU Charles Warlow, MD, professOT of medical neurolog:,' H,lf,JJI'I'i(IJ()() 180 11'(13 Methodological issues common to both single and muIticentre iríais .Randomised or non-randomised trealment comparison .Blind or open treatrnent allocalion and outcome assessment .lntenlion .Sample to treal or on-treatment analysis, or both size and duration of follow up .Subgroup anaIysis .Defined entry and exclusion crileria .Amount of data lO collect on randomised and non-randomised but eligible patients .D"fin"J UUl,ul11"" .Widespread applicability of lrial results; what to do about non-randomised but eligible patients .lnterim anaIysis and the Tole of an independent data monitoring cornmittee .How intensively to look for adverse effects of treatrnent .How intensively to monitor compliance .Computerisalion of data .Sharing of original data with overview groups looking at similar treatrnents (meta-analysis) .lnvolvement with sponsoring company, if any. problems, visit centres,and ser up and supervise the tría! office. This personwilI usua!Iy,but not always,be a doctor, presumablywhoever wanted to ser up the tria! in the first place,who is going to be in the samejob long enough to see the trial through to the end. Medica!Iyqualified principal investigators must t~. to enter their own patients into the trial to maintain credibility and to experienceand share the practical problems with the other coIlaborators. Writing tbe protocol This is a job for one person, usually the principal investigator, and not a cornmittee. Of course, that person will need comments and advice from all the centres,tbe trial statistician,and outside expens in the field, andthis mayall require one or more collaborators' meetings and often several versions of the protocol before everyoneis satisfied. Becausethere is no rule againstmodifying a protocol if unforeseenproblems ariseduring the trial (provided suchmodifications are Leadership A multicentre trial must havean identifiable leader sensibleand not data dependent)there is no reason or principal investigator, who should be able to to delay writing it and getting staned while some cornmand the respect of collaborators and have centresfinally make up their minds whether to ;oin in. the time and energy to keep the whole enterprise Indeed, it mar be desirable to involve more centres under control. He or she must not lose interest in once the trial is underway and some of the teething the dull bit betweengetting staned and getting the problems soned out, although the more centres that results. Running a trial by cornmitteeis disastrous.A areconcernedfrom the very stan the better. It doesnot prestigiouschairpersonof asteeringcornmitteemay be always hold, however, that centres concerned in a "political" advantageor evena necessity,but such a writing the protocol are more likely to súck to it than personis unlikely to ron the trial on a dayto daybasis. centres ioining later and having to accept ir. In Someone mus¡ get the funding, gather together international trials the protocol mar have to be centres, run meetings, answer questions, solve translated into several languages, which can be 8.\\J VOLl'ME300 20 JA~1:ARY 1990 surprisinglvexpensive. Fonunately, if it is \\Titten in English translation is now hardly necessary, at ¡eastnot in Western Europe, which is a huge advantage to us in Great Britain and a great credit to OUT medical colleagues on the continent. Finding the centres Having accepted that a multicentre rather than a single centre trial is scientifically necessaryto solve a particular treatment problem, friends and colleagues from a fe\\"centres usually get together to discuss a protocoloFrom there, other friends and colleaguesare brought in from centre to centre and from country to country unúl enoughcentres are found to saúsfythe sample size requirements in a reasonableperiod of time. After about three years recrwtment gets tedious and may fall off, but follow up is usually less Úffie consuming and may need to continue much longer, depending on the treatment. Key people in a country are often very successful in recruiúng their own compatriots,far more so than an outsider. Advertising in medicaljournalsand through specialistorganisations canalsohelp. Pharmaceutical companies may, through their naúonal and internaúonal networks, approach numerous potential collaborators simultaneously. Howeverthe centresare found, they must be seriously interested and reasonablycompetent in the field but not necessarilyspecialists.In any eventspecialistsmay already be involved in their own compeúng studies; relaúve non-specialistsin district generalhospitals or generalpracúcemay be extremely keento collaborate -and make very effecúve collaborators. Usually they have more patients than teaching hospitals; they may haveno other wayto take pan in medical research,and to be involved with specialists in the field who are organising the trial is often educative. But whoever the collaboratorsare, they are all equal when it comes to recruiúng patients, and it is theír trial not the principal invesúgators' whose role is organisaúonal and catal~"tic:there should beno "star billing." Randomisation Randomísation must be centralised and is best done by telephone or perhaps by a coroputer link to some central point which may be the triaJ office or, if 24 hour cover is required, a hospital ward or switch board. This is the only way for the triaJ organisers to have an irnmediate binding record of who has been randornised, when. and froro which centre. Sealed envelopes or locaJly held randoroisation lists are simply not good enough. Issues that may presentparticular difficulties in multicentre trials .Leadership .Wriúng tbe protocol .Finding the centres .Randomisation .Patient registration .FoUow up .Trial coordinator .Trial office .Visiting tbe centres .CoUaborators' meetings .Quality control¡ .Costl. Writing th.: papers. BMJ VOLUME 300 Superfluous data must not be collccted and nor should "add on" studies be permitted without adequate human and financial resources. Indeed, all the collaborators must have a very clear idea of the aim oi the trial and concentrate on that. Nevertheless. it may be possible for a few particularly interested collaborators to collect more data than the others, but this should be optional and certainly not to the detriment of paúent recruitment or answering the basic trial quesúon. FoUow up Follow up must :lIso be simple and require liule more Iban what would be done in routine clinical practice. ObviousJy, the outcome must be recorded and measuredbut tbis may not require leng¡hy follow up, and sometirnesit can be obtained from routinely collected data (such as death certificates, cancer registries)thus disposing with the need for follow up at all. Some follow up might be more simply, and even more accurately, obtained by telephoning the patient rather Iban by contacting the patient's doctor. Overelaboraterecording of and testing for compliance should be avoided as should unnecessaryblood tests. Too frequent and too detailed folIo\\" up may add very litde to the statistical power and may kill recruitment in a multicentre trial once trial patients and collaboratorsrealisewhat they have 1ft themselvesin for. Wriuen forms should be no longer Iban one side of A4 paper; if they are they probably will not be completed fully or reliably. In any event trials should not interfere with routine clinical practice. Trial coordinator This is the key person even if the tria! is not large enough to merit a full ÚIne appointment. The job is administrative, not medical. She, for it is seldom he, stands at the centre of the tríal and must be commined, energetic, sensible, well organised, have some knowledge of computing, and be able to work flexible hours. T o collect the data, organise them, and transmit them to the statistician she must be meticulous and even obsessiona!, but she must algo be good with people so that she can ron the trial office and maintain harmonious relationships with the distant collaborators. She must be adroit at dealing with the awkward (and some doctors can be remarkably awkward) and flexible and energetic enough so rhat she can hop on and off trains and planes and pUl up with a certain amount of discomfort as she travels from centre to centre extracting data from iorgetful coliaborarors, encouraging them ro randomise more patients, and generally nurruríng esprir-de-corps, while still retaining a sense of humour. Patient registration If large numbers are required and the budget is Trialoffice limited patient registration must be simple, practical, The trial office,organisedand supervisedby the trial and quick. Collaborators should have to do little more coordinator, has the task of collecting, checking, than what is normally required in routine c\inical and entering the data from the distant centres and practice. Patients must be identified (name, sex, date requestingmore informaríanif (hereareinconsistencies of birth), data on a few important prognostic variables or omissions.The office must supply (hecollaborators collected, and possible prespecified subgroups with all the necessarydocumentarían(entry and follow identified. It is often feasible to collect a11or some of these data on the telephone before treatment a11ocation up forms, freepost envelopes,sticky labels, etc), and possibly even (he trial medication, dispatch regular is made so obviating the need to complete, post, check, newsletters and listings on missing data, listings of code, and punch data entry forms. It also means when patients are due for their next follo~' up, and that me treatment a11ocationis not made unless and answerquestions from (he many centres. The office until the entrv data are recorded centrallv. Another must be available, friendly, helpful, knowledgeable, advantage is that data recorded bejore randomisation reliable, and efficient, perhaps in more (han one are unbiased with respect to any knowledge of languagefor internacionaltrials; al(houghdoctorsmay treatment a11ocation.Computer networks which do the speakexcellentEnglish, the samedoesno! necessarily samething are another possibility but more expensive. 20 jANUARY 1990 181 apply to their hospital telephonists. The collaborators must seethe office as a "black box" into which data go in and out of which informaúon comes. Any strife, inefficiency, or chaos should remain hidden. Such an office needs to be clase to the principal investigator and, if possible, to the staúsúcian, although with the avaiJability of modero high speed computer links this mar not be so imponant now. The office requires secretarial and clerical staff, a telephone and answering machine independent of a busy hospital or university switch board, a facsirnile machine (Fax), a photocopier, several rnicrocomputers,printers, filing cabinets, and adequate storage space. It must have enough resources to do its job properly and take as much as possible of the burden of the trial organisaúon from the collaborators. The resources required will gradually increase through the recruitment period after which they will stabilise but probably not decline. Visiting the centres This is time consuming and expensive but must be done, often regularly, for the following reasons: to discuss any local problems arising from understanding and implementing the protocol; encourage recruitment to the trial and so]ve problems relating ro rhis; collect outstanding data, although wíth advance warning these usually appear in the trial office shortly before the visit; discover neighbouring centres interested in joining the tríal; reiriforce the role of each centre, without which the tríal would faíl; and keep everyone informed of what is goíng on in the trial as a whole. The meeting must be arranged well in advance because it is importanr to ger all the local collaborators around a rabie. But the vísít need not last long, perhaps an hour or so. Remember, the local collaborators are busy and the trial is unlikely to be their majar preoccupation or priority. Who does the visits depends on the circumstances; the trial coordinaror is usually the best person to collect data and the principal investigator best at sorting out scientific and medical problems, but a regional coordinator can be helpful if the centres are widely dispersed. Whoever ít ís must not waste time; often several centres can be visited in a day and work can be done on trains. Trains with sleeping compartments perrnit an early start to the day. A voíd visíts during the surnmer when collaborators are on holiday and the trains are full of additional passengers, and avoid the depths of winter when snow and fog dislocate transporto Don't carry too much wíth you but take enough to read in case you gel delayed; take train and 'plane timetables and be prepared to change your plans; sort out how to get money quickly in foreign countries; and make sure the trial office and your farnily know your whereabouts. etc. Funhermore, it may be scientificalIy attractive (and indeed cornmercialIy attractive if registration fees and sponsorshipare fonhcoming) to organise a symposiumor educacionalmeeting alongside the trial meeting. Quality control In a single centre trial it is imponant that the trial treatment is properly described, well delivered, and any complications fairly assessed. This is even more imponant in multicentre trials, when there are cenain to be differences among centres, both real and due to chanceo Even drug treatment may differ among centres because of different storage conditions or possibly circumstances of delivering it to the patients. But of more concem is variation among centres in other non-drug treatments being evaluated such as surgef)", speech therapy, psychotherapy, etc. Obviously all the centres must agree to standardise more or less such treatments so that about the same amount is given for about the same time and that it is reasonably uniformo Such non-drug treatments, however, will never be exactly the same in all centres and it is counterproductive to insist that they are; as long as they are roughly similar no problem.; \vill ari.;e and the trial result will be applicable to other centres giving roughly, but never exactly, the same treatment. Monitoring uniformity of treatment is difficult but at least it helps to monitor any irnmediate complications, such as postoperative morbidity, and to have regular meetings of the collaborators. If a centre is perfonning badly (inadequate treatment, excessive complications, not enough patients to ensure competence, etc) then it should stop randornising patients, but not, of course, stop following up those already randornised. Clinically sensible cointerventions, such as drug regimens adrninistered during a trial of a surgical procedure, do not have to be exactly the same in each centre because randornisation (stratified by centre if necessaf)') will ensure that they are used in the same proponion of "treated" and "control" patients across the trial participants. Cost Mulúcentre trials are expensive but they need not be prorubiúvelyso provided that collaborators are reasonableand do not attempt to re-equip their entire departmentand undertake major non-trial projects at the expenseof a sponsoringpharmaceuúcalcompany. lndeed, from the point of view of unit cost per paúent randomised or per patient year of follow up multicentre trials should be cheaperthan single centre trials as they gajo from economies of scale. But, of course, the trial office and collaborators' meetings must be properly funded and the collaborators Collaborators' meetings themselvesreimbursed for any extra work (wruch is Collaborators' meetings are needed to write, very linle in a well designed trial) over and above and if necessary,rewrite the protocol; to discuss rouúne clinical practice. There are various ways of diseasedefinÍtions, measurememsof outcome, and doing trus: a lump sum per patient randomised; a sum documemation; to discuss any imerim results; per completedata form received by the trial office;-a to encouragetreatmem; to plan new trials; and to formula basedon the number of patients randomised encourage a group identity and common purpose. to support a researchnurse, etc. Whatever is done the They are, however, time consuming to attend, collaboratorsshould be paid only for extra work done extremelytime consumingto organise.and expensive. and not work that they saythey will do. In many rrials. however, it is difficult to obtain proper funding; trus They should therefore be frequem enough to fulfil their purposebut not so frequem asto be prohibitively applies particularly to treatments from which no profits are to be made-for example, non-patented expensive. It mar be sensible-or evenessemial-to organiseregionalor national collaborators'meetingsif drugs, surger:-',physiotherapy-so that it is totall: thereare alarge number of widelydispersedcentres. It inappropriate for trial funding to be left entirel~'to the mar alsobe possibleto link trial meetingsto othersthat pharmaceutical industry. Although a sponsoring collaboratorsare likely to be going to an~"Way. suchas compan~'must nor be concernedin either data analysis national specialist societies. imernational meetings. or publicarion. it canbe extremely helpful in collecting BMj VOLl"."iE 300 20 JA~LARY 1990 baseline (but 7101outcome) data, maintaining trial discipline, and encouraging recruitment through its own networks of medical representatives and researchers. Whatever the cost of a trial, it should be compared \\.ith the casI of the disease being treated and, perhaps. against the casI of non-medical endeavours such as 10\\. altitude military aircraft training or unemployment benefit. By any such comparisons multicentre trials are usuallv extremelv inexpensive and may lead to the rejection ~f expensiv~ but inetTective treatments-for example, extracranial to intracranial bypass surgef). for the prevention of stroke-and not al\\.ays to the introduction of more expensive health careo Writing the papers Like the protocol, tbis is a job for the principal invesugator, not a committee. Naturally, it will be necessaryto have many discussions with the tríal statistician and trial coordinator and commentsand advicefrom all the collaboratorsasnumerousdrafts are produced. It is crucial, however, that in the end aJIthe results are published under the name of all the collaborators;without them there would have beenno trial at all, and they did the work. Although current1y unfashionablein somequarters, the whole philosophy underlying multicentre trials is that group effort takes precedenceover individual effort; only by acting as a group can the individuals get answersto therapeutic questions which affect their own individual pauents. Of course,any centrecan publish its own results but there must be no "star billing" for author:; ",'hen the results ofthe whole trial are presented. Conclusions Betore starting a multicentre trial the foIlowing questions must be answered affirmatively: Is the therapeutic question really important, preferablyeven a burning issue?Are you sure there is no better way of answering ir? Can you gel enough centres together? Are you likely to gel the resources?Have you gol the time?Do you reaIly want to do ir? If so, thengo ahead, but first visit one or two successfulmulticentre trial organisationswhich will give you far more idea of the problemsand pleasuresiban I havebeenableto within the contextofthis article. Remember,only fools faíl to learo from others' mistakes.And once you gel started, always keep thinking about how the tríal can be done more efficiently and effectively, lessexpensively,more quickIy with a greater recruitment cateand with less extra work being done by the coIlaborators. At the same time remember that esprit-de-corps is what countsmore thananything else: look afterir. 1 thank other devoteesof multicentre trials with whom 1 haveworked and who commentedon the manuscript (Livia Candelise,David Chadwick, Rory Collins, BarbaraFarrell, Jan van Gijn, Adrian Grant, Peter Sandercock,and Brenda Smith), and also Richard Peto, who persuadedme and other c1inicians,that really large sample sizesare crucial to make any senseof treatrnents which have modest yet clinically imponant benefits. But it is 1 who take the blame for alI the opinionsexpressedin the anicle. Fibromuscular dysplasia of renal arteries: a neglected cause of acote loin pain Simon Stinchcombe, Adrian R Manhire, Michael C Bishop, Roger H S Gregson i i An arteriallesioD sbould i be suspected in patients i witb acute loÍD pain if : there is no evidence of ureteric ot-struc!;o" O" urograpby Department of Radiology, Queen's Medical Centre, NottinghamNG72UH Simon J Stinchcombe,BM, senior registrar Roger H S Gregson,FRCR, consultant Departmeots of Radiology aod Urology, City Hospital, Nottingham NGS lPB Adríao R Manhire, FRCR, consuilanl MichaelC Bíshop, FRCS, consullanl Corres¡:x>ndence10: Dr Slinchcombe. Sr ,\led] 1~,300 BM} ¡1!3- VOLUME300 Fibromuscular dysplasia of the renal arteries is usually associated with hypertension, but occasionally it may presem with acute loin pain due to occlusion of the renal arteries or embolisation of small peripheral vessels. The preseming symptoms may then be inuIslmguIsnaDie lrum a":Ull: ..:a!..:uiuusubstru..:uun ur acute pyelonephriús. Early diagnosis of arterial occlusion is essemial to prevem permanem damage to the urine gavenegativeresults. On intravenousurography the right kidney was thought to be non-functioning. On ultrasonographythe size of the kidney was nonnal with no dilatation of the pelvicalicealsystem.Ureteric obstruction due to a radiolucent calculus was diagnu~eJ,anJ he \\.a~lrealeJ cun~er\"al1\"e¡y. The next day he developed a temperatureof 38°C, and despite 24 hour treatment with intravenous antibiotics he did not improve. A right retrograde pyelokidney. We report on three young previously fit patients who gram was nonnal, but a -Tc diethylenetriaminepresemed with acute loin pain, in whom renal infarcúon penta-acetate(-Tc DTPA) renogram showed that occurred in relation to fibromuscular dysplasia of the there was no perfusion of the right kidney. Renal renal arteries. We also discuss how paúems with loin angiographyshowed that the right renal artery was pain should be investigated so that arterial occlusion completelyoccludedby thrombus 3 cm from its origin; can be diagnosed early. there \\'assomeirregularity of the wall of the left main renal artery, and two stenoseswere seenin one of the segmentalvessels.The findings \\.ereconsistentwith a Case1 diagnosisof fibromuscular dysplasia.Despite the five A 33 year old man was admitted with a four hour day interval after the onset of symptomswe anempted bistory of acute right testicularpaÍn. Testicular torsion to revascularisethe kidney with intra-arterial strepwas diagnosedinitially, but an exploratory operation tokinase 5000units/hour. The thrombus in the right disclosedno abnormality. Two daysafter admissionhe main renal artery was dissolved, disclosing a 75% wasstill in pain, wbich had movedto the right loin, and stricture, which wastreated \\"ith balloonangíoplasty. he developed frank haematuria. lnvestigations dis- Unfortunately pericathéter thrombus fonned, which closedaraisedwhitecell count (20.7x 10./1)but normal embolised distally, and peripheral perfusion to the plasmaureaandelectolyteconcentrations.¡\1.icroscopic kidnev could not be restored. The patient was given examinationof the urine showedincreasednumbersof long íenn anticoagulant treatment and discharged. red and wbite cells and granular casts.Culture of bis Detailed investigation subsequentlyfailed to disclose 20 jANUARY 1990 183