Health services based trials (HSB): introductory course Elina Hemminki THL, Universities of Helsinki and Tampere (National Institute for Health and Welfare) Helsinki May 11,12,18 2009 Department of Public Health, Helsinki Structure of the course Monday 11, Elina Hemminki • Introduction of participants and interests • Introductory lecture • Assignment of group/ individual work • Start of group/ individual work Tuesday 12, Diana Elbourne, Nea Malila, Minna Kaila Monday 18, student presentations • presentations + discussion • what next Volume and credits Guided work 12 hours Independent work 10 hours • reading and commenting an article • project plan (shared work) • course diary (a brief description what you learnt/ wondered of the lectures or the topic) Credits (to be clarified): if all of the above: ..credits. If 60-90%: ..credits Outline of the introductory lectur • Trials: Experimental vs non-experimental • Various kinds of trials • HSB: why, features • Examples Material: • 3 papers for exercise 1 • Others which may be of interest • List of further reading (later) • Copies of presentations Trials: Experimental vs nonexperimental 1. Introduction 2. Doing a trial 3. Difficulties 4. Additional issues 5. Impact of trials 6. Exercises What does a trial mean? • Intervention to group(s), created by the researcher (><natural trials) • Usually, two or more (>< uncontrolled trial) • Good trials: comparable groups (same calendar time) • Most common allocation method: randomization (why: also unknown predictors) • Different kinds of trials History • First modern clinical trial 1940s (TB and streptomycin, MRC in UK) • blood letting, scurvy and vitamin C • later non-clinical health trials • agriculture etc earlier • www.jameslindlibrary.ccccc Why trials? • Unbiased information • Especially important in case of therapy and other intentional activities (selection bias) • “All new interventions in health care (and those old not previously studies) should be introduced via trials” • Simple in theory, difficult in practice Advantages • With an example • postmenopausal hormone therapy • used for decades based on nonexperimental data • millions of women exposed for a preventive therapy; health (and monetary) costs large HT • (Female) hormone drug therapy during and after menopause, HRT • Estrogens (+ progestins + other), extracts, DES, other synthetic • 80 yr + for symptoms • 40 yr + against aging • 30 yr + for preventing diseases (all major chronic diseases) • North America (USA) --> Western & Northern Europe --> whole world DDD/1000 inhab. 60 50 Finland 40 30 20 10 0 1980 1984 1988 1992 Year 1996 2000 The Nurses’ Health Study 121,700 females age 30-55 in 1976-1994, risk of cardiovascular deaths . OR (95% Cl) Crude Adjusted Never Current HRT 1.0 0.35 (0.25-0.49) 1.0 0.47 (0.32-0.69) NEJM 1997;336:1769-75 Past HRT use 0.84(0.67-1.05) 0.99 (0.75-1.30) WHI, Women’s health initiative • 16 000 ”healthy” women; RCT, 8 yr prevention of cardiovascular diseases • stopped after 5 yr: ineffective, harms • 7 extra heart infarcts per year per 10 000 women • 8 extra breast cancers • 8 extra brain insults • 18 extra deep vain thrombosis WHI, Women’s health initiative • less fractures • less colon cancer • (no major effect on metabolic diseases, dementia, well-being) Other trials • Nachtigall et al. 1979: + • Hemminki et al. 1997: ? • HERS 1998: • Interim report from WHI 2000: 0 • (WHI 2002: -) • WH1-2 2004: - Small secondary prevention trials EVTET 2000: WEST 2001: 0 PAPWORTH 2002 ESPRIT 2002 etc. (Michels ym 2003) Why a difference in observational and experimental studies? • several papers trying to solve what was the cause. • an explantion (of general relevance: bias9 Bias in observational studies of Oestrogen and Heart Disease Source Effect on OR Prescribing bias More educated Higher social class Osteoporosis No disease* Healthier Compliance decrease decrease decrease decrease decrease decrease decrease * diabetes, hypertension, cardiovascular After Barrett-Connor, BMJ 1998; 317:457-461 CDP- RCT; secondary prevention of CHD, clofibrate,Males 30-60 (n=2789) 5 yr Mortality, Placebo Group Adjusted for 40 baseline characteristics 67% “GOOD” ADHERERS* 15 % 16 % 33% “POOR” AHDERERS 28 % 26 % Adjusted OR death =0.64 attributable to compliance * > 80 % prescribed dose NEJM 1980; 303: 1038-1041 Doing a trial Doing a trial: Simple in theory, difficult in practice • Dependent on others (service providers, customers, politicians, press etc) /laboratory, archives, interviews • On line research, things are changing while research is going on, a moving train • Ideological opposition, “guinea pigs “ • Costs, bureaucracy • Researcher skills/ administrative skills General principles • • • • • • Forward in time, like a cohort study Define the intervention Create the groups Expose them to intervention(s)/ control Define the outcomes, measure them, Compare the groups Forward in time • like a cohort study • starting from exposure (intervention) • parallel in time; cross-over-design ______________ • uncontrolled trials: “comparison” group may be the situation before or after; I do not recommend Define the intervention • • • • • • • Important to be explicit how feasible? Original and compromised how applicable after the trial? together with those who implement think all elements, including the soft ones describe the intervention in detail monitor whether the intervention is as planned; report changes; correct if possible Create the groups • Target population (different for mechanistic and practical trials) • Allocation: randomization if possible (takes care also unknown confounders) • Unit: an individual, a department, a village • Power calculations: how many ---> impact on all aspects of the study • Power calculations different for individual and cluster trials • What is possible? Expose to intervention(s)/ control • Craftsman skills: weak link often • Method and skills needed: depends on the intervention and context • Contamination (control group) Define outcomes, measure them • Mechanistic vs. practical: to understand what happens/ is the intervention useful • Often: too much intermediary, laboratory, process etc. outcomes • Too little health, unanticipated outcomes • Costs of intervention (important for decision) • Costs of data collection • An issues: How much determined by data source Compare the groups • Simple if groups similar (randomized and large numbers): “cross-tabulations” • Analysis by intention to treat (compliance, drop-outs) • Main outcomes such that data from everyone possible to get • Sub-analysis by compliance etc • Overall outcome from all dimensions (for decision making) Typical challenges or problems • providers/ others responsible • balance between feasibility and generalizibility • changes in environment, time (too early/late) • compliance, dilutation • drop-outs, loss to follow-up • costs and dependence on financiers Persuasion of service providers/ others responsible • Skills not often trained, nor typical for researchers • Conflicting interests; knowing the politics and march order • Routines help, others’ experiences • Unconventional approaches or challenging current practices: ? • Shopping around (Researcher X moved to town Y because it had good chief physician) Balance between feasibility and generalizibility • Generalizibility: can the results of the trial be transferred to practice/ other contexts? • Often a compromise what would be best and what can be done • Stepwise • Using other type of evidence to fill the gap between own trial and the practice Changes in environment, time • “It is too early to study an intervention until it is too late” • “The results of trials tell about history” • Environment may change (political, administrative, health, services, public perception, funding…, CHIMACA) • Competing groups may publish their results (WHI and Wisdom) • A problem especially with long-term exposures/ follow-ups Compliance, dilutation • Trial participants do not do as planned The groups start resembling each other. Control group wants; intervention does not want. • Especially important for trials when - intervention based on behavior changes - long term intervention - hot topic Drop-outs, loss to follow-up • How serious threat: depends on the context and the intervention; long follow-up • Anticipation: is there a way to get at least some information from all • Compromises in terms of other aspects (e.g. comprehensive information) • Routine record keeping / special data collection • Drop from intervention/ from data collection Costs and dependence on financiers • Cost-efficiency of the trial • Costs of the study an aspect in all methodological decisions • Bias in topics by the interest of potential financiers • Service providers should be responsible: trials a part of routine services (see Chalmers papers) Practical issues • Permissions (ethics, data protection etc) • Insurance, liability • Intervention = drug: international detailed rules • Data collection: simple/ many aspects, administrative data (register based trials) • Financing/ time (which first: field or money?) Difficulties • Depends on the intervention and the design • Mechanistic vs. practical vs HSB • Imagination in the design may help • Teaching of clinicians/ others responsible for activities studied • Public relations, press • Will be easier in the future? • EU and drug trials What have I learnt? • Have several research questions (vs Peto) • Visit field frequently; make space for finding problems (as it is vs. it should be) • Be ambitious, but do not expect success • Have imagination • Look for interest groups’ motives • Need for PR and alliances • Find practical tips from previous research Impact of trials • • • • Valued in medical literature Cochrane collaboration Other meta-analyses Trial registration, trial reporting rules all trials, regardless of results should be published • “Negative” vs. “positive” results Impact on practice • ? Like any research: • “if winds are favorable “ • if key-people are involved • if (financial) interests/ incentives Optimistic: with time the impact will come _______ Example of HT trial HT sales in USA after publishing WHI-I-trial 2002 • 40 % reduction in sales of all estrogen-progestin pills July - October (prescriptions) • 56 % in sales of Prempro® • competing drugs: a small increase Source: The Associated Press, November 2002 (web) HT myynti Suomessa 1980-2004 60 50 40 30 20 10 0 19 80 19 82 19 84 19 86 19 88 19 90 19 92 19 94 19 96 19 98 20 00 20 02 20 04 DDD/1000 inhabitants HT use in Estonia and in Finland Comments of WHI-I trial in Finnish medical journals Not so enthusiastic • Important • One of many • Not for preventing CVD 7 / 13 4 5 Possible reasons for Finnish physicians’ defensive position • fear of loosing personal credibility (experts) • fear of loosing professional credibility (among patients and lay-persons) • too much at stake in clinical work • too much at stake research wise • education and push from competing firms Pragmatic vs explanatory trials, mechanistic vs practical • Originally by Schwartz and Lellouch 1960s • Criteria changed but basic idea the same: two types of trials at the end of the continuum • Impact on all phases • Simplified: practical brings answers for practice; main problem in generalization Outline for HSB-trials • • • • What HSB is and why it is important? Do HSB trials need promotative actions? Is Finland the promised land of HSB? Examples What does "HSB trial" mean? • The purpose is to find, through experimental research, the best form of action in health services* • It aims for practice * in a broad sense, including activities outside health services aiming at promoting and preserving health, treating diseases, rehabilitation and (medical) nursing Terms in Finnish • kokeelliset tutkimukset terveydenhuollossa (KTT) • terveydenhuoltoon upotetut kokeet • terveydenhuollon interventiot • kokeelliset tutkimukset terveydenhuollon käytännössä • kokeelliset tutkimukset osana terveydenhuoltoa • jne. Terms in English • • • • • • • health services based trials (HSB) complex interventions (Elbourne) community trials interventions in health care social experiments cluster (randomized) controlled trials randomized field tests + quasi-experiments (methods other than randomization to allocate groups) • randomized social experiments • field experiments • trials for policy research Typical features 1 • Experimental research = health/ health care intervention is given/ received in such a way that different groups can be reliably compared. The researcher decides grouping. • HSB-trial definition is overlapping with other experimental trials and evaluation research. • Classifications are continuums. Typical features 2 • Explanatory/ mechanistic trials: individual randomization, informed consent, researchers responsible for intervention, new exposures • HSB-trials: cluster randomization/ other allocation, no informed consent, "system" responsible for intervention, exposures already in practice • Natural experiment: allocation independently from research, no permission, "system" responsible for intervention, exposures old or new • Practice change: no group allocation (usually before-after), no permission, "system" responsible for intervention, exposures old or new Typical features 3 Explanatory trial HSB-trial Permission: from target group Intervention: researcher Payer: research financer Target group: feasibility, informativness Results: specific (placebo a problem) Applicability: Effect in ideal circumstances more general health care system health care (future) target all effects (trial effect problem) effect in practice Typical features 4 • Allocation: often other than (individual) randomization • Unit: often collective (institution, group, community) rather than a person or her part (ear, eye, leg, arm) • applicability important (generalizibility sure only for the population studied) Why HSB-trials are important? An addition to traditional trials The key-tool for evidence based medicine is RCT In their current form, RCTs (individuals as units) best suit for studying (single) technologies in selected populations Currently, information is lacking of: 1. impact of technologies in practice 2. systems, packages, treatment strategies Why HSB should be facilitated? This far few HSB-trials Likely important for evidence based health care (judged from the impact of RCTs) Some hindering factors which could be changed Facilitating factors • Need for evidence based health care • Coverage with evidence/ only in research development (big health care payers, USA, UK) • Increasing health care costs • Finnish public health care system • Research valued by service providers • Well-educated population • Good health registers Hindering factors 1 • Researchers’ visions, know-how, attitudes • costs • research norms and laws Visions and know-how. Only few trials this far --> not an established research tradition, "Have not thought of“, “Is real research”: teaching (this course). Costs may be lowered by 1. Integrating research into normal services 2. collecting outcome data from routine records, e.g. registers Hindering factors 2 Research regulation needs change to allow HSB-trials and proper ethical evaluation, both within Finland and internationally. A key issues: informed consent (from who, should it be asked) E.g.: giving health education, offering screening, organizing services, changing environment Finnish HSB trials • • • • • • • • • • • • • Cancer screening trials (Hakama et al) Malila EBMeDS (supported decision making in health centers) Kaila LATE (quality in health centers) Yhteispeli (well-being of school-children) Isolation in psychiatric care ? Psychically ill parents' children (Solantaus) Young men at risk of marginalization (Stakes) Hospital discharge practices (Marja-Leena Perälä) Vaccination trials (KTL) Anti-smoking intervention among youth (KTL) Ergonomy in kitchens (TTL) Employment paths (TTL) Promotion of sexual health(KTL) Register based trial on birth education Elina Hemminki, Kaija Heikkilä, Tiina Sevon, Päivikki Koponen Hemminki et al 2008, BMC Health Services Research 8:126 (tabled) Childbirth classes 1 • 20 maternity centers in Helsinki matched for pairs and randomly allocated to intervention and control. • Intervention: further education to nurses and leaflets to mothers • Outcome: mode of delivery • Data collected from registers • Childbirth classes: an existing program • Responsibility of municipalities to organize and finance; content not defined Childbirth classes 2 • Called research originally; by the advice of service providers (administration) divided into program (intervention) and research (data collection). • Design: to guarantee data with small money (routine registers) • Maternity centers invited by service administrators and researchers • Invitation to the head nurse of each district with letters and phone calls; personal contacts by a midwife researcher • sent to be handled in ethics committees; did not succeed (rules did not apply) Childbirth classes 3 For • researchers received administrative backup from the city health administrators • no press coverage Against • lack of research administration rules • reluctant service providers (public health nurses) >< competition between midwife and nurse professions • No request from women Childbirth classes 4 • No protest on only half of the centers being exposed • Reluctance of the field (not felt important) • Professional battle Which rules apply? Medical research or developing health care? Is approval from ethics committee needed? Meetings and discussions with City health planners, people responsible for developing maternity health centres The body coordinating research activities within Helsinki city health services (TUTKA) Helsinki University coordinating ethics committee Gynaecology and obstetrics committee TUTKA: “Development project, no ethics committee statement is required” KIVA school trial • Financed by the Ministry of Education, Turku University, Cristina Salmivalli • Developing a program to prevent bullying at school grades 1-3, 4-6, 7-9, evaluating it, 2006 • "HSB": Program first introduced to 117 schools; how allocated to intervention and control? What are outcome measures? • All Finnish schools can get the intervention autumn 2009 • KiVa Koulu web-pages: the trial is difficult to identify (hidden), not in English Problems in law and rules Sir Bradford Hill 1963: "there is no one way of doing clinical trials ethically, and giving detailed advice as if there were, will harm both research and ethics". BMJ 1963, 5337: 1043-9 Key actors 1: ethics committees • one of the gate-keepers: what can and cannot be made • starting point: participant protection • time and cost implications on research; no responsibility to research financers • human rights --> scientific content • ethics --> surveillance of law and rules • committees over-do (protection) and under-do (motives for research) • no "formal decision" --> no easy appeal Relation between research and service • different criteria for research and routine care (accepted technology) • easier to use without research than to start with research (time, money, know-how, rewards) --> care experiments • "I need permission to give a drug to half of my patients, but not to give it to all of them" (Richard Smithells) Exercise 1 Read the 3 indicated papers. Select one, write 1 page (spacing 1) what (new) you learnt, what you agree/ disagree/ wonder. Send it to me, latest on Sunday 17 evening by e-mail Exercise 2 • Two people together (if not even, one 3 people) • Select one topic of the list or choose your own; let me know on Tuesday your topic + who form a pair. • Prepare a presentation for next Monday, either a power-point or a paper to be distributed, 10 min • Divide the work among yourself, discuss the plan, present it. • Focus: population, unit, allocation, implementation, outcome measures, data collection, costs, payer, actors, permissions • Be prepared to comment other people’s presentations. Exercise 2 topics 1. Can health education to school-children by ordninary health services reduce use of sweets (candy, soft drinks, cookies..)? 2. Does it matter if .. (päivystys) in health centers is done by regular physicians or outsider firms (reppufirmat)? 3. Is it better to run maternity centers combined to child wellfare centers or to family planning. 4. Can use of psyvchotropic drugs among old in institutions be prevented by adding social activities