Dr. E is a primary care physician (general practitioner) with an extensive patient list. He is approached by researchers leading a large epidemiological study into the association between asthma and heart disease. He is asked for de-identified patient data on all his patients with a history of asthma concerning their age, sex, age at first diagnosis with asthma, current medication, and cardiovascular history. He is assured that the research ethics committees of both his local university hospital and the researchers’ own institution has approved the study. He is offered payment for the administrator’s time required to prepare this data.
Epidemiology may be defined as “the study of the distribution and determinants of disease in human populations” (Dunn, 2003, p. 34). There is no very tight distinction between epidemiology and other data collection and analysis practices in public health, but for practical purposes epidemiology can be considered a research activity, which is to say that it is concerned with producing generalizable scientific knowledge. A standard textbook of epidemiology (Farmer et al., 1996, p. 6) defined the principal uses of epidemiology as:
The investigation of the causes and natural history of disease, with the aim of disease prevention and health promotion.
The measurement of health care needs and the evaluation of clinical management, with the aim of improving the effectiveness and efficiency of health care provision.
Both of these can be read as service objectives (What caused this disease outbreak? How can we assess the health needs of this group of patients?) and as research objectives. For the purposes of this chapter, we are interested mainly in epidemiological research. Epidemiological research has a number of different approaches, which include experimental methods (principally randomized controlled trials) and observational methods (principally cohort studies, case–control studies, cross-sectional studies, ecological studies, and analyses of routine data). The ethical issues involved in experimental methods in epidemiology are discussed in Ch. 27. This chapter will be concerned mainly with observational methods in epidemiological research.
Classical examples of epidemiological research include John Snow’s identification of the waterborne nature of the cause of cholera in the early nineteenth century, Richard Doll and colleagues’ establishment of the causal link between smoking and lung cancer in the late 1940s, and the definition of acquired immunodeficiency syndrome in the early 1980s.
Because epidemiological research is concerned with the causes and management of disease and ill-health in populations, there is an associated generic risk that it may place community interests in conflict with the interests of individual patients and citizens. Because epidemiological research is concerned with characteristics of populations, a central methodological preoccupation of epidemiologists is with bias. In particular, epidemiologists are concerned with biased sampling from populations, which would produce inaccurate representations of the variables of interest in a population, and false or misleading interpretations of associations between such variables. In a cohort study, for instance, if significant numbers of the participants in a cohort were “lost to follow-up” or withdrew their consent to the use of their records in a study, this could fatally undermine the study. For this reason, many epidemiologists are keen to use designs that minimize the risk (as they see it) that their sample will be biased in construction or in operation. Two ways that bias can arise are that a significant number of people might refuse to participate or that a significant number of people might drop out during the course of a study. The major ethical debates in epidemiology have, therefore, concerned consent and methods of recruitment and retention in studies (Hart et al., 1997).
Other ethical issues do arise. In epidemiological studies that involve collecting new data from study participants (such as by measuring blood pressure or collecting genetic samples), an important question arises when person-specific data are collected that indicate that the person may have a health problem or risk which was not foreseen by the person or the researcher – so-called “incidental findings.” Does the researcher or research team have a duty of care to the patient that would require disclosure of this information? What if there is nothing that can be done about this information? What if this information represents a risk to third parties? This problem can arise across the study as a whole once the findings of the study are interpreted (Richards, 2003). For instance, in genetic epidemiology, it can happen that a gene is identified that seems to raise the risk of heart disease in those that have a given allele. Should all those people in the study with this allele be informed? Do they have a right to this information? (Sharp and Orr, 2004).
Aside from individual-level ethical challenges, there may also be population-level challenges. While epidemiological research can be a powerful tool for identifying inequities in healthcare, epidemiological research can reinforce inequities both through study design and through the (mis)interpretation of results. For instance, an epidemiological study may be constructed to investigate a health issue in a particular ethnic group, where there is no particularly good reason to think that this ethnic group is different from the population at large from a health or medical point of view. The interpretation of the results of such a study (either by the researchers themselves or by some audience of the research) may then be used to reinforce stigma against this ethnic group (Singer et al., 2003). Although epidemiology is at least as much a social science as a medical and natural science, it is relatively unusual for the social impact of epidemiological research to be considered.
The most notorious example of unethical epidemiological research is probably the Tuskegee Syphilis Study. This study, initiated in 1932, was a study in a cohort of African-American men of the natural history of syphilis. The men were not told or were misled about the nature of the disease they had; they were denied treatment, discouraged from seeking alternative medical advice or treatment, even when effective treatment for syphilis became available in the 1940s, and were followed up until 1972 when the study was closed owing to the outcry when the study came to public attention (Reverby, 2000). Although this study is an extreme example, which is hardly representative of modern epidemiological research, it does illustrate in a salutary way the ethical issues described in this section.
To this author’s knowledge, no jurisdictions have specific laws or regulations for epidemiological research as such. In this respect, epidemiological research may be considered less regulated than clinical trials. However, many activities that form central parts of epidemiological research do fall under regulation: either relating to research in general or framed with other purposes in mind. An example will suffice to explain this. Firstly, many epidemiological studies involve the collection of new data from research participants. For example, participants in a study of blood pressure and mortality may be asked to undergo blood pressure measurements and blood samples, in addition to giving their consent for access to their medical records for long-term follow-up. These blood samples may be processed for DNA analysis in the context of a search for genetic factors that may partially explain health outcome variations over the course of the study. The additional measurements and invasive procedures to collect data and blood samples would fall under relevant law and regulations relating to research with human subjects in general (the Declaration of Helsinki [World Medical Association, 1964], Code of Federal Regulations [US Department of Health and Human Services, 2005], the common law of consent and applicable professional guidance in the UK, and so on). The processing of patient data for research purposes would fall under data protection legislation and confidentiality law in Europe, or the US Health Insurance Portability and Accountability Act (1996) and associated regulations in the USA. In both Europe and the USA, this class of legislation was not framed with research purposes in mind, although both bodies of legislation make mention of research in passing. The archiving of human biological samples and their use for purposes known at the time of collection, or re-use for purposes determined subsequent to collection by the original research team or some other group, may be governed by common law provisions relating to consent, by novel legislation such as the UK Human Tissues Act (2004), or by regulations drafted pursuant to existing law and regulation ad hoc. (For more information on this issue, see Ch. 23.) It is arguable that the regulation of epidemiological research is somewhat unstructured and difficult to apply given the rapidity with which epidemiology has developed since the early 1980s under the influence of new technology (both information and data-processing and biomedical research techniques) and the scaling up of research from studies based on single communities to large-scale multisite projects.
In contrast to experimental methods in medical research, relatively little has been written about the ethics of epidemiological research. It is not a central focus of the Declaration of Helsinki, for instance, although the provisions of the Declaration relating to non-therapeutic research procedures would apply to epidemiological studies, which involve the collection of new data from research participants (World Medical Association, 1964). However, the Council of the International Organizations of Medical Sciences (1991) has issued international guidance on ethics in epidemiology (these are currently being revised). From the early 1990s onwards, a series of ethical guidelines were produced at national and international levels, most importantly the statement of Good Epidemiological Practice of the International Epidemiological Association (2002 [updated in 2004]). There have been a number of academic reviews of ethical principles in research, and several specialist epidemiological associations have produced guidance on ethical issues in their specific areas of epidemiological research (Coughlin, 2007).
Given the complex situation regarding the regulation of epidemiological research, there has been a growth in empirical studies that seek to examine the awareness among researchers of existing guidance and ethical principles in epidemiological research, the development of consensus within the research community and in society at large about what guidance and principles are required, the specific ethical issues in research, and how best to make the trade off between individual patient, community, researcher, and societal interests in epidemiological research (Prineas et al., 1998; Wynia et al., 2001; Kessel, 2003). This growth in empirical research as a way of defining issues and generating consensus has been particularly notable in relation to research involving genetic databases and tissue sample collections, and to a lesser extent in relation to research involving the use of patient data without consent or with generic prospective consent (i.e., a consent to the future collection and use of data without the researcher returning to re-seek consent from participants in connection with each future use) (Richards et al., 2003; Tutton and Corrigan, 2004; Eriksson and Helgesson, 2005; Matsui et al., 2005; Stolt et al., 2005).
Empirical research can be useful in a number of ways: it can identify the issues that are at stake in practice (as opposed to what theoreticians think they should be); it can evaluate the relative importance of different concerns; it can gauge how widely particular views are held; and it can act as a kind of surrogate for a poll on policy decisions to be made. Although this last use of empirical data is frequently regarded as disreputable, in the context of epidemiological research it may be important. Since such research may depend for its ethical justification on appeal to the public interest, it may be important to establish both the content of the public interest (what is it in this case?) and how far the public interest is consistent with the aggregate of individuals’ interests (Ashcroft, 2004). The latter can be established by survey or polling, although this should be handled with care (Rose, 2001; Wendler, 2006).
No good systematic reviews have been conducted of empirical research in this area, so findings need to be handled with care. For example, Richards and colleagues (2003) established that research participants in a breast cancer epidemiology study would welcome feedback of the findings of research. Yet while acknowledging that individual feedback would be desirable, they recognise the problems this might create. In a later study, Dixon-Woods et al. (2006) found that participants wanted individual feedback and found general feedback unhelpful. The current state of empirical research considering ethics in epidemiology is, therefore, rather inchoate. At its best, it can help researchers to frame their own thoughts. It is too early to suggest that empirical research can lead policy making in this area, but it can perhaps be a counsel of humility regarding how far we know the ethical answers.
It is clear that many practice and policy issues remain to be resolved in epidemiological research, and that guidelines and regulations may not be entirely helpful either in conducting epidemiological research or in setting good practice standards. Nonetheless, some basic principles apply generally. Firstly, the general principles of research ethics apply to epidemiological research as they do to any biomedical and public health research. Consequently, so far as is possible, research should be carried out with the voluntary and informed consent of the research participants, in such a way as to minimize harm, risk of harm, and inconvenience to them, while maximizing the benefits to science and society that could accrue from the successful conduct of the project. Epidemiological research will rarely directly benefit the research participant, although some recruitment and data-collection processes do involve what is effectively a health screen for the participant. In this case, it will be important to ensure that this is conducted with proper information to the participant, and to a clinically competent standard. It is necessary to have a policy on the feedback of unexpected findings of such a screening process to the participant (Illes et al., 2006).
The main challenges lie where either it is not possible to adhere to the normal ethical standards in research or there is not yet a consensus on what those standards require. Examples of the latter include the as-yet-unresolved controversy on the scope of consent to future uses of research samples in genetic database research and the debate on feedback of individual test results from research where those results are not validated to clinical quality standards, their interpretation is as yet uncertain, or there is no clinical intervention to address any health problem they may indicate (Duncan et al., 2005).
The standard examples of research that may depart from the standard of informed voluntary consent are research projects that involve collection of patient data from records without the express prior consent of the patient. Most health systems contain a vast amount of useful clinical data, which can be a very valuable resource for epidemiologists. In some cases, data can be collected in an anonymized form, such that the researcher cannot, in principle or practice, identify any individual from the data collected. It can be argued that this does not, in fact, violate any privacy right of the patient. In other cases, the data can be collected in a form that could in principle be decoded to allow identification of the patient, although in practice this would be very difficult, and coding and firewalls can be put in place to make this as difficult as possible. Here, the justification might be based on the minimal risk of any actual harm to the patient even pursuant to their possible identification by a malign third party. This justification would essentially accept that a privacy right of the patient may have been violated but hold that this was at worst a technical breach only, outweighed by the value of the research and the minimal risk of harm to the patient. Some commentators would go further and argue that there is not even a technical breach of a patient’s privacy right in such a case. There are standard arguments here. One is to hold that some breaches of confidence are only breaches of “prima facie” rights, since patient confidentiality can always be overridden in the face of a clear public interest, such as protection of third parties from serious harm or in the administration of justice (Gostin, 2000). Another is to hold that individual rights of privacy arise from communal interests, rather than the other way around: the “communitarian” approach (Rubinstein, 1999). This type of justification may be applied where patient information is identifying but where there are privacy rules and safeguards to protect the patient’s privacy from invasion by unauthorized external parties. In any case, it is clear that there is a very strong presumption in favor of patient confidentiality and privacy, and in favor of seeking patient’s consent to participation even in minimal-risk database research. Part of the case for this presumption lies in the common interest in securing patient trust and support for medical research. This may, in turn, underwrite such practices as “broad” consent to re-use of tissue samples, by ensuring that the public at large understands and supports medical research as such and accepts that detailed “narrow consent” may be a costly and inefficient brake on an activity that is widely supported by the public (O’Neill, 2002). The present author’s research suggests that most participants are less concerned with issues around consent than they are with governance. So long as they believe that the study is appropriately governed, and driven by a concern with patient welfare rather than political or commercial interests, they have a high level of trust in this type of research (Williamson et al., 2004).
The role of the research ethics committee or institutional review board here may be particularly important as it can act as an independent check on the claim that the research represents an appropriate balance between individual interests, the interests of the researcher, and the public interest.
In the light of the above discussion, the main concerns of Dr. E will be as follows. Firstly, would passing on data collected from his patient’s notes in this way be lawful and consistent with relevant professional guidance? The fine detail here may vary between jurisdictions. But in principle, although this information is not explicitly identifying, in that names and addresses are not collected, a third party could potentially identify one or more of the patients in the data table from the information given. So what the doctor needs to satisfy himself about would be the nature of the procedural and technical protections in place to prevent researchers or unauthorized third parties from using the information in that way. He would also need to satisfy himself that this is not a breach of confidence or, if it is, that it is a justified breach. For practical purposes, assurance by a research ethics committee that the study has their ethical approval may satisfy him on this point. Nonetheless, primary care physicians would normally be legally responsible for this decision, and liability in tort for a breach of confidence would not be removed by the fact of ethical approval from an ethical committee. The central issue, therefore, would be whether patient consent is required for this data processing. This is, in part, a legal question. So far as it is an ethical question, we are to consider first whether the patient has a right to be asked for his or her consent. Most would consider that it would be good practice to seek the patient’s consent; however, this might introduce significant bias into the sampling, and moreover we have some research evidence to suggest that patients regard collection of their data for this sort of purpose as important and they are surprised by the request for their consent (Robling et al., 2004). Thus Dr. E could be advised to take the following approach. Firstly, make it generally known to his patients that medical research studies using their data, subject to independent ethical review and careful information security controls, are carried out with his practice’s cooperation. They can be asked to notify the practice if they do not wish their information to be used in this way, thus giving them an opt-out. However, the opt-out would be a generic opt-out, so that there is no risk (from the scientific point of view) of “consent bias” in recruitment to particular studies (Medical Research Council, 2000). Secondly, he should ensure that he is satisfied that the risk of patient identification has been minimized. On that basis, he can be satisfied that any breach of patient confidentiality and risk to patients’ interests has been minimized and he can agree with the request. The issue of whether paying his practice for the administrative work required in producing the data may raise a question of whether he is being induced to hand over this information, but since it is only a question of reimbursement for time spent on research at the expense of other practice activities, this is not a significant concern.