Diabetes mellitus constitutes a major healthcare problem globally. In the United States, 10.3 million people were reported with this disease in 1997; and an additional 5.4 million people were suspected to have diabetes, but were not diagnosed nor treated 1. The total number of individuals with identified diabetes is expected to increase to approximately 16 million by 2010 2, and 22 million by 2025 3. Most patients have type 2 diabetes mellitus (T2DM), characterized by a combination of insulin resistance and progressive loss of insulin secretion. In approximately 58% of patients with T2DM, the disease is not sufficiently controlled based on American Diabetes Association (ADA) recommendations 4. As a result, patients experience substantial morbidity and mortality due to complications resulting from chronic exposure to hyperglycemia 56789 although major clinical trials have demonstrated that improved glycemic control can significantly reduce the risk of complications 1011.

Oral antidiabetic agents (OAs) are the current mainstay of T2DM management, but their long-term efficacy is limited by the progressive and irreversible loss of β-cell function, whereas insulin therapy remains continually effective. However, initiation of insulin injections frequently is restricted as a "last resort" for a variety of reasons such as fear of needles, inconvenience, and technique. Indeed, several studies have shown delays of 4 years or longer in the uptake of insulin therapy by patients uncontrolled with multiple OAs 1213. This delay may accelerate β-cell failure and disease progression and hasten development of complications, highlighting the need for improved patient acceptance of insulin therapy so it may be used in a timelier and appropriate manner.

Non-invasive methods of insulin delivery, such as inhaled insulins, may increase the acceptance, and thus earlier initiation, of insulin therapy. The first inhaled human insulin (INH, Exubera^® (insulinhuman [rDNA origin]) Inhalation Powder) was recently approved by the European Commission and the US Food and Drug Administration to deliver effective glycemic control in adult patients with type 1 or type 2 diabetes. Phase 3 trials have demonstrated that INH is as efficacious as subcutaneously (SC) injected regular human insulin in type 1 diabetes (T1DM) 1415 or T2DM 16, and superior to OAs in T2DM 17181920. INH offers a number of potential advantages 14151617181920212223242526272829, including a time-action profile that is more suited for mealtime insulin administration compared with SC regular human insulin 2122. Another advantage of INH is that patients with T2DM who are poorly controlled with OAs are able to attain good glycemic control by simply adding INH to their oral regimen 1819. In a 12-week study by Rosenstock and colleagues, adjusted treatment group differences from baseline (9.5–9.6%) in glycosylated hemoglobin (HbA_1c) of -1.2% and -1.7% were observed when INH was substituted for combination OA therapy or added to combination therapy, respectively 18. Furthermore, Phase 3 trial results indicate that INH lowers HbA_1c to ADA target levels (HbA_1c < 7%) in more than one third of patients after only 3 months of treatment 161820.

Evidence from completed Phase 2 and 3 studies up to 4 years in duration support that insulin administration via the lung is generally well tolerated in patients with T1DM or T2DM 1213141516171819202324. A slight, but clinically insignificant, decline in lung function was observed to a greater extent in patients treated with INH compared with control patients; this decline was nonprogressive and reversed after discontinuation of INH therapy 141516192324. The only significant clinical adverse effect associated with INH was cough, which was generally characterized as mild to moderate in severity, decreased over time, and was not associated with decline in lung function 14151619. In extensions of several Phase 2 studies, patients appear to be satisfied with INH, as evidenced by more than 85% of subjects choosing to remain on INH 29. Subjects also demonstrated a preference for INH to SC insulin or OAs when offered the choice 25262728.

This combination of attributes suggests that INH may be a valuable addition to the treatment options available to patients and physicians in attaining glycemic control. However, these data on efficacy and safety cannot address the overall health benefits and pharmacoeconomic impact of INH in society. No randomized trials have examined the potential advantages of increased uptake of insulin due to avoidance of injections.

This paper describes the Real World Trial of INH, which was designed to overcome the limitations associated with traditional study design. The Real World Trial has adopted a novel design in order to estimate the effects of the availability of INH on physician and patient acceptance of insulin, health status (e.g. glycemic control), and health economic parameters.

Estimating the potential benefits of INH on improved insulin acceptance and glycemic control requires adaptation of conventional randomized trial designs. The Real World Trial will estimate the effects of insulin acceptance on glycemic control while preserving the benefits of randomization to reduce bias. Its design takes into account the need for internal and external validity that the results are attributable to the intervention (i.e. INH availability) and that they may be applied to general practice. This design is practical because patients and physicians have real life choices that are normally "designed out" and removed in a "classical" clinical trial. In this trial, physicians are able to freely use the new treatment without some of the restrictions of a traditional trial; physicians and patients can interact normally in discussing, designing, and optimizing treatment regimens. The study also will encompass a variety of clinical practices including community specialist and primary care to better reflect the expected post-marketing use of the product. The findings from the Real World Trial will provide important information that can be generalized for clinicians and healthcare policymakers in distinguishing the role of INH by identifying the association between INH availability/acceptability and glycemic control, and establishing the effects on resource use. Recruitment for the Real World Trial concluded in spring 2006 after recruitment of 739 patients, and results are expected in 2007.

A questionnaire-based feasibility study examining the theoretical effects of the availability of inhaled insulin on treatment in 779 patients with T2DM who were not achieving target glycemic control with their current therapy was recently completed 32. This randomized, controlled trial provided supportive insight into the design of the Real World Trial, as well as evidence that inhaled insulin may lead to improved uptake of insulin. Treatment preferences expressed by subjects of the feasibility study, by number of OAs, are described in Table 1. In each group, 77% of patients had HbA_1c values greater than 10%, and the majority of patients were receiving treatment with one or more OA along with dietary and lifestyle advice. Patients were three times more likely to choose insulin therapy when inhaled insulin was available (odds ratio [OR]: 4.16; 95% confidence interval [CI]: 2.93 to 5.95, p < 0.0001) and significantly fewer patients in the group offered inhaled insulin chose to make no change to their therapy compared with patients offered standard treatments only (OR: 0.49; 95% CI: 0.36 to 0.67, p < 0.0001). The proportion of patients choosing insulin in both groups increased with the number of OAs currently being taken, and inhaled insulin was the most frequently chosen treatment option. Aversion to incorporating SC insulin into therapy remained strong, despite high HbA_1c levels, and the availability of insulin pens in some countries. The enhanced willingness of those offered the option of inhaled insulin treatment to change to a more appropriate therapy increases the potential for achieving improved glycemic control and reducing complications, associated morbidity, premature mortality, and increased cost of diabetes.

One of the key decisions that required consideration in the design of the Real World Trial was the question of the appropriateness of patient- or cluster-level randomization. Rather than allocating patients between treatment groups using a random process, cluster-level randomized trials allocate patients in groups 33. The rationale for cluster-level randomized trials is apparent when treatments cannot be applied to individuals, for example in a trial of fluoridation of water supplies. Similarly, a cluster trial would be appropriate when there is a risk of contamination of treatments between groups, for example where a healthcare professional was expected to provide different levels of healthcare promotion advice to different patients.

Thus, cluster-level randomized trials are necessary in certain circumstances, but they are associated with a certain disadvantage – we cannot consider patients allocated in this way to be independent of each other. This lack of independence must be addressed in the analysis. The degree to which clustering affects the design of a study depends upon the extent to which patients are clustered within their units of allocation and the number of patients per unit. However, patient per patient, a clustered trial always will be less efficient than an individual patient randomized trial. In addition, there are potential problems associated with the implementation of cluster-level randomized trials. For example, when a trial intervention is considered particularly attractive or of interest to investigating clinicians, the disappointment engendered from being allocated to the control rather than experimental condition may affect the behavior of the investigator. Thus, control group investigators may recruit fewer patients or different patients from those recruited by the experimental group investigators, or they may experience greater loss to follow up. The treatment of interest in the Real World Trial may be applied to individual patients without the risk of contamination; at the time of the study enrollment INH was available only as an investigational product. Thus, there is no requirement to move to a cluster design and the associated issues regarding trial quality and loss of efficiency.

The novel design of the Real World Trial, randomizing patients to scenarios where different treatment options are available for the management of their condition, is necessary in order to provide an estimate of the potential utility of the intervention in standard practice. Trials required for the registration of a new product generally do not provide such information, which is increasingly necessary for healthcare policymakers and clinicians in an environment where healthcare resources are limited. It is very likely that we will see many more such trials in the future.

Nick Freemantle has received funding from Pfizer Inc for his work on the design of this trial.

Thomas Strack is an employee of the sponsor, Pfizer Inc.

NF and TS conceived of the study and participated in its design and helped to draft the manuscript. All authors read and approved the final manuscript.