|Home | About | Journals | Submit | Contact Us | Français|
To synthesize the cost-effectiveness (CE) of interventions to prevent and control diabetes, its complications, and comorbidities.
We conducted a systematic review of literature on the CE of diabetes interventions recommended by the American Diabetes Association (ADA) and published between January 1985 and May 2008. We categorized the strength of evidence about the CE of an intervention as strong, supportive, or uncertain. CEs were classified as cost saving (more health benefit at a lower cost), very cost-effective (≤$25,000 per life year gained [LYG] or quality-adjusted life year [QALY]), cost-effective ($25,001 to $50,000 per LYG or QALY), marginally cost-effective ($50,001 to $100,000 per LYG or QALY), or not cost-effective (>$100,000 per LYG or QALY). The CE classification of an intervention was reported separately by country setting (U.S. or other developed countries) if CE varied by where the intervention was implemented. Costs were measured in 2007 U.S. dollars.
Fifty-six studies from 20 countries met the inclusion criteria. A large majority of the ADA recommended interventions are cost-effective. We found strong evidence to classify the following interventions as cost saving or very cost-effective: (I) Cost saving— 1) ACE inhibitor (ACEI) therapy for intensive hypertension control compared with standard hypertension control; 2) ACEI or angiotensin receptor blocker (ARB) therapy to prevent end-stage renal disease (ESRD) compared with no ACEI or ARB treatment; 3) early irbesartan therapy (at the microalbuminuria stage) to prevent ESRD compared with later treatment (at the macroalbuminuria stage); 4) comprehensive foot care to prevent ulcers compared with usual care; 5) multi-component interventions for diabetic risk factor control and early detection of complications compared with conventional insulin therapy for persons with type 1 diabetes; and 6) multi-component interventions for diabetic risk factor control and early detection of complications compared with standard glycemic control for persons with type 2 diabetes. (II) Very cost-effective— 1) intensive lifestyle interventions to prevent type 2 diabetes among persons with impaired glucose tolerance compared with standard lifestyle recommendations; 2) universal opportunistic screening for undiagnosed type 2 diabetes in African Americans between 45 and 54 years old; 3) intensive glycemic control as implemented in the UK Prospective Diabetes Study in persons with newly diagnosed type 2 diabetes compared with conventional glycemic control; 4) statin therapy for secondary prevention of cardiovascular disease compared with no statin therapy; 5) counseling and treatment for smoking cessation compared with no counseling and treatment; 6) annual screening for diabetic retinopathy and ensuing treatment in persons with type 1 diabetes compared with no screening; 7) annual screening for diabetic retinopathy and ensuing treatment in persons with type 2 diabetes compared with no screening; and 8) immediate vitrectomy to treat diabetic retinopathy compared with deferred vitrectomy.
Many interventions intended to prevent/control diabetes are cost saving or very cost-effective and supported by strong evidence. Policy makers should consider giving these interventions a higher priority.
The cost of diabetes in the U.S. in 2007 was $174 billion (1). Many interventions can reduce the burden of this disease. However, health care resources are limited; thus, interventions for diabetes prevention/control should be prioritized. We wanted to compare the effectiveness and costs of various interventions to find those that were the most effective for the least expense. Cost-effective analysis is a useful tool for this purpose. Such analyses consist of compiling incremental cost-effectiveness ratios (ICERs), which are calculated as a ratio of the difference in costs to the difference in effectiveness between the intervention being evaluated and the comparison intervention.
With the same health outcome indicator, ICERs of interventions are comparable. Therefore, these ICERs can make it easier to decide how to allocate resources. Although many cost-effectiveness (CE) analyses of diabetes interventions have been published, their qualities and conclusions vary. A systematic review, which appraises individual studies and summarizes results, would aid policy makers and clinicians in prioritizing interventions to prevent or treat diabetes and its complications.
Few investigators have conducted systematic reviews of the CE of diabetes interventions (2–5). The systematic review presented here, following the Cochrane Collaboration's protocol (6), includes all English language studies available from 1985 to May 2008. The interventions included only those recommended by the 2008 American Diabetes Association (ADA) Standards of Medical Care in Diabetes (7).
We searched the Medical Literature Analysis and Retrieval System Online (MEDLINE), Excerpta Medica (EMBASE), Cumulative Index to Nursing and Allied Health Literature (CINAHL), PsycINFO, Sociological Abstracts (Soc Abs), Web of Science (WOS), and Cochrane databases to identify relevant studies. We created a search strategy involving medical subject headings. The key words—and what each indicated—were:
Database searches were based on matches in all four keyword categories. Reference lists of all the included articles were screened for additional citations, and Diabetes Care was reviewed manually, issue by issue, as the journal was expected to be highly relevant.
Criteria for inclusion in the review were 1) original CE analysis; 2) intervention directed toward patients with type 1, type 2, or gestational diabetes mellitus (GDM) and recommended in the 2008 ADA standards for medical care (7); 3) outcomes were measured as life years gained (LYGs) or quality-adjusted life years gained (QALYs); and 4) publication in the English language occurred between January 1985 and May 2008 (2). To ensure that only studies with acceptable quality were included, we limited the analysis to studies considered good or excellent according to a 13-item quality-assessment tool based on the British Medical Journal authors' guide for economic studies (8).
To make ICERs comparable across the studies, all costs are expressed as 2007 U.S. dollars with adjustment from other currencies, as needed, using the Federal Reserve Bank's annual foreign exchange rates (9) and from other cost years using the Consumer Price Index (10). If a study did not mention the year used in cost calculations, we assumed cost was as of one year before publication. ICERs were expressed as dollars per QALY or dollars per LYG and were rounded to the nearest hundred dollars per QALY or LYG.
Interventions were classified based on the level of CE by convention as described in the literature (2,11,12)—cost saving (an intervention generates a better health outcome and costs less than the comparison intervention) or cost neutral (ICER = 0); very cost-effective (0 < ICER ≤ $25,000 per QALY or LYG); cost-effective ($25,000 < ICER ≤ $50,000 per QALY or LYG); marginally cost-effective ($50,000 < ICER ≤ $100,000 per QALY or LYG); or not cost-effective (>$100,000 per QALY or LYG)—and whether evidence for the intervention's CE was strong, supportive, or uncertain as described below.
There were two grades of evidence included in the “strong” group. Grade 1 was defined as 1) CE of the intervention was evaluated by two or more studies; 2) study quality was rated good or excellent; 3) effectiveness of interventions based on well-conducted, randomized clinical trials with adequate power and generalizable results or meta-analysis or a validated simulation model; 4) effectiveness of interventions rated as level A (clear evidence from well-conducted, generalizable, randomized controlled trials that were adequately powered; compelling nonexperimental evidence, i.e., the all or none rule developed by the Centre for Evidence-Based Medicine at the University of Oxford, U.K.) or level B (supportive evidence from well-conducted cohort studies or supportive evidence from a well-conducted case-control study) according to the 2008 ADA standards of medical care (7); and 5) similar ICERs reported across the studies. Grade 2 was defined as the same as Grade 1 except that the CE was based on only one study and the study was rated as excellent.
We called the level of evidence “supportive” if only one study, rated lower than excellent, evaluated the CE of the intervention or if the effectiveness of the intervention was supported by either level C evidence (supportive evidence from poorly controlled or uncontrolled studies, or conflicting evidence with the weight of evidence supporting the recommendation) or expert consensus (level E) in ADA recommendations (7). The term “uncertain” was used to describe interventions with inconsistent evidence about CE across studies.
We reported the study results in two ways: 1) summarizing the key features and results for each included study; and 2) synthesizing the CE of the interventions based on the classification criteria described above. For the summary, we grouped interventions based on their intended purposes: a) preventing type 2 diabetes among high-risk persons; b) screening for undiagnosed type 2 diabetes and GDM; c) management of diabetes and risk factors for complications; d) screening for and early treatment of complications; and e) treatment of complications and comorbidities. We considered cases where the same intervention was applied to different populations or was compared with different interventions as different specific interventions and reported the ICERs separately. This was because both incremental costs and effectiveness of an intervention, and thus the ICERs, varied if the population and/or comparison group differed. If the CE of an intervention was evaluated from different study perspectives, we report the ICERs separately. We presented the ICERs in subgroups if their ICERs differed substantially from base-case analysis, and original studies reported the ICERs this way. If the study reported the ICERs only for population subgroups, we provided a range and, when available, trend of the ICERs. Finally, if a study used both LYGs and QALYs as study outcome measures, we reported the ICER in both costs per LYG and QALY.
In reporting the synthesized results, we applied the following rules: 1) We used the median ICER to represent the CE of an intervention if the intervention was evaluated by more than one study. 2) We reported the ICERs from the longer analytical time horizon if the intervention was evaluated from both short- and long-term perspectives. This was appropriate since many of the benefits of most diabetes prevention and control interventions would come from preventing diabetic complications, which occur later in life. 3) We chose the health care system as our primary study perspective for the purpose of cross-study and cross-intervention comparisons. This study perspective included all the medical costs incurred no matter who paid. 4) If the ICERs of an intervention differed substantially between the U.S. and other developed countries (mainly European countries, Australia, and Canada), we reported the summary results separately by labeling the ICER for the U.S. or for the other countries. 5) If the trial on which the CE of an intervention was based was conducted in a mixed population with type 1 or type 2 diabetes, we assumed the CE was the same for both types of diabetes.
The search yielded 9,461 abstracts. After reviewing the abstracts and subsequent reference tracking, we narrowed the focus to 197 possible original CE studies. Further review of the full text resulted in 56 CE studies that met our inclusion criteria. Figure 1 depicts the data abstraction process.
Table 1 shows the detailed description of the CE studies that we included according to intervention type (13–70). We first grouped similar interventions together, then arranged them chronologically and by the first author's last name. Some studies that evaluated multiple interventions appear in more than one category. The information used to describe each study included the intervention being evaluated; comparison intervention, population, and country setting; data sources for the effectiveness of the intervention; study methods; quality of the study; analytical time horizon; discount rate (a rate that is used to convert future costs and benefits into their present values); and ICER.
Thirty-nine of the 56 studies took a long-term analytical time horizon, such as 20–30 years or lifetime. Nearly all of the studies with the long-term horizon used simulation modeling. Only one study was conducted in a developing country (Thailand) (57). There were 48 excellent studies and 8 good studies. Only three studies took perspectives other than the health care system.
The interventions evaluated in these CE studies covered a wide range: lifestyle and medication therapy to prevent type 2 diabetes among high-risk individuals (eight studies); screening for undiagnosed type 2 diabetes or GDM (three studies); intensive glycemic control (12 studies); self-monitoring of blood glucose (one study); intensive hypertension control (four studies); statin therapy for cholesterol control (five studies); smoking cessation (one study); diabetic health education program (two studies); diabetes disease management program (two studies); screening to prevent diabetic retinopathy (five studies); optimal foot care to prevent foot ulcer and amputation (two studies); ACE inhibitor (ACEI) or angiotensin receptor blocker (ARB) therapy to prevent diabetic end-stage renal diseases (ESRD) (15 studies); comprehensive interventions using a combination of several of the above secondary prevention interventions (two studies); and interventions treating diabetic retinopathy and foot ulcers (two studies).
The classification of the interventions based on their level of CE and strength of evidence is presented in Table 2. For each intervention, we also described the number of studies that evaluated the CE of this intervention, its comparison intervention, and the study population in which the intervention was implemented. We reported the median and range of the ICERs across the studies.
Twenty-six interventions were classified as supported by strong evidence concerning their CE (Table 2). Among these, six interventions were cost saving, eight were very cost-effective, six were cost-effective, two were marginally cost-effective, and four were not cost-effective. These interventions consisted of primary prevention, screening for undiagnosed type 2 diabetes, diabetic risk factor control, early prevention of diabetes complications, and treatment of diabetes complications.
The six cost-saving interventions with strong evidence were 1) ACEI therapy for intensive hypertension control, as in the UK Prospective Diabetes Study (UKPDS), in persons with type 2 diabetes compared with standard hypertension control; 2) ACEI or ARB therapy to prevent ESRD for type 2 diabetes compared with no ACEI or ARB therapy; 3) early irbesartan therapy at the stage of microalbuminuria to prevent ESRD in people with type 2 diabetes compared with treatment at the stage of macroalbuminuria; 4) comprehensive foot care to prevent ulcers in mixed population with either type 1 or type 2 diabetes compared with usual care; 5) multi-component interventions for diabetic risk factor control and early detection of complications compared with conventional insulin therapy for persons with type 1 diabetes; and 6) multi-component interventions for diabetic risk factor control and early detection of complications compared with standard glycemic control for persons with type 2 diabetes.
Of the eight very cost-effective interventions with strong evidence, six were for persons with type 2 diabetes, one for persons with type 1 diabetes, and one for a mixed population with type 1 or type 2 diabetes. Interventions for type 2 diabetes included: 1) primary prevention through intensive lifestyle modification; 2) universal opportunistic screening for undiagnosed type 2 diabetes in African Americans between 45 and 54 years old; 3) intensive glycemic control as implemented in UKPDS; 4) statin therapy for secondary prevention of cardiovascular disease; 5) smoking cessation; and 6) annual screening for diabetic retinopathy and early treatment of it. The intervention for type 1 diabetes was annual screening for diabetic retinopathy and treating the positive cases. The intervention for mixed population of type 1 and type 2 diabetes was immediate vitrectomy to treat diabetic retinopathy compared with deferral of vitrectomy.
The six cost-effective interventions with strong evidence were 1) one-time opportunistic targeted screening for undiagnosed type 2 diabetes in hypertensive persons aged 45 years and older compared with no screening; 2) intensive insulin treatment for persons with type 1 diabetes compared with conventional glycemic control; 3) UKPDS-like intensive glycemic control applied to the U.S. health care system among adults younger than age 54 years with type 2 diabetes compared with conventional glycemic control; 4) intensive glycemic control by a Diabetes Prevention Program (DPP) type of intensive lifestyle intervention in persons with newly diagnosed type 2 diabetes compared with conventional glycemic control; 5) statin therapy for primary prevention of cardiovascular disease in persons with type 2 diabetes compared with no statin therapy; 6) multi-component interventions including insulin therapy, ACEI therapy, and screening for retinopathy in persons with type 1 diabetes compared with intensive insulin therapy.
The two marginally cost-effective interventions with strong evidence were 1) intensive glycemic control for all U.S. residents with type 2 diabetes diagnosed at age 25 years and older compared with usual care; and 2) screening for diabetic retinopathy every two years compared with screening every three years in persons with type 2 diabetes.
The four interventions with strong evidence of not being cost-effective were 1) one-time universal opportunistic screening for undiagnosed type 2 diabetes among those aged 45 years and older compared with no screening; 2) universal screening for type 2 diabetes compared with targeted screening; 3) intensive glycemic control in the U.S. setting for patients diagnosed with diabetes at older ages (55–94 years of age) compared with usual care; and 4) annual screening for retinopathy compared with screening every two years. All these studies were for type 2 diabetes.
There were 18 specific interventions for which their CEs were based only on “supportive” evidence. Among them, 15 were each supported by one CE study, 13 were supported by level C or level E evidence, and five were supported by level A or B evidence as defined in the 2008 ADA standards of medical care in diabetes (7). For those interventions with level A or B evidence, the CE of each intervention was evaluated by one study with a quality of being “good.”
In terms of the level of the CE, 10 of the 18 specific interventions based on “supportive” evidence were cost-saving, including 1) screening using the sequential method (50-g glucose challenge test followed by 100-g glucose tolerance test [GTT]) for GDM in 30-year-old pregnant women between 24–28 weeks' gestation compared with no screening; 2) screening for GDM using the 100-g GTT method compared with no screening; 3) the sequential method compared with 75-g GTT screening for GDM; 4) 100-g GTT compared with 75-g GTT screening for GDM; 5) diabetes self-management education for persons with type 1 diabetes compared with no education; 6) full-reimbursement policy for ACEI for patients with type 1 diabetes compared with patients paying out-of-pocket; 7) full-reimbursement policy for ACEI for patients with type 2 diabetes compared with patients paying out-of-pocket; 8) screening using a mobile camera at a remote area and processing data in a reading center compared with a retina specialist's visit in a mixed population of type 1 and type 2 diabetes; 9) screening for diabetic nephropathy and ensuing ACEI or ARB therapy in persons with type 1 diabetes compared with no screening; and 10) intensified foot ulcer treatment in a mixed population with type 1 or type 2 diabetes compared with standard treatment.
Seven of the 18 specific interventions were very cost-effective: 1) primary prevention of type 2 diabetes in women with GDM history through intensive lifestyle intervention compared with usual care; 2) universal opportunistic screening for type 2 diabetes in African Americans aged 25–44 years compared with no screening; 3) 100-g GTT compared with the sequential screening method for detecting GDM in 30-year-old pregnant women between 24–28 weeks' gestation; 4) diabetes self-management education for persons with type 2 diabetes compared with no education; 5) disease management programs using specialist nurse–led clinics to treat and control hypertension or hyperlipidemia in patients with type 2 diabetes in a city in England or a culturally sensitive case–management training program to control diabetes and its risk factors in a Latino population with both type 1 and type 2 diabetes in a U.S. county compared with usual care only; 6) self-monitoring of blood glucose (SMBG) three times per day compared with no SMBG in type 2 noninsulin users; and 7) SMBG once per day compared with no SMBG in type 2 noninsulin users. One of the 18 specific interventions was cost-effective, i.e., the use of metformin to prevent type 2 diabetes in obese persons with impaired glucose tolerance compared with standard lifestyle intervention. No interventions in the “supportive” evidence category were “marginally cost-effective” or “not cost-effective.”
Current evidence is uncertain on how the CE of screening for undiagnosed type 2 diabetes would change with the age of those screened. Two studies evaluated the CE of screening for undiagnosed type 2 diabetes; one study reported that cost-effectiveness ratios (CERs) increased with initial screening age (16) while the other reported that they decreased with screening age (35).
Our systematic review showed that, with few exceptions, ADA-recommended interventions for preventing or treating diabetes and its complications were cost saving, very cost-effective, or cost-effective (i.e., with an ICER of less than $50,000 per QALY or LYG), although the strength of evidence varied. Generally, interventions that cost less than $50,000 per QALY are considered an efficient use of resources and worth recommending (11). Interventions with strong evidence for being cost saving, very cost-effective, or cost-effective should be considered for implementation. Interventions with supportive evidence for being cost saving, very cost-effective, or cost-effective should be adopted if extra resources are available or if similar interventions with strong evidence are unavailable or infeasible in a specific setting.
The one intervention recommended by the ADA that was shown as not CE was screening for type 2 diabetes of all U.S. residents aged 45 years and older. When considering allocating resources efficiently, universal screening for undiagnosed diabetes should be undertaken with great caution. The high CE ratio for universal screening for undiagnosed type 2 diabetes was primarily attributable to the small gain in health benefit. For example, screening everyone aged 45 years and older gained only 0.003 QALY per eligible person compared with no screening. However the additional costs associated with screening and early treatments were relatively large ($564 per person). Although detecting and treating diabetes earlier can prevent future diabetes-related complications and their associated medical costs, such savings are relatively small ($57 per person). Combining the health benefit and costs would yield an ICER of more than $1 million per QALY (35). An alternative to broad screening is to focus on screening persons with additional risk factors, such as hypertension. Such targeted screening is shown to be cost-effective when compared with no screening or universal screening.
Intensive glycemic control for all U.S. residents with type 2 diabetes diagnosed at age 25 years and older is marginally CE. However the cost-effectiveness of this intervention varies by age at the time of the diabetes diagnosis. The intervention is cost-effective in persons diagnosed at 25–54 years of age. However, intensive glycemic control for those diagnosed with diabetes at 55 years of age and older is not cost-effective. In fact, this result is consistent with the ADA's recommendation of less stringent A1C goals for patients with limited life expectancies.
The ADA recommended annual eye screening for diabetic retinopathy. This recommended intervention is very cost-effective compared with no screening in persons with type 2 diabetes. If considering the efficient allocation of resources, however, screening every other year might be a better alternative. Screening annually leads to a small health benefit but results in a moderate additional cost. For example, Vijan et al. (69) showed that, compared with a 2-year screening, annual screening among persons at moderate risk (65 years old with A1C level 9%) resulted in an increase of 2–3 days of sight at a cost of $540–690 per person. However the ADA also stated in its recommendation that “less frequent exams (every 2–3 years) may be considered following one or more normal eye exams.”
For the interventions with uncertain CE (including optimal age of starting screening for type 2 diabetes), following the current treatment guidelines may be the best option until more evidence on their CE is available.
The CEs of 43 ADA-recommended interventions were evaluated. Of these, 25 were in the “strong” evidence category. This number would probably have been larger if we had used less stringent criteria to define evidence as being strong. For example, evidence on the CE of using metformin to prevent type 2 diabetes among high-risk individuals was considered “supportive” in our current classification even though the efficacy of the intervention was shown by well-conducted multi-center large clinical trials in different country settings (71,72), and its CE was evaluated by “excellent” CE studies (25,34). This intervention was considered to have supportive evidence because it ranked lower in the ADA recommendations (7).
Among all the interventions considered, evidence for the CE of primary prevention through intensive lifestyle modification was the strongest regarding the quantity and quality of the CE studies and efficacy data. Several well-conducted clinical trials have shown the efficacy of intensive lifestyle modification in preventing diabetes in different country settings, such as the U.S. DPP (71), Finnish Diabetes Prevention Study (73), China Da Qing Diabetes Prevention Study (74), and Indian DPP (72). Eight cost-effectiveness studies (seven of them rated as excellent quality) have been conducted by different groups in different countries based on data from these well-conducted clinical trials (15,25,34,36,41,50,59,66). The results from these studies consistently showed that intensive lifestyle modification in persons with impaired glucose tolerance was cost saving or very cost-effective in the long run (15,25,34,36,41,50,59). Even in a short-term and one-on-one consulting setting, the intervention remained cost-effective (66). The intervention would be more cost-effective than existing studies show if the cost of the lifestyle intervention could be reduced. This might be achieved by changing the setting in which the intervention is provided. Only one study found a DPP-like intervention to be marginally cost-effective (25). Even in this study, however, the intervention would have been very cost-effective (23) if done in the type of group environment that is most likely in a real-world setting. A group-based, DPP-style lifestyle intervention partnership with the YMCA costs $275 to $325 per participant in the first year compared with $1,400 in the one-on-one setting of the DPP trial (75). Preventing diabetes, in particular by lifestyle modification, is not only effective but also a very efficient use of health care resources.
The CE of an intervention can vary by country setting. For example, intensive glycemic control (with a goal A1C level of 7%) in type 2 diabetic patients diagnosed at 25 years of age and older was marginally cost-effective in the U.S. but very cost-effective in other developed countries. Although the efficacy data of all studies of intensive glycemic control in type 2 diabetic patients were based on the same UKPDS data, the cost data were based on how residents of the different countries used health services and the cost of those services. The incremental cost of intensive glycemic control was much higher in the U.S. than in the U.K. because of different practice patterns. Patients outside the U.S. did not receive diabetes disease management services and had less frequent self-testing and physician office visits than their U.S. counterparts at the time these studies were conducted. If using the health services as described in the UKPDS setting but with the U.S. cost of these services, the CE of the intensive glycemic control in the U.S. would resemble that of other developed countries.
Future economic evaluation of diabetes interventions should consider the following. First, more studies are needed to evaluate the CE of interventions that fell in the “supportive” evidence category. For studies with weaker efficacy data, further efficacy studies are needed. Second, there are also 38 interventions recommended by the ADA but they have not been evaluated for their CE or the studies did not meet the inclusion criteria for our review (list is available upon request from the authors). The CE of these interventions should be assessed. Third, more CE studies are needed that address interventions in real-world settings. For example, few studies considered attrition rate, noncompliance, and dropout rates in evaluating CE. Fourth, more studies are needed to evaluate the CE of public policy changes. Only two studies evaluated public insurance reimbursement of ACEI therapy and both found this intervention to be cost saving. Finally, the CE of multiple interventions needs to be evaluated. In most real-world settings, patients receive multiple interventions simultaneously. Nearly all previous studies only evaluated the CE of a single intervention.
This review's conclusions should be used with caution. First, our conclusions are based on available information up to May 2008. More studies have been published since then. In addition, data on both the effectiveness and cost of an intervention could have changed since the time the original study was conducted. Using the newly available data could change our current conclusion. For example, in our review, we concluded that the CE of optimal age to start screening for type 2 diabetes was uncertain. A recently published CE study on age at initiation of screening for type 2 diabetes, released after our analysis was complete, might change that conclusion (76). Another example is the large decrease in costs for metformin, statins, and ACEIs. Studies that evaluate CE using current costs might look more favorably on interventions that include statins and ACEIs than those reported here. Reevaluating the costs and benefits of these interventions, using current-day costs, is beyond the scope of this study. Second, when using the results and conclusions of our review, readers need to be certain that terms are understood correctly. For example, “intensive insulin treatment” in our review meant “multiple insulin injection” or “insulin infusion.” Developments in medical technology might make continuous glucose monitoring systems, which record blood glucose levels throughout the day and night, more common. Drugs such as TZD Byetta and Gliptin, not available at the time covered by this review, are increasingly used to achieve intensive glycemic control. The CE of treatment with these and other new devices and drugs are unknown. New CE analyses are needed for these new interventions. Third, not everyone will necessarily agree with our classification criteria. Different classification criteria might have changed some conclusions. Fourth, most of the CE studies are based on simulation modeling. Although good-quality simulation modeling can provide information at a much lower cost than clinical trials, models are based on assumptions and represent a simplification of—and therefore might depart from—reality. Fifth, these CE studies use different methods, which could account for some differences in CERs. If the results from different models were consistent, we would have more confidence in the conclusion on the CE of the intervention. Sixth, we used the same threshold for the classification of the CE of interventions regardless of whether the ICERs were expressed as dollars per LYG or dollars per QALY, although they are different measures. The studies that reported costs per LYG did not incorporate the impact of the intervention on quality of life into the analysis. If they did, the cost per QALY could be higher, lower, or the same depending on the relative magnitude of the health benefit of the intervention on quality of life. Seventh, the interpretation of the CE of an intervention must include consideration of variables such as study population, comparison interventions, and country setting. Lastly, our recommendations are based on the CE of the interventions and not their efficacy; therefore, these recommendations are not necessarily the same as the ADA recommendations.
The importance of CE in decision making should not be overstated. CE is only one aspect to consider. CE analysis does not address the distribution of costs and the benefits of an intervention, societal or personal willingness to pay, social and legal aspects, or ethical issues associated with each intervention. All these aspects are important in formulating public policy. The good news is that our study shows that a majority of the recommended diabetes interventions provide both health benefits and good use of health care resources.
The authors conducted this project as part of their jobs as employees of the Centers for Disease Control and Prevention (CDC). The CDC is a federal agency in the U.S. government. The authors have no financial interest in this project.
Parts of this study were presented at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009, and at the Division of Diabetes Translation 2007 Annual Conference, Atlanta, Georgia, April 30–May 3, 2007.
We thank Drs. Sue Kirkman, Richard Khan, William H. Herman, John Anderson, Susan Braithwaite, Dan Lorber, and Vivian Fonseca for reviewing the earlier version of this manuscript and providing valuable comments. We also thank Elizabeth Lee Greene for her invaluable editorial assistance.
The findings and conclusions in this report are those of the authors and do not necessarily reflect the official positions of the Centers for Disease Control and Prevention.