pg79

These pages are best viewed with Netscape version 3.0 or higher or Internet Explorer version 3.0 or higher. When viewed with other browsers, some characters or attributes may not be rendered correctly.

A_1c Is Our Best Outcome Measure: Let's Use It
Matthew C. Riddle, MD, and Diane M. Karl, MD

These days we are all talking about measuring the outcomes of treatment for diabetes. The pressure to do so is strong. Our patients want and deserve the best, and more than ever before we are asked to justify the costs of medical care by proving its benefits. But how should we measure the effectiveness of our efforts for diabetes?

Various recommendations, guidelines, and criteria for audit have been proposed. Unfortunately, they are mostly unappealing and ineffective.

The American Diabetes Associations (ADA) most important statement of this kind is the Standards of Medical Care for Patients with Diabetes Mellitus.1 This is a fine summary of current tactics, but it is so detailed that few clinicians, whether primary physicians or endocrinologists, have the time and motivation to adhere strictly to it at every encounter with a patient. Moreover, it contains relatively few objective items that lend themselves to audit by record review. Not surprisingly, these standards are not applied as widely as intended.

At the other end of the scale, the Health Plan and Employer Data and Information Set from the National Committee for Quality Assurance (HEDIS) criteria include just one diabetes-specific measure: completion of a yearly dilated fundiscopic examination. While appropriate and auditable, this by itself cannot reflect the adequacy of diabetes care in general.

At present, there is no consensus on how to assess the effectiveness of care for diabetes in traditional or emerging systems of health care. Each group seems to be devising its own methods.

Part of our disarray stems from mixing different kinds of outcome measures. Three types of measures can be distinguished: process measures, medical events, and physiological measures.

Descriptions of physicians (and patients) actions are measures of the process of medical care. They are only indirectly related to good or bad medical outcomes, but are valued for their proven or presumed association with these outcomes. Examining feet and referring for nutritional assessment or education are examples. The great advantage of process measures is that they are easy to audit, hence their current popularity. Their disadvantages include favoring efforts to satisfy the auditor rather than help the patient, such as wasteful procedures done to minimize malpractice risk. Worse, they may not yield the information desired.

What we really want to know is the impact of medical care on the number of well, productive days in people's lives, and conversely, the rate of premature death and disability. In practical terms this means counting adverse medical events: hospitalizations, myocardial infarctions, strokes, amputations, auto accidents, transplants, cases of blindness, and the like. These are firm, objective measures, but collecting them reliably is difficult and costly. Also, these outcomes reflect medical care given years previously, and the information they offer comes too late to benefit the patients concerned.

Both process measures and studies of medical events have been disappointing so far. We need measures that are objective, diabetes-specific, and link present medical practices to future health outcomes, yet are easy to obtain and inexpensive.

Do we have any such measure? Actually, we do. It is a physiological measure: glycosylated hemoglobin. Evidence that elevation of glycosylated hemoglobin, now usually expressed as HbA_1c, correlates with and predicts bad medical outcomes is already extremely strong and grows steadily.^2-5 This should be no surprise, since the A_1c assay detects in blood cells a chemical process that occurs in most tissues and mediates much of the harm resulting from hyperglycemia.⁶ Laboratory methods for A_1c are now well standardized, and their cost is likely to decline. Office-sized machines are now available that measure A_1c in fingerstick blood in ⁵ minutes, with accuracy acceptable for most purposes. Test results are quantitative, easily entered into data handling systems, and thus easily retrieved and analyzed.

Why don't we routinely use A_1c to assess the quality and effectiveness of care? Why indeed? It's the best measure we've got. There is a strong case for using A_1c as our main measure of outcome, both for individual patients and for populations of patients. To consider this idea, first examine Figure 1.

Figure 1 shows the natural history of non-insulin-dependent diabetes mellitus (NIDDM), the most common form of diabetes, as it might be tracked by measurement of A_1c. The individual, in this case a woman, was born with a strong genetic vulnerability to NIDDM. HbA_1c, reflecting mean blood glucose, exceeded the normal range (4-6%) at age 20 in this case, with further increases during two pregnancies complicated by gestational diabetes. After age 35, her A_1c exceeded the 7% level, ADA's target for good control of diabetes.¹ At age 40, it climbed above 8%, the level at which further action to improve control is deemed mandatory. Her diabetes was diagnosed only at age 50, after 30, 15, or 10 years of hyperglycemia as defined by these three levels of A_1c.

The ages at which these milestones of glycemia are passed may be appropriate only for this one patient, but the pattern and the delay in diagnosis are typical. We know that half the people with diabetes in this country are undiagnosed.⁷ Figure 1 emphasizes the interval of hyperglycemia preceding diagnosis in which these undiagnosed patients are living. This long, hidden interval explains why so many patients already have diabetic complications when NIDDM is diagnosed.⁸

Figure 1. The natural history of NIDDM in a woman, as shown by her HbA_1c pattern over time. The normal range for HbA_1c (4-6%) is shown for reference. The broken lines indicate the 7% and 8% levels that the ADA has identified as the "taret" and "needs action" levels for managing diabetes.

At diagnosis, the patient changed her dietary and exercise habits. There was improvement of A_1c, followed by later failure of this approach. She then took a sulfonylurea, once again with initial success followed by secondary failure. With both lifestyle therapy and oral agents, years passed before the need to go on to the next form of treatment was obvious. By the time she began taking insulin, another two decades had passed, during which A_1c exceeded 8% most of the time.

Again, this woman's experience is lamentably typical. Population surveys commonly show the average patient under treatment for diabetes has an A_1c over 8%. For example, one medical outcomes study reports the mean total glycosylated hemoglobin for established patients to be about 9.5%, which corresponds to 8.3% A_1c.⁹ By the time insulin was started, she was quite old and likely burdened by diabetic complications. The horse was out of the barn, so to speak.

This pattern of late diagnosis, ineffective treatment, and too many complications must be changed. Fortunately, we now have more ways to treat diabetes, both insulin-dependent diabetes mellitus and NIDDM. Just as important, we have A_1c to measure success for both individual cases and for systems of health care. Here are three ways routine use of A_1c can help: 1) early diagnosis, 2) timely intensification of treatment, and 3) identification of cases needing special attention.

From a public health viewpoint, making the diagnosis of NIDDM earlier has a very high priority. Early intervention for NIDDM should improve glycemic control during the active years of mid-life, and thereby reduce the burden of diabetic complications later on. But making the diagnosis comes first, and random glucose, fasting glucose, and glucose tolerance testing have failed as ways to identify NIDDM in clinical practice.

Random glucose is far too variable, and high values are routinely ignored. Assessing fasting glucose is surprisingly difficult. Patients must come to the lab or office early in the morning, and often confound the test by late evening meals or early morning coffee with milk. Glucose tolerance testing is even more problematic. It requires fasting, an early appointment, several hours, some discomfort, and considerable cost. Whether because of the shortcomings of these tests or for other reasons, we have not been diagnosing NIDDM early enough.

Whether A_1c measurement should replace the other ways to make the diagnosis is now under debate. Change is slow, but A_1c should prevail. It's an inexpensive, nonfasting test, potentially using a single fingerstick blood sample. It doesn't correlate perfectly with an oral glucose tolerance test, but neither does a repeated oral glucose tolerance test. In contrast to a tolerance test, A_1c identifies the patient's actual glycemic experience instead of a glycemic profile after an unphysiological stimulus. A_1c over 7% could be used to identify cases of NIDDM. With this conservative standard, well above the upper limit of normal (6%), false diagnoses would be rare. In doubtful cases, glucose tests could be used for confirmation.

Aside from improving medical care for individuals with NIDDM, figures from routine A_1c screening could be used to assess the quality of services offered by a health-care system. For example, a low rate of diabetes reported in a system's clientele, such as the 2–3% value often cited, should be regarded as a red flag. It suggests continued underdiagnosis and nontreatment, since the actual percentage of diabetes in most populations is 5% or more. By contrast, a high rate of diagnosed diabetes in a health system might indicate that diabetes is taken seriously.

The most valuable part of the ADA Standards may be the recommendations regarding use of A_1c for monitoring patients with diabetes. They include routine measurement of A_1c for all patients, with measurements at least every 3 months for patients treated with insulin. The results evaluate the success or failure of the present treatment, and allow timely increases of dosage or addition of new therapies. The target levels, as mentioned above, are quite specific: 7% is the desired level, and above 8% requires a change of tactics. Adherance to these recommendations should reduce the years of persistent hyperglycemia without medical action that are currently routine.

If A_1c were universally tested in a health system, the results would reflect the whole systems performance, as well as each patient's status. The mean A_1c in patients with diabetes, stratified by duration of diabetes and severity of illness, could be compared between physicians, between clinics, or between health maintenance organizations. Values averaging over 8% would reveal ineffective treatment of the population studied. Lack of adequate A_1c follow-up data would imply inattention and probably poor glycemic control. Such data may predict later medical events and decline of functional status better than the process measures that now dominate medical audits. The effectiveness of A_1c versus other measures as predictors of medical events should be tested prospectively.

Diagnosis and early treatment of diabetes lend themselves to standards and algorithms. Standard protocols for diabetes education and selection of initial medications, guided by A_1c and glucose testing, can be implemented throughout a primary care system. Many patients should respond well to standard medication regimens with follow-up testing. If the average A_1c is over 8% in a given diabetic population, less than half the patients have values in the acceptable range, and there's a lot of room for improvement.

One kind of systematic intervention is described by Mazze and colleagues.¹⁰ The figures given in this report are 10.2% A_1c before intensification of management in a primary care setting, and 8.8% 6 months later. This corresponds to improvement from about 25-40% of patients adequately controlled. With more time, more patients may achieve acceptable control with algorithm-based management, though no one knows just how many.

However, experience has shown that managing diabetes is not easy. If a primary care system could demonstrate as much as 60% success with a cross-section of diabetic patients, it would be an impressive achievement. That would leave 40% of patients with A_1c persistently over 8%. Regular measurement of A_1c would easily identify those patients. Here's where diabetes specialists can be put to best use. Diabetes units staffed by endocrinologists, educators, and dietitians whose careers are devoted to diabetes could provide primary or parallel services for patients not succeeding with standard management. Patients might also be sent to these centers because of severe hypoglycemia or repeated diabetes-related hospitalizations, but inability to keep A_1c below 8% should be the main, objective basis for referral.

An undetermined fraction, maybe 10% of patients with diabetes, will remain uncontrolled despite all efforts. These are patients with severe medical, psychosocial, or financial problems. The rest, perhaps 30% of patients with diabetes, are incompletely responsive to a simple approach, but potentially successful with more intensive and individualized services. By directing the right patients to these services, we may increase the percentage with acceptable control toward 90%. This concept also could be tested. The potential human and economic benefits justify doing so.

While Figure 1 shows the natural history of NIDDM that we want to change, Figure 2 shows an improved pattern we may hope to achieve by systematic management based on A_1c testing. After gestational diabetes as a warning sign, intensive lifestyle changes might prolong the time before A_1c exceeds 7%. Regular A_1c tests could identify the onset of asymptomatic NIDDM in young adulthood. Early and combined use of oral agents might keep A_1c below 8% for many years before insulin becomes necessary.

Figure 2. Possible improvement of the lifetime position of HbA_1c shown in Figure 1 by earlier diagnosis and more aggressive management of NIDDM.

The goal will be to maintain excellent glycemic control of NIDDM during the early part of its natural history, when it has often gone unrecognized or undertreated, and thus reduce lifetime medical disability and expenses. Patients who are unable to keep A_1c values consistently below 8% should be referred to diabetes units to determine why ordinary measures have failed and to provide more intensive treatment.

This scheme can be tested by structured health systems with varied medical resources and the ability to collect and analyze data for populations. In addition to its place in clinical research such as the NIDDM Primary Prevention Trial now underway, A_1c testing should be a routine part of health screening and audits of the effectiveness of care for diabetes.

1American Diabetes Association: Position Statement: Standards of care for patients with diabetes mellitus. Diabetes Care 19 (Suppl 1): S8-15, 1996.

2The DCCT Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term complications of insulin-dependent diabetes mellitus. N Engl J Med 329: 977-86, 1993.

3Klein R: Hyperglycemia and microvascular and macrovascular disease in diabetes. Diabetes Care 18: 258-68, 1995.

4Kuusisto J, Mykkanen L, Pyorala K, Laakso M: NIDDM and its metabolic control predict coronary heart disease in elderly subjects. Diabetes 43: 960-67, 1994.

5Andersson DKG, Svardsudd K: Long-term glycemic control relates to mortality in Type II diabetic patients. Diabetes Care 18: 1534-43, 1995.

6Schwartz JG: The role of glycohemoglobin and other proteins in diabetes management. Diabetes Reviews 3: 269-86, 1995.

7Harris MI: Undiagnosed NIDDM: clinical and public health issues. Diabetes Care 16: 642-52, 1993.

8Harris MI, Klein R, Welborn TA, Knuiman MW: Onset of NIDDM occurs at least 4-7 years before clinical diagnosis. Diabetes Care 15: 815-19, 1992.

9Greenfield S, Rogers W, Mangotich M, Carney MF, Tarlov AR: Outcomes of patients with hypertension and non-insulin-dependent diabetes mellitus treated by different systems and specialties. J Am Med Assoc 274: 1436-44, 1995.

10Mazze RS, Etzwiler DD, Strock E, Peterson K, McClave CR, Meszaros JF, Leigh C, Owens LW, Deeb LC, Peterson A, Kummer M: Staged diabetes management. Toward an integrated model of diabetes care. Diabetes Care 17(Suppl. 1): 56-66, 1994.

Matthew C. Riddle, MD, is head of the Section of Diabetes at the Oregon Health Sciences University, Portland. Diane M. Karl, MD, is co-director of the Diabetes Treatment Center at Providence Medical Center, Portland, Oreg.

Note of Disclosure: Dr. Riddle serves on an advisory board for Bayer Pharmaceuticals, a subsidary of which (Miles Diagnostics) developed an instrument for measuring hemoglobin A_1c levels.