Current issues of ACP Journal Club are published in Annals of Internal Medicine

Quality Improvement

Self-monitoring of blood glucose was no more effective but more expensive than urine testing for improving glycemic control in NIDDM

ACP J Club. 1991 Jan-Feb;114:30. doi:10.7326/ACPJC-1991-114-1-030

Source Citation

Allen BT, DeLohng ER, Feussner JR. Impact of glucose self-monitoring on non-insulin-treated patients with type II diabetes mellitus. Randomized controlled trial comparing blood and urine testing. Diabetes Care. 1990;13:1044-50.



To compare the relative efficacy and cost of self-monitoring of blood glucose and urine glucose in managing patients with non-insulin-dependent diabetes mellitus (NIDDM) who were not treated with insulin.


Randomized, unblinded trial with monthly follow-up for 6 months.


Outpatient clinics at a university-affiliated Veterans Affairs hospital.


Patients had NIDDM (fasting plasma glucose level, 8.8 to 22 mM), no history of ketoacidosis, current treatment with diet with or without oral hypoglycemic agent, and no active infection or serious concurrent illness. 61 men (mean age 58 y) were randomized. 7 patients were not evaluated: 5 were inappropriately randomized (did not meet inclusion criteria) and 2 patients dropped out for unknown reasons.


Both groups initially received diet instruction from a dietician who emphasized increased fiber intake. The urine-testing group was expected to achieve negative urine glucose before meals. The blood glucose group was to achieve blood values < 7.7 mM fasting and < 8.8 mM before lunch and dinner. Physicians used an algorithm to adjust diet, exercise, oral agents, or insulin based on the patient's glucose profile.

Main outcome measures

Fasting plasma glucose, glycosylated hemoglobin, weight, cost, and compliance.

Main results

No statistically significant differences existed between groups in any outcome measure except cost. Within each group, significant improvements in fasting glucose (mean change 1.4 mM [SD 3.2 mM], P < 0.03, for the blood-testing group; mean change 1.5 mM [SD 2.8 mM], P < 0.01, for the urine-testing group) and glycosylated hemoglobin (mean change -2.0% [SD 3.4%], P < 0.01, for the blood-testing group; mean change -2.0% [SD 2.4%], P < 0.001 for the urine-testing group) were noted. Similar numbers in both groups achieved a 25% reduction in glycosylated hemoglobin or fasting plasma glucose, or both, during the 6- month study {44% in the blood-testing group vs 41% in the urine-testing group, 95% CI for the 3% difference, -29% to 22%, P = 0.78}.* The study had a 95% power to detect a 25% improvement in any of the major outcomes. Compliance, measured by glucose record keeping and clinic attendance, was similar between groups. Self-monitoring of blood glucose was estimated to be 12 times more expensive than urine testing in the first year, with purchase of a reflectance meter, then 8 times more expensive thereafter.


Self-monitoring of blood glucose was no more effective but much more expensive than urine testing in improving glycemic control in a closely supervised, comprehensive program.

Sources of funding: Veterans Administration Health Services Research and Development Service, and A. W. Mellon Foundation.

Address for article reprint: Dr. J. R. Feussner, Health Services Research Field Program, Veterans Affairs Medical Center, 508 Fulton Street, Durham, NC 27705, USA.

*Numbers calculated from data in article..


Allen and colleagues in this careful study found that patients improved control of their blood sugars when provided with a program of diet instruction and self-monitoring of blood or urine glucose. The finding that both forms of monitoring were equally effective is intriguing and could reverse the trend to encourage more patients with NIDDM to measure their blood sugars, a more painful and expensive method. However, like many good studies, the findings raise more questions that must be considered before self-monitoring of blood glucose is abandoned for patients with NIDDM.

Could the positive results be because of the extra attention that both groups received? Or could it be the fact that patients were selected because of their initially high blood sugars that would have come down anyway ("regression to the mean")? Perhaps more important, was self-monitoring of blood glucose "set up"? That is, what is the rationale for using this method when a patient's sugars are usually high enough to cause sugar to spill in the urine? Similarly, what is the usefulness of urine testing when the sugars are usually below the renal threshold but too high for good diabetic control? Unfortunately, there were not enough participants in this range to determine the relative usefulness of the 2 monitoring methods.

The investigators have provided a stimulating opportunity to think through the logic of home monitoring. While waiting for answers to the questions raised by their study, I will be a less enthusiastic advocate of self-monitoring of blood glucose for patients with NIDDM. Rather, I will ask my patients with NIDDM to monitor urine sugars before meals until the tests are usually negative and then ask them to switch to self-monitoring of blood glucose.

R. Brian Haynes, MD, PhD
McMaster UniversityHamilton, Ontario, Canada