Screening to prevent renal failure in insulin dependent diabetic patients: an economic evaluationBMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7020.1595 (Published 16 December 1995) Cite this as: BMJ 1995;311:1595
- Bryce A Kiberd, associate professor of medicinea,
- Kailash K Jindal, associate professor of medicinea
- Correspondence to: Professor Kiberd,5077 RC Dickson Building, Victoria General Hospital, Halifax, Nova Scotia, Canada B3H 2Y9.
Objective:To examine the conditions necessary to make screening for microalbuminuria in patients with insulin dependent diabetes mellitus cost effective.
Design:This economic evaluation compared two strategies designed to prevent the development of end stage renal disease in patients with insulin dependent diabetes with disease for five years. Strategy A, screening for microalbuminuria as currently recommended, was compared with strategy B, a protocol in which patients were screened for hypertension and macroproteinuria.
Intervention:Patients identified in both strategies were treated with an angiotensin converting enzyme inhibitor
Main outcome measures: Strategy costs and quality adjusted life years (QALYs).
Results: The model predicted that strategy A would produce an additional 0.00967 QALYs at a present value cost of $261.53 (1990 US$) per patient (or an incremental cost/QALY of $27041.69) over strategy B. The incremental cost/QALY for strategy A over B was sensitive to several variables. If the positive predictive value of screening for microalbuminuria (impact of false label and unnecessary treatment) is <0.72, the effect of treatment to delay progression from microalbuminuria to macroproteinuria is <1.6 years, the cumulative incidence of diabetic nephropathy falls to <20%, or >64% of patients demonstrate hypertension at the onset of microalbuminuria, then the incremental costs/QALY will exceed $75000.
Conclusions: Whether microalbuminuria surveillance in this population is cost effective requires more information. Being aware of the costs, recommendation pitfalls, and gaps in our knowledge should help focus our efforts to provide cost effective care to this population
Treatment with angiotensin converting enzyme inhibitors at the start of microalbuminuria must delay progression to macroproteinuria by on average 1.6 years to be cost effective
The positive predictive value of the test is important; physicians must also avoid inappropriate testing
Twice yearly testing for microalbuminuria after the start of treatment with angiotensin converting enzyme inhibitors is probably not cost effective
Diabetic nephropathy is the leading cause of end stage renal failure in many countries.1 2 Several studies have demonstrated the importance of control of blood pressure in delaying progression to end stage renal failure.3 4 More recently captopril, an angiotensin converting enzyme inhibitor, was shown to have the additional benefit of a delay in progression in patients with overt diabetic nephropathy.5 There are many reports showing that most patients pass through a stage of “incipient” nephropathy (microalbuminuria) before developing overt diabetic nephropathy.6 7 8 More recently treatment with an angiotensin converting enzyme inhibitor at the stage of microalbuminuria has been shown to prevent the progression to clinical proteinuria.9 Ideally if patients destined for diabetic nephropathy could be identified and treated early to delay progression, valuable health care savings and improvements in patient outcomes could be realised. Whether microalbuminuria is a good predictor is controversial.10
Two cost effective analysis studies support screening for microalbuminuria.11 12 These models, however, assumed a high incidence of diabetic nephropathy in the population, incorporated high efficacy rates of treatment to prevent end stage renal disease, and did not specifically examine the effect of false positive results of microalbuminuria on cost effectiveness. This study examines these variables and also examines whether screening for microalbuminuria (>20 £g/minute) and treatment with an angiotensin converting enzyme inhibitor is cost effective compared with a strategy not requiring microalbuminuria testing.
This economic evaluation compared two strategies for the follow up and treatment of patients with insulin dependent diabetes mellitus. Strategy A consisted of screening for microalbuminuria in patients with more than five years of diabetes and treatment with an angiotensin converting enzyme inhibitor (equivalent to captopril 25 mg thrice daily) if two out of three tests were positive (>20 £g/minute).5 13 Strategy B consisted of treating patients with hypertension or macroproteinuria, or both (dipstick >0.3 g/l or positive Albustix confirmed with >300 mg/day or >200 £g/minute proteinuria). A Markov model incorporating relevant outcomes for a cohort of patients with insulin dependent diabetes was created based on previous models.11 12 Strategy A (fig 1) is similar to these previously published models. A portion of the patients destined for diabetic nephropathy will be hypertensive at the onset of microalbuminuria.14 15 Therefore a portion of patients in strategy B will be treated with angiotensin converting enzyme inhibition at the same time as in strategy A without the costs of screening for microalbuminuria. The remaining patients destined for diabetic nephropathy in strategy B will either develop hypertension sometime thereafter and be started on the drug or develop macroproteinuria and be treated. Patients in strategy B destined to diabetic nephropathy without concomitant hypertension are assumed to progress from microproteinuria to macroproteinuria at a faster rate (compared with strategy A) because of a delay in treatment (fig 1).
To determine which of these two strategies were superior, probabilities, costs, and utilities were assigned to each treatment and health state (tables 1 and 2). Direct costs from the perspective of the patient and provider (third party and government) were included. Utilities were used to calculate QALYs (cost effective and cost utility are used interchangeably in this report).16 Traditionally utilities vary between 1.0 (perfect health) and 0 (death). As there are no published utilities for our purpose, we determined utilities from 17 health care workers (four nephrologists, six house staff, six nurses, and one social worker) not associated with the study. Six states were presented to each subject in random order (present health; hypertension requiring medication; and insulin dependent diabetes alone, requiring blood pressure pills, with a functioning renal transplant, or on dialysis). Subjects first ranked and then assigned values to the states by using the time trade off method.17 Costs and outcomes were then calculated over a 60 year period. A sensitivity analysis was performed to calculate the threshold values for choosing strategy A over B. This report used guidelines recommending that a technology costing <$15000 US dollars (<$20000 Canadian) per additional QALY be readily implemented and that one costing>$75000 (>$100000 Canadian) be considered poor evidence for adoption.18
The baseline model predicted the life expectancy for a diabetic subject from the time of diagnosis for strategy B would be 39.56 years. For this strategy a patient was also predicted to accure 12.51 QALYs from the time of screening. Strategy A produced an additional 0.1287 years but only 0.0097 more QALYs (3.5 days). Only 0.039 of the additional undiscounted life years produced accrued in the first 30 years. A physician following the recommendations of strategy A must screen and treat 103 (1/0.0097) patients over a 60 year period to produce an additional QALY. From a different perspective he or she must screen and treat 423 patients for 10 years to prevent one year of dialysis. The incremental cost/QALY of A over B was $27041.69. As costs differ widely from centre to centre a wide range of values were examined (table 3). Costs for screening and treatments with drugs and for renal failure as well as the effect of drug treatment on quality of life had important effects; most changes, however, maintained the incremental cost/QALY in an acceptable range.
Screening tests for microalbuminuria have a high coefficient of variation and can yield positive results in normal subjects.14 19 20 The potential to diagnose patients falsely with incipient nephropathy will lower the positive predictive value of the test. The sensitivity and specificity of the test are reported to be high (>0.90); the precision (wide confidence intervals) of these estimates, however, is low.14 With an annual incidence of microalbuminuria of 1.7% the positive predictive value for a single test (0.97 sensitivity and 0.96 specificity) is 0.3.21 To what extent repeat testing increases the positive predictive value is unclear (not likely to exceed 0.8). If the predictive value is low many patients will be treated needlessly. The positive predictive value must be >0.72 if strategy A is to be preferred (fig 2). Lower screening costs with poor predictive value or more frequent testing (at higher costs) with modestly better predictive value will both result in higher costs/QALY. Even if annual drug costs drop to $250, strategy A will not be cost effective if the predictive value falls below 0.65.
The Collaborative Study Group showed that captopril slows progression of diabetic nephropathy in patients with renal impairment.5 There was no evidence that earlier treatment had an additional effect. The baseline model states that treatment will delay the transition from microalbuminuria to macroproteinuria by two additional years. For strategy A to be preferred, delayed treatment under B must result in premature macroproteinuria by >1.6 years (fig 3). If early treatment delays the transition by four additional years more QALYs are produced for less cost in strategy A.
A recent study reports that the cumulative incidence of diabetic nephropathy 25 years from initial diagnosis has fallen, from 30% to <9%.22 It is not clear whether this same change is being seen elsewhere. This model predicts a lifetime (25 year) cumulative incidence of 36.2% (15.7%) and 37.8% (16.8%) for strategy A and B, respectively. If the cumulative lifetime incidence falls below 20% (25 year to 8.1%) and all other assumptions hold, however, strategy A would not be cost effective (fig 4).
A number of patients will develop hypertension before or at the time of onset of microalbuminuria.14 15 Therefore screening for hypertension will already detect many patients destined for diabetic nephropathy and eliminate the need to screen for microalbuminuria in some patients. If the percentage of such patients with concomitant hypertension exceeds 64% then strategy A is not cost effective (fig 5).
Recently a report showed that patients with poor glucose control (haemoglobin A1c >8.1%) developed microalbuminuria early and frequently.23 These patients may well have proteinuria on the basis of poor control. If the rate of development of sustained “true” microalbuminuria (0.051) was three times our baseline value (0.017), however, even a lower positive predictive value of 0.65 would result in a cost/QALY of only $24310. The present recommendations also include follow up microalbuminuria testing twice a year. If this practice were eliminated without compromising efficacy the cost/QALY falls from $27041 to $18821.
The high incidence of renal failure in patients with insulin dependent diabetes and the high costs and consequences of end stage renal failure make screening an attractive technology. If screening and treatment costs are low then quick calculations show that even a short delay will be “cost effective.” Undeniably better blood pressure control and the additional benefits of captopril will forestall the progression in patients with overt nephropathy. Although not the purpose of this report, our model does show that initiation of treatment at the start of hypertension or macroproteinuria not only increases QALYs but also reduces costs compared with conventional treatment (data not shown). Screening for microalbuminuria, however, introduces variables which impact on cost effectiveness. These include balancing more distant future savings with upfront costs (discounting), balancing present health over future health, and estimating efficacy without the benefit of a randomised control trial. Two previous studies concluded that screening would be cost effective.11 12 The European study did not discount outcomes whereas the Joslin Clinic study did not assign utilities, and neither study considered the effect of false labelling, the fact that hypertension would reduce the need for screening in some patients, nor the impact of a change in the incidence of diabetic nephropathy.
This analysis shows screening to be cost effective if our baseline values and assumptions are correct. Screening for microalbuminuria will not be cost effective if these values and assumptions deviate significantly (and quite realistically) from baseline. From the above, research should be directed at the following.
Firstly, the effect of early treatment with angiotensin converting enzyme inhibitor (onset of microalbuminuria) to delay the onset of diabetic nephropathy and subsequent renal failure must be quantified. Simply stating there is a delay is insufficient.
Secondly, the positive predictive value and costs of the best microalbuminuria test should be determined. The positive predictive value will be much lower when testing is done annually (low incidence) than done in an unselected cross section of diabetic patients. At present several methods are being used. Some advocates suggest using two different screening methods with each patient visit.24 The trade off between higher screening costs (more testing) and better precision in diagnosis must be carefully considered (fig 2).
Thirdly, updated information on the cumulative incidence of diabetic nephropathy in North America and Europe should be collected.
Fourthly, the incidence of associated hypertension before, at the onset of, or just after the development of microalbuminuria should be determined.
Fifthly, the effects of treatment with angiotensin converting enzyme inhibitors on quality of life may be contentious. None the less, economic appraisals include this disutility in their analysis and further research in this area is warranted.25 26
Finally, the additional impact of non-compliance, inappropriate testing, and loss to follow up must be considered on the overall costs of screening. Large numbers of patients will be tested and either refuse treatment or be lost to follow up. Nearly half of all patients identified with microalbuminuria in a recent large study were lost to follow up.15 This increases screening costs without improving outcomes.
For those centres already screening there are several important recommendations. Firstly, to be cost effective screening must maintain a high positive predictive value. To reduce false positive results screening should be done in patients who are otherwise well.24 27 Poor glucose control, acute illness, high protein intake, urinary tract infection, and heavy exercise are causes of increased albumin excretion in diabetic patients who would not necessarily have incipient nephropathy. Despite best intentions and well known warnings, many tests are being inappropriately ordered.
Secondly, there is no evidence to support microalbuminuria testing in patients already on angiotensin converting enzyme inhibitors, those in whom the drug and other antihypertensive agents would be contraindicated or not considered, or those with macroproteinuria. Further research is needed to determine whether quantifying the effect of treatment on microalbuminuria is important. It is not known whether it is better to start treatment at a low dose and monitor or start at a higher dose and not retest. Given the high coefficient of variation many samples (at additional cost) would be needed to ensure confidence in monitoring.14 19 28
Thirdly, each centre should strive to minimise drug and screening costs and ensure compliance. Longer prescription orders, cheaper drugs, and careful testing are important. Finally, early angiotensin converting enzyme inhibition does not assure that a delay in progression will occur. There is some evidence that the response may also depend on blood sugar control, and good glucose control remains an important goal.29 30
Like many other promising screening strategies the evidence is not always available to support widespread implementation.31 32 None the less, there is some evidence to support screening for microalbuminuria in insulin dependent diabetic patients. What remains to be determined is at what level and how. With ongoing research more precision in the above analysis should be possible. Until then this analysis shows screening for low levels of microalbuminuria to be unproved and potentially costly. We hope this report provides a balanced perspective on screening for microalbuminuria and focuses the issues for practising physicians.
Conflict of interest None.