The results of the decision tree are presented in a nomogram: A, the difference in the annual cost between any 2 strategies of nonsteroidal anti-inflammatory drug (NSAID) use; B, the difference in the annual rate of clinical upper gastrointestinal tract (UGI) events (symptomatic ulcer, perforation, bleeding, or obstruction) between any 2 strategies of NSAID use; and C, the incremental cost-effectiveness ratio, which is the additional cost per clinical UGI event prevented. An incremental cost-effectiveness ratio that incorporates the cost of clinical UGI events (in addition to the cost of the drugs) can be obtained by deducting the assumed cost of a UGI event.
A hypothetical example of how to use the nomogram pictured in Figure 1 (see "Results" section for details). Strategies with smaller incremental cost-effectiveness ratios are more cost-effective.
El-Serag HB, Graham DY, Richardson P, Inadomi JM. Prevention of Complicated Ulcer Disease Among Chronic Users of Nonsteroidal Anti-inflammatory DrugsThe Use of a Nomogram in Cost-effectiveness Analysis. Arch Intern Med. 2002;162(18):2105-2110. doi:10.1001/archinte.162.18.2105
Nonsteroidal anti-inflammatory drugs (NSAIDs) are associated with an increased risk of clinical upper gastrointestinal tract (UGI) events, namely, symptomatic ulcer, perforation, bleeding, and obstruction. Our objective in this study was to compare the cost-effectiveness of several strategies aimed at reducing the risk of clinical UGI events in NSAID users.
A decision tree model was used for patients requiring long-term treatment with NSAIDs to compare conventional NSAID therapy alone with 7 other treatment strategies to reduce the risk of NSAID-related clinical UGI events (cotherapy with proton-pump inhibitor, cotherapy with misoprostol, cyclooxygenase [COX]-2–selective NSAID therapy, or Helicobacter pylori treatment followed by each of the previous strategies, including conventional NSAID treatment, respectively). The outcome measure is the incremental cost per clinical UGI event prevented compared with conventional NSAID treatment over 1 year.
The use of a COX-2–selective NSAID and cotherapy with proton-pump inhibitors were the 2 most cost-effective strategies. However, the incremental cost associated with these strategies was high (>$35 000) in persons with a low risk of clinical UGI event with conventional NSAIDs (eg, 2.5% per year). If the baseline risk of clinical UGI events is moderately high (eg, 6.5%), using a COX-2–selective NSAID becomes the most effective and least costly (dominant) treatment strategy, followed closely by cotherapy with a daily proton-pump inhibitor. Because small changes in costs or assumed efficacy of these drugs could change the conclusions, the incremental cost-effectiveness ratios between any 2 strategies were presented in a nomogram that allows the flexible use of a wide range of values for costs and rates of clinical UGI events.
The risk of clinical UGI events in NSAID users depends on their baseline risk, the added risk associated with the individual NSAID, and the protection conferred by cotherapy. A nomogram can be used to incorporate these factors and derive estimates regarding cost-effectiveness of competing strategies aimed at reducing the risk of clinical UGI events.
NONSTEROIDAL anti-inflammatory drugs (NSAIDs) are commonly used to relieve symptoms of arthritis and soft tissue inflammation. It is estimated that in 2000, there were more than 111 million NSAID prescriptions filled in the United States at a cost of approximately $4.8 billion.1 The NSAIDs have been associated with an increased risk of clinical upper gastrointestinal tract (UGI) events, namely, symptomatic ulcer, bleeding, perforation, and obstruction.2,3 The frequency of these complications in average-risk persons is relatively low (1%-5% per year of drug use).3- 8 Several risk factors increase the risk of clinical UGI events associated with NSAID use. These include older age, history of peptic ulcer disease, use of anticoagulants or steroids, and possibly Helicobacter pylori infection.6,9- 16
Various measures have been used to reduce the risk of clinical UGI events in NSAID users.4 Medical cotherapy with high-dose histamine2 receptor blockers, proton inhibitors, or misoprostol reduces the incidence of endoscopic duodenal and gastric ulcers.6,17- 21 Misoprostol, in addition, reduces the risk of NSAID-associated clinical events.6 Recently, treatment with cyclooxygenase-2–selective NSAIDs (coxibs) have been shown in double-blind, randomized, controlled trials to be associated with a significant reduction in frequency of clinical UGI events compared with conventional NSAID therapy.7,8 Finally, treatment and eradication of H pylori among infected individuals might reduce the frequency of complications related to peptic ulcer disease.22
In the presence of several effective strategies that reduce the number of NSAID-related UGI clinical events, cost becomes an important factor in choosing the optimal strategy.23 Several cost-effectiveness studies have addressed this issue.4,24- 27 Typically, cost-effectiveness studies present their results for a set of assumptions regarding the magnitude of risk of clinical UGI events, the reduction in frequency of complications associated with the use of a certain strategy, and the cost of that strategy. Difficulty arises when one attempts to apply results of published cost-effectiveness studies to an individual case because (1) the risk of developing clinical UGI events in individual NSAID users ranges between 1% in low-risk patients and 50% in patients with several risk factors; (2) costs differ among hospitals, pharmacies, and countries, and costs of drugs change over time; (3) the estimates for the efficacy of the strategies outlined above in reducing the number of clinical UGI events may vary significantly in the published literature; and (4) most cost-effectiveness studies predate the introduction of coxibs.
In the present cost-effectiveness study, we have presented the results in a nomogram that enables physicians to tailor the treatment strategy to each patient's specific circumstances and to compare the cost-effectiveness of any 2 strategies aimed at reducing the risk of clinical UGI events. The values in the nomogram include a wide range of possibilities for the risk of clinical UGI events, the assumed efficacy of the treatment strategy, and the cost of the drug and/or strategy. We provide several clinical examples to illustrate the use of the nomogram as an aid to clinical decision making.
A decision tree model was created using Excel (Microsoft, Redmond, Wash)28 to determine the most cost-effective strategy for reducing the risk of clinical UGI events (symptomatic ulcer, perforation, bleeding, and obstruction) in NSAID users under base case assumptions. A time frame of 1 year was used, and therefore we did not adjust dollar amounts to reflect the time value of money by assigning lower values to dollars paid in the future than to dollars paid in the present. The perspective of the analysis is that of a third-party payer.
The base case scenario involves a 55-year-old person with osteoarthritis who requires NSAID therapy for 1 year. The overall risk of clinical UGI events, which is assumed to be 2.5%, equals the baseline risk combined with the risk conferred by conventional NSAID intake.4 The following alternative strategies of NSAID treatment were compared: (1) conventional NSAID (800 mg of ibuprofen 3 times daily—this is the conventional NSAID used throughout the present study); (2) conventional NSAID plus a single dose of regular-strength proton-pump inhibitor (PPI; 30 mg of lansoprazole daily—this is the PPI used throughout the present study); (3) conventional NSAID plus 200 µg of misoprostol 3 times daily; (4) coxib (100 mg of celecoxib twice daily); and (5) through (8), bismuth subsalicylate/metronidazole/tetracycline combination (Helidac; Prometheus Laboratories Inc, San Diego, Calif) and PPI treatment for H pylori twice daily for 2 weeks followed by each of the first 4 treatment strategies, respectively.
These treatment strategies were assumed to have similar anti-inflammatory efficacy and a similar degree of osteoarthritis pain relief. The 3 main outcomes of all the strategies were the total number of clinical UGI events, the total cost of drugs, and the total cost of the strategy when the cost of a clinical UGI event is considered. The probability of clinical UGI events associated with each strategy and cost of drugs is outlined in Table 1. We also calculated the incremental cost-effectiveness ratios between strategy 1 (reference strategy, conventional NSAID therapy alone) and each of the other 7 strategies. The incremental cost-effectiveness ratio is the extra cost incurred by using an alternative strategy to reduce a single clinical UGI event. For example, the incremental cost-effectiveness ratio between strategy 2 and strategy 1 is calculated as the cost of strategy 2 minus the cost of strategy 2 divided by number of clinical UGI events prevented in strategy 2 minus the number of clinical UGI events prevented in strategy 1.29
All the assumptions of costs and probabilities were made for the 1-year time frame. The assumptions concerning the incidence of clinical UGI events in different settings were derived from published literature, and in cases of significant differences in these estimates, the opinion of an expert (D.Y.G.) on the published literature was used (Table 1). The costs of drugs were reported from the Red Book30 for 1999 (Table 1) and the cost of clinical UGI events from Laine.4
The 3 variables included in the sensitivity analyses were (1) the baseline risk of clinical UGI events with conventional NSAID use alone; (2) the risk of clinical UGI events with other strategies of NSAID therapy; and (3) the cost of the drugs. The overall risk of clinical UGI events equals the baseline risk of clinical UGI events with conventional NSAID use multiplied by the risk reduction conferred by using alternative NSAID strategies. Incremental cost-effectiveness ratios were calculated for a wide continuous range of possibilities for risk of clinical UGI events (2 variables) and the cost of the drugs (1 variable). A 3-way sensitivity analysis is presented as a nomogram. The overall difference between any 2 strategies in the risk of UGI events varied between 0.1% and 5% while the difference in cost varied between $10 and $2000 (cost in 1 year).
A wide range of values for the difference in clinical UGI events, difference in cost, and incremental cost-effectiveness ratio was presented. The functional relationship z = x/y can be linearized by a transformation of variables as log(z) = log(x) − log(y). Hence, depiction of this relationship by a nomogram requires an exponential scaling of the axes representing the values of x, y, and z (the first 2 increasing and the third decreasing). Once the position of 1 labeled value for each of 2 of these axes is chosen (eg, x and y) and the length of a doubling interval on 1 of these 2 is chosen, the exact scalings for all 3 axes are uniquely determined.
To facilitate the estimation of the overall risk of UGI events in a given individual, we have provided risk estimates associated with older age and history of peptic ulcer disease, H pylori infection, use of steroids, and use of anticoagulants (Table 2). These estimates were based on our interpretation of published literature. The effect of these variables was assumed to be additive (ie, no interactions). However, different assumptions can be substituted by the user within the ranges presented in the nomogram.
Table 3 outlines the annual cost when only the cost of the drugs used in the 8 treatment strategies is considered. These estimates are presented for the base case assumption of 2.5% annual risk with the use of conventional NSAIDs. The same table includes data on the frequency of clinical UGI events prevented and the incremental effectiveness (in reducing clinical UGI events) between strategy 1 and all alternative strategies. The cost-effectiveness is expressed as the incremental cost-effectiveness ratio between strategy 1 and each of the alternative strategies. The cost of strategies was the lowest for a conventional NSAID (strategy 1) and highest for a combination of a conventional NSAID and a daily PPI (strategies 2 and 6). The frequency of clinical UGI events was highest for conventional NSAID (strategy 1) and equally low for a combination of conventional NSAID and a daily PPI (strategy 2) or a coxib (strategy 4). However, the incremental cost-effectiveness ratio was lowest for coxib (strategy 4) and highest for a combination of conventional NSAID and PPI (strategies 2 and 6).
Changes in the baseline risk of clinical UGI events resulted in large changes in the estimated effectiveness and cost-effectiveness measures. For example, if the baseline risk of clinical UGI events is moderately high (6.5%), the incremental cost-effectiveness ratio is lowest with strategies 4 and 5 (ie, most cost-effective in reducing clinical UGI events). This example is illustrated in the right 3 columns of Table 3.
Table 4 outlines incremental cost-effectiveness ratios for the different strategies when the cost of clinical UGI events is considered in addition to the cost of the drugs (ie, different values from Table 3). At an assumed cost of $28 000 for each clinical UGI event, the incremental cost-effectiveness ratio was $69 600 between strategies 1 and 3 and $35 200 between 1 and 4. In other words, adding misoprostol to a conventional NSAID treatment regimen (strategy 3) prevents 1 clinical UGI event at an additional cost of $69 600, while using coxib (strategy 4) instead of a conventional NSAID prevents a single UGI event at an additional cost of $35 200.
The estimates for incremental cost-effectiveness ratio in Table 4 differ from those in Table 3 by a fixed value of $28 000, which is the assumed cost of a clinical UGI event. Therefore, we chose to present the cost of the drugs (rather than the cost of the strategy) in the sensitivity analyses.
The nomogram (Figure 1) consists of 3 vertical lines representing (A) the difference in the annual cost of the drugs, (B) the difference in the annual risk of clinical UGI events between any 2 strategies, and (C) the incremental cost-effectiveness ratio, or additional cost per clinical UGI event prevented between the strategies. Estimates for risk factors of clinical UGI events among NSAID users based on the published literature are presented as fold increase in risk in Table 2. These estimates can be used to calculate the value in Figure 1 B (difference in risk of clinical UGI events).
The following example serves to illustrate the use of the nomogram in comparing the cost-effectiveness of 2 competing strategies for NSAID use in reducing clinical UGI events (Figure 2). A 50-year-old man with a history of bleeding peptic ulcer disease has a 25% annual risk of a clinical UGI event while undergoing therapy with 375 mg of Naprosyn (Roche Pharmaceuticals, Nutley, NJ) 3 times daily (2.5% baseline risk with conventional NSAID × 10). Naprosyn is assumed to cost $654 per year. The same person has a 12.5% risk of a clinical UGI event if Naprosyn is combined with 20 mg of omeprazole daily (an assumed 50% reduction). At health maintenance organization (HMO) A, the additional cost of omeprazole is $1393 per year (Figure 2 A), and the difference in the annual rate of clinical UGI events is 12.5% (Figure 2 B); therefore the incremental cost-effectiveness ratio is $11 144 (calculated by connecting the marks on Figure 2 A and Figure 2 B and extending the line to intersect with Figure 2 C). Alternatively, this person could be given a coxib, 100 mg of Celebrex (G.D. Searle & Co, Skokie, Ill) twice daily, at an annual cost of $1029 (Figure 2 A) with the same reduction in clinical UGI events (12.5% on Figure 2 B) and therefore an incremental cost-effectiveness ratio of $9700. If the same patient receives his care at a different HMO (B) where omperazole is offered at a lower annual cost of $900 (Figure 2 A), then the calculated incremental cost-effectiveness ratio becomes $7200 (Figure 2 C), and adding omeprazole to the regimen becomes a more cost-effective strategy. Similarly, if the user disagrees with the 50% reduction in clinical events with coxib and assumes a 70% reduction rate instead, the point on Figure 2 B changes to 19.3%, and the incremental cost-effectiveness ratio will be calculated at $5322.
The nomogram in Figure 1 includes only the annual cost of drugs. An incremental cost-effectiveness ratio that incorporates the cost of clinical UGI events can be obtained by simply deducting the assumed cost of a UGI event from the incremental cost-effectiveness ratio in the nomogram. For example, if the cost of a clinical UGI event in the previous case scenario is $15 000, then the incremental cost-effectiveness ratio of omeprazole (at HMO A) is $3573. In this case, the coxib strategy becomes the most effective and least costly (dominant) strategy.
Three variables affected the cost-effectiveness estimates for reducing clinical UGI events in long-term NSAID users: (1)the cost of drugs, (2) the baseline risk of clinical UGI events with conventional NSAID therapy, and (3) the degree of protection conferred by alternative strategies. We have presented the results of this model in a user-friendly nomogram. In constructing the nomogram, we used a wide range of probabilities for these influential variables that allow the user to obtain an instant estimation of the incremental cost-effectiveness ratio of reducing clinical UGI events between any 2 competing strategies of NSAID treatment (the additional cost needed to prevent 1 event). The incremental cost-effectiveness ratio between strategies was the outcome of the model and the nomogram. The incremental cost-effectiveness ratio is an important measure of cost-effectiveness29; in the present study it refers to the additional cost of alternative strategies to avoid 1 clinical UGI event compared with the use of conventional NSAID therapy.
We assumed that once a clinical UGI event, for example a bleeding ulcer, develops in an NSAID user, the clinical course is similar regardless of the initial strategy of NSAID use. Therefore, for practical purposes, all the "downstream" events (hospitalizations, endoscopy, consultations, or death) and their associated costs will be determined by the probability of developing a clinical UGI event. We have shown that the calculated incremental cost-effectiveness ratios follow the same trend whether we use only the cost of drugs or the overall cost of the strategies including the cost of UGI events. The costs of drugs are obtained more easily and accurately and will provide the user with readily calculated incremental cost-effectiveness ratios. However, to obtain an incremental cost-effectiveness ratio that incorporates the cost of clinical UGI events, these costs could be simply deducted from the incremental cost-effectiveness ratio obtained in the nomogram. The cost of clinical UGI events ranges between $20 000 and $30 000 and could vary greatly between practitioners, institutions, and countries.25,26
This model has several advantages. It can be used as a simple cost-minimization model to determine the least expensive drug(s) to prevent a clinical UGI event in NSAID users. Most important, however, the model is constructed to make this determination for patients with a wide range of baseline risks for clinical UGI events. The attractiveness of this method of presentation is in its universality: it can be applied to existing drugs, new drugs, and combinations of drugs. In addition, while the decision tree uses specific predefined estimates based on our interpretation of published literature, users of the nomogram can enter whatever assumptions they wish based on their own interpretation of the literature and cost estimates in their own practice settings. The relative simplicity of the model in having only 2 predicting variables has made the nomogram presentation possible. The presence of more than 2 predicting variables would have made the linear presentation impossible. Therefore, this type of presentation may not be applicable to more complex models.
The overall finding of this study is rather straightforward: it is cost-effective to use relatively expensive medications such as coxibs or to add a PPI to regimens for patients with a high risk for clinical UGI events. Including the cost of clinical UGI events in the analysis strengthens the argument further. Even in persons with an intermediate risk of clinical UGI events with conventional NSAIDs (6.5% per year), the use of coxibs becomes a more effective and less costly strategy than conventional NSAID therapy. We assumed in this model that coxibs were slightly less expensive than adding a PPI to a conventional NSAID regimen, but these costs are similar, and it is possible that in some practice settings a certain PPI would be less expensive. The use of the nomogram could determine the most cost-effective strategy in these settings.
Accepted for publication February 27, 2002.
Dr El-Serag is the recipient of Veterans Affairs Health Services Research & Development Career Development Award RCD00-013-2.
Corresponding author and reprints: Hashem B. El-Serag, MD, MPH, Department of Veterans Affairs Medical Center, 2002 Holcombe Blvd (152), Houston, TX 77030 (e-mail: email@example.com).