Cost Modeling for Alternate Routes of Administration of Opioids for Cancer Pain

May 1, 1999

The economic considerations relative to neuraxial infusion can be looked at with different types of economic models, including cost-minimization, cost-effectiveness, and cost-benefit analyses. A theoretical predictive model

ABSTRACT: The economic considerations relative to neuraxial infusion can be looked at with different types of economic models, including cost-minimization, cost-effectiveness, and cost-benefit analyses. A theoretical predictive model was developed about 2 years ago using a computer spreadsheet and based on four levels of supportive data. The model shows that the breakeven point at which it becomes less expensive to administer opioids with an intrathecal/implanted pump, rather than an epidural/external pump, is between 3 and 6 months after the start of pain management. In addition, the breakeven point between systemic treatment and spinal delivery with an implanted system is between 1 ½ and 2 ½ years from start of pain treatment. [ONCOLOGY 13(Suppl 2):63-67, 1997]


The economic considerations relative to spinal infusion can be assessed with different types of economic models. What will be discussed in this article are cost-minimization analyses: how much does it cost to deliver a drug by one route vs another? This is the bottom line for the payers—assuming the outcomes are equal.

Most published data concern charges—what is charged to an insurance company or to a patient—not what the real cost is. Cost is a different issue. An example of cost would be how much money it costs a hospital to support an operating room, including the indirect cost of the operating room itself, the rent on the space, the upkeep, the cleaning of the space, the supplies, the salaries for the staff, etc. ‘‘Charges’’ is what is billed for an operation. Hopefully, there is a profit margin in those charges, although increasingly with managed care and discounted fees, the difference between cost and charges may in fact approach 0, or even a negative dollar amount.

Cost-utility analysis groups together all sorts of disparate and possible outcomes and end points, not just, for example, survival time or drug use, but a patient’s satisfaction, too. Cost-utility analysis basically asks: for the amount of money paid, what is the utility to the patient in an overall sense? Cost-utility analysis looks at patient satisfaction vs the cost of achieving that satisfaction.

Cost-effectiveness models look at the costs of different means to achieve one particular outcome, for example, tumor shrinkage. The means in this case could be radiation vs surgery vs chemotherapy. Cost-effectiveness models would evaluate the different associated charges or costs of these therapies to get to that one outcome.

Cost-benefit analysis offers more of a societal viewpoint, and might be more relevant to Medicare or government. What is the bottom line for society? What are the costs, and what are the benefits to society in terms of dollars? Using a noncancer example, what would be the cost of various treatments to get a sick person back to work? That would be the specific outcome, as in cost-effectiveness. The cost-benefit, however, would be the benefit to society of the person returning to work. An employed person would generate taxes for the government and more productivity for our country as a whole. So, cost-benefit analysis on that particular back-to-work issue is a much larger picture than just simply getting the patient back in his or her job.

The Bedder et al Study

The bulk of analyses relative to neuraxial infusion have been cost-minimization studies. The Bedder et al study from 1991, which is a bit dated now, is an example.[1] The Bedder et al study was a comparison of an externalized epidural catheter vs an implanted pump. The exteriorized Du Pen epidural catheter was placed in five patients with cancer pain who had a survival of 2.5 months. The SynchroMed pump was implanted in 15 patients, 7 patients with cancer pain who survived 5 months, and 8 patients with noncancer pain. Records of actual services provided and home care vendor quotations were reviewed, but these were charges not costs.

What were the charges to third-party payers for the two scenarios, given the disparate nature in terms of the patient mix and even the number of the patients in the two cohorts? With an external catheter, the costs incurred include a pump rental fee, drug fee—which is the cassette preparation and the drug-dispensing fee for that cassette—dressing supplies, and nursing visits. This raises the issue of how these patients are managed as outpatients. With an implanted pump, the costs incurred pertain to the drug itself, as well as the actual supplies, namely the pump, and the professional fee for implanting the device.

The Bedder et al study found that the initial cost for an implanted pump is about $15,000 and the initial cost for an externalized catheter placement is $9,000 (Table 1). The $9,000 cost could probably be lowered significantly depending on the approach used in placing the catheter and getting the patient in and out of the hospital. Monthly follow-up is the critical issue, however. In this particular paradigm in 1991, the charge associated with the monthly follow-up for care of an externalized epidural system was about $2,000 a month, as opposed to $273 a month for an implanted system.

Obviously it costs more money initially to implant the pump, although the upkeep of this implanted system is significantly lower than the upkeep of the externalized system. In such cases there will always be a breakeven point (Figure 1). If a patient survives and requires spinal delivery longer than that breakeven point, there would be less cost, or cost minimization, incurred with the implanted system.

Keep in mind, however, that these calculations are based on assumptions of how this externalized system is treated. This particular technique can be reduced in various ways to save money. The Europeans have found ways, whether it be epidural or intrathecal, of lessening some of the expenses. In many European countries, however, there are still charges associated, but they are hidden charges because they are assumed by the government. Nevertheless, money is being spent if somebody is actually doing work, even if it does not actually show up on a ledger and get charged to a hospital or to a patient. Such hidden costs need to be evaluated.

M. D. Anderson Theoretical Predictive Model

Bedder et al’s model was taken from actual vendor quotations of charges. The model that was put together at M. D. Anderson about 2 years ago is based on a theoretical modeling viewpoint.[2] We took a computer model—basically just a spreadsheet—and looked at actual charges for various items, then put these charges into the model and worked out issues of how they come together. This is really a theoretical predictive model based on actual charges for various entities.

There are five different parameters in this model:

initial charges for screening and start of opioid delivery;

ongoing drug charges in the follow-up period;

ongoing service in the follow-up period (such as nursing home health care);

professional fees for follow-up visits to the outpatient clinic; and

readmission or complication charges.

Relative potency can be evaluated across routes of administration (oral vs intravenous, epidural, intrathecal) and among different drugs—hydromorphone (Dilaudid), fentanyl (Duragesic, Actiq), and sufentantil. Even the fentanyl patch vs extended release morphine can be evaluated in this scenario. The potency of epidural delivery is approximately 10 times greater than that of intravenous, and that of intrathecal is 100 times greater than intravenous. This means that 10 times less drug is required by epidural administration vs intravenous, and 100 times less drug is required with intrathecal vs intravenous administration (Table 2).

Data to Support Model

This is a theoretical predictive model, but wherever possible, actual data were used to support the particular elements on a basis of levels, from the strongest data to the weakest. Level 1, the strongest level of support, is published material with median figures for geographical areas across the United States. For example, if you want to look at a Current Procedural Terminology (CPT) physician charge for a particular procedure, you can find manuals that have both the CPT code and a median, with various standards of deviation for charges for that particular CPT code across the United States.

There are a number of publications that list hospital charges, for example, what a standard hospital bed costs per day in the United States. The Red Book lists average pharmaceutical wholesale charges.[3] Likewise, there are published surveys that actually give some idea of what the commercial markup is for the wholesale price for drugs.

Level 2 supportive data include wholesale price quotations from various manufacturers. These price quotations do not summarize all data, but do provide information about what a specific device would cost. Level 3 data include surveys in multiple geographical areas from hospitals and national home health care agencies—asking hospitals or agencies around the country what they charge in various situations. At level 4, there are no published data, rather a best estimate based on the clinical experience of leaders in the field. This is at least useful information in the absence of other data.

Most of the information used to support this theoretical model, was obtained from level 1 and 2 data. There are few data from level 3, and rarely, information from level 4.

Relative Contributions of Cost Components

Relative contributions of different components of cost are important in comparing delivery routes of opioids. Relative contributions of components for oral morphine, transdermal fentanyl, subcutaneous or intravenous morphine delivery, epidural external pump, and intrathecal implanted pump are compared at 1 month in Figure 2.

For oral morphine, the drug accounts for slightly more than 50% of the total charge at 1 month. Home care is also significant, whereas clinic and readmission charges are perhaps less significant. Transdermal fentanyl, as you would expect, looks almost identical in terms of the cost breakdown. With a subcutaneous or intravenous delivery and an epidural external pump, however, the drug is less of a component of cost, and the home care percentage of charges at 1 month becomes more significant.

The initial charges for the epidural external pump are moderately significant. As in the Bedder et al model, with an intrathecal or an implanted pump, most of the charges at 1 month still represent the initial charges for the cost of the pump, with very small charges for drugs, clinic, home care, or readmission. The relative percentages obviously change with time.

To predict costs, one can use this theoretical predictive model, which is basically a spreadsheet, and plug in the average charge for a hospital bed in the United States per day; charge for SynchroMed pump; average markup across the United States for a pump; and average price for the drug, etc. Costs can be predicted for 1 month, 2 months, and 3 months. The initial charges stay constant; they are incurred at the time therapy starts and therefore stay fixed. But as length of treatment progresses, charges for the drug, for example, become more cumulative and these drug charges become a larger percentage of the cost pie.

Bottom-Line Approach

An example of a bottom-line approach may be informative: A dose of 10 mg per hour of intravenous morphine or its equivalent based on routes or drugs, or transdermal fentanyl, yields a fairly linear accumulation of charges over time. At almost all points evaluated up to about 25 months, on a pure cost-minimization level, oral morphine is going to be less expensive than any of the other options.

The expensive options are either intravenous or epidural externalized systems with morphine. Although epidural externalized systems are fairly inexpensive at the start, charges for home health care, the drug, the cassette, and the pump rental add up rather significantly. This results in a fairly rapid dose escalation over time. Using this particular paradigm at 15 months, a patient with an externalized system may have $100,000 or $120,000 of total charges related to the drug, the pump, home health care, and all other charges involved in his or her treatment.

The model in this case uses the same morphine preparation for epidural and intrathecal administration, regardless of whether the pumps are internalized or externalized. The resulting cost for the drug is trivial.

For the externalized pump, the expense is in the home health care issues and the pump rental and cassette. For the implanted pump, there is a fairly large initial charge—anywhere from $15,000 to $20,000 because of the device and the operation to implant it. The subsequent charges for home health care are virtually nil, however. Built into this cost model are some readmissions and a certain percentage of complications. Despite these factors, it is still fairly inexpensive for the upkeep at this point. The cost curves would cross if the treatment is continued long enough. In a noncancer patient, therefore, one could argue that for somebody who is on a fairly large dose of extended release morphine, strictly in terms of cost minimization, it would be cheaper to put the patient on an implanted pump if he or she will be receiving the drug for 2, 3, or 4 years.

Effect of Variations on Breakeven Point

What if the dose varied from 5 mg up to 20 mg per hour of intravenous morphine? The basic concepts and the breakeven points timewise are about the same. The cross-point between the external and implanted pumps is anywhere from 2 to 4 months, depending on how the parameters are varied.

The advantage of a theoretical model like this, as opposed to using actual charges, is that you can start asking some interesting questions. For example, if instead of $15,000 the intrathecal initial charges go as high as $30,000 to get a patient in and out of the hospital; if the pump is marked up 100%; your professional fees are high; your hospital charges are high—how much do any of these scenarios really change the breakeven point between the external and implanted pumps? According to this model, increasing initial charges from $15,000 to $30,000 changes the breakeven point from maybe 1 month about 3 or 4 months. It does not significantly change.

Similarly, the effect of an intrathecal equipment markup anywhere from 25% to 100% on the pump is trivial in terms of breakeven points.

Externalized pump per diem rates are actually the one component that may make a difference in breakeven points. There is a small variation in pump per diem rates, going from $100 to maybe $200, including the pump rental, and home visits by nurses. At the higher rates, the externalized system becomes very expensive over a time period.

For dose escalation, assume 10 mg per hour of intravenous morphine that is constant over the life of the patient. One can make a very simple assumption that there is an escalation in the dose of 10% a month, which says if one were taking 60 mg of oral morphine one month, the next month one would be taking 66 mg a month. That’s not such a large increase in clinical practice.

What is interesting is that on the theoretical model, the curves for the different routes of administration begin to go up exponentially (Figure 3). The breakeven point between intravenous or epidural with an externalized system vs an implanted system really does not change at all, but the breakeven points for the oral and transdermal routes do increase to about 14 months.

These curves do not imply that there is an indication on pure cost basis for an implanted pump vs oral morphine. This is so even with the more expensive oral drugs. With methadone, the oral route curve just drops down to almost a trivial amount.

The drug charge with an implanted pump is small. It is 1% or 2% of the total charge. You could triple that charge or cut it to one-tenth and you just would not see much of an effect on total charges. The drug charge for the epidural system is a very small percentage because most of the increase in charges comes from home health care. All of the drugs are being increased by 10%, but in terms of the whole total charge, they are just a drop in the bucket. So, if you had 2 drops in a bucket or 1 drop in a bucket, there’s not much difference.

Back Pain Syndrome Study

Another study from a slightly different angle looked at intractable pain from failed back surgery syndrome.[2] This is a noncancer setting, but it was the same concept as the M. D. Anderson theoretical predictive model—looking at direct cost of intrathecal morphine via implantable pump vs medical management alone. Various clinical scenarios and calculations for charges were evaluated in very much the same way.

With intrathecal morphine therapy, basically one looks at short-term complications and ongoing care, at the various charges and base case charges associated with delivery of the drug, and at the total expected cost. One of the interesting points about this particular model for back pain was to take a cost assumption of a base case, which is the average, and then look at 35% lower cost and 35% higher cost to get some idea of the sensitivity of these numbers.

For the back pain patient there is an intersection at about 20 months . This would suggest, if these are correct assumptions, that for a noncancer patient, at somewhere between 18 and 24 months on pure charge alone, it is cheaper to have an implanted pump than to go with systemic therapy.

Drugs are not really the big component of the charges here. The charges here are related to doctor visits, physical therapy, emergency room visits, and readmissions. The assumption is made that if you are getting better medical management or pain management, there is less supplemental therapy being used. That is a slightly different situation because it is in the noncancer setting and it actually includes a little bit of the effectiveness issue between the different routes. It is not a pure cost-minimization study there.


One can estimate costs by using models based on using actual charges, as in Bedder et al’s article, or by using base charges to do computer modeling. Some interesting figures resulted from the three different types of models reviewed—Bedder et al’s, the M. D. Anderson theoretical predictive model, and the back pain model. There appears to be a breakeven point between 3 and 6 months when it is less expensive to go with an implanted system than with an externalized system. In addition, if one compares systemic treatment vs spinal delivery with an implanted system, that breakeven point is 1½ to maybe 2½ years. That is probably not terribly significant on a pure dollar basis in the cancer setting, but it actually might be reasonable in a noncancer setting. This is not meant to suggest that patients are treated on the basis of what is the cheapest, but it is important to look at what the charges are when we do treat patients.

Faculty Questions

Elliot Krames, MD: I think that there have been some studies that show the average dose for morphine is about 200 to 250 mg. Your model supposed 600 mg or more. May we then extend your model from 18 months to probably 2½ to 3 years?

Samuel Hassenbusch, MD PhD: Actually that was looked at. With the model, we created four different graphs, from 20 mg per hour to 5 mg per hour and you can carry that down to 1 mg per hour of intravenous morphine or its equivalent. The point is that if you look at the breaks between externalized systems and implanted systems, it really does not change significantly. If you look at the oral opioids vs an implanted system, those breaks did not actually change significantly either, but they are still 1½ to 2 years and that is using MS Contin as opposed to methadone, which of course would drop that virtually to nil.

Dr. Krames: If cost was an issue, then clearly even at 600 mg per day oral morphine for at least 18 months or 2 years is cost-effective. Would you agree with that?

Dr. Hassenbusch: That’s true, although one has to quickly point out that it is not a trivial amount of money. Extended release morphine is not cheap.

Dr. Krames: The conclusion from your data is that at any rate at 2 months to 4 months, subcutaneous therapy or epidural therapy is not, in terms of cost, a viable option when you have to go from oral therapy to some other form of therapy. Would you say that?

Dr. Hassenbusch: Yes. All of the studies—Bedder et al’s plus the others presented here—show the breakeven point anywhere from 3 to 5 months no matter how you vary the numbers.


1. Bedder MD, Burchiel K, Larson A: Cost analysis of two implantable narcotic delivery systems. J Pain Symptom Manage 6(6):368-373, 1991.

2. de Lissovoy G, Brown RE, Halpern M, et al: Cost-effectiveness of long-term intrathecal morphine therapy for pain associated with failed back surgery syndrome. Clin Ther 19(1): 96-112, 1997.

3. Red Book-1994, Medical Economic Data. Montvale, New Jersey, 1994.