Patient subgrouping may be necessary for identifying the individuals eligible for a new drug as well as for estimating changes in condition-related costs. This may include a breaking out of the incident and/or prevalent population by age, condition severity, or history of previous treatment failure. As noted earlier, the new drug’s indication or restrictions on reimbursement or both might dictate this subgrouping.
Information for identifying patient subgroups by condition severity or based on a reimbursement recommendation may not be easy to obtain. The manufacturer of the new drug will generally have estimated the size of these subgroups, but these estimates may not be based on publicly available sources. As a result, these esti- mates might not be considered credible to use in the budget-impact analysis. The health technology assessment agencies or health plans may also estimate the size of the relevant population subgroups. However, these estimates also may need data to which access is limited.
In the case of reimbursement being restricted to only those with a specific level of condition severity or with a particular stage of disease (e.g., an indication limited to individuals with HIV infection with multiclass drug resistance and no remaining fully suppressive treatment regimens, individuals with multiple sclerosis with relapsing disease only, or individuals with chronic plaque psoriasis for whom con- ventional immunosuppressants have failed), information about the disease stage and/or treatment history may be taken from published epidemiology studies or esti- mated using disease-progression models. Types of studies that provide this informa- tion include large cross-sectional observational database and registry studies that estimate the proportion of patients at different disease stages (e.g., Pugliatti et al.
2006, for multiple sclerosis or Buist et al. 2008, for COPD) or long-term disease- progression modeling studies from which the proportion of time in different disease stages or on different lines of therapy can be estimated and applied to those with the indication of interest. Since these latter estimates might not be available in the pub- lished literature, the budget-impact analyst might need to develop a disease- progression model to estimate these values.
In Box 3.5, we present an example set of estimates of population prevalence by disease severity for COPD.
Box 3.5 Prevalence by Condition Severity
A budget-impact analysis is being developed for a new maintenance treatment in chronic obstructive pulmonary disease (COPD). As part of this budget- impact analysis, patients who are eligible for this new maintenance treatment need to be identified. Patients with COPD are typically categorized as having mild, moderate, severe, or very severe COPD, which is based on a patient’s lung function. Only patients with moderate COPD plus at least one exacerba- tion per year or those with severe or very severe COPD regardless of the number of exacerbations per year qualify for this new maintenance treatment.
An example of a common marketing indication that requires patient subgrouping and one in which information in order to identify those subgroups may not be read- ily available is the approval for a specific line of treatment where failure of earlier lines of treatment is required for eligibility. With this marketing indication, the num- ber or proportion of patients who may be eligible for second, third, and/or subse- quent lines of treatment must be estimated.
Consider a new drug that is approved for second-line treatment of an acute con- dition. One possible way to estimate the proportion of patients with a diagnosis and a history of failed first-line treatment would be to multiply the number of patients who receive treatment each year by 1 minus the success rate of the standard first- line treatment or 1 minus the weighted average of success rates for the mix of stan- dard first-line treatments.
To determine the size of the eligible population for the budget-impact analy- sis, we need to identify prevalence by COPD severity.
Buist et al. (2008) set out to estimate the prevalence of COPD in those aged
≥40 years using a population-based sampling plan and survey and spirometry testing before and after administration of 200 micrograms of salbutamol in a minimum of 600 participants (300 men and 300 women) in 12 sites across various countries around the world. In this study, they estimated the preva- lence of COPD by sex and disease severity and its risk factors.
This study provides consistent and credible estimates for COPD preva- lence by disease severity. Specifically, it reports a prevalence of 1.9% for no airflow obstruction, 1.1% for mild COPD, 1.4% for moderate COPD, and 1.0% for severe/very severe COPD within the USA site. If we needed to split the prevalence of severe/very severe COPD or if we needed to understand the prevalence of moderate COPD plus at least one exacerbation per year, we could combine these data with data from Hurst et al. (2010). Hurst et al.
(2010) performed a large observational study in which they examined the occurrence of exacerbations in patients with different levels of disease sever- ity. They reported that 39% of patients with moderate, GOLD stage 2 COPD had at least one exacerbation in the previous year. Using these data, we can estimate the prevalence of moderate COPD plus at least one exacerbation per year in the USA at 0.55% (=1.4% × 39.0%).
For estimating the prevalence of severe versus very severe COPD, we can use data from Hurst et al. (2010) to estimate the percentage of patients who have severe versus very severe COPD from the full study population. The numbers of patients in the study with severe and very severe COPD were 900 and 293, respectively. As a result, we have 75.4% (= 900/[900 + 293]) of severe/very severe patients have severe COPD and 24.6% (= 293/[900 + 293]) of severe/very severe patients have very severe COPD. In the USA, this translates to 0.75% (= 1.0% × 75.4%) with severe COPD and 0.25%
(= 1.0% × 24.6%) with very severe COPD.
S. Earnshaw and J. Mauskopf 46
For a chronic condition, the proportion of prevalent patients seeking active treat- ment and eligible for second-line or subsequent lines of treatment may be estimated by multiplying the condition prevalence by 1 minus the ratio of the mean time on first-line treatment to either the mean life expectancy or the mean total time on active treatment. If there are no published estimates of the mean time on first-line treatment, a treatment pathway model could be constructed based on published clinical trials and observational studies that provide data on discontinuation and treatment failure rates.
In Box 3.6, we present two examples of estimating the number of patients on later lines of treatment, one for an acute condition and one for a chronic condition.
Box 3.6 Estimating Number of Patients on Second-Line Treatment Acute Condition
Patients with a specific bacterial infection will be reimbursed for a new antibiotic if it is used for second-line treatment only. Currently, there is only one antibiotic approved for treatment of the bacterial infection that has an 80% probability of successfully eradicating the bacteria. The new antibiotic has been proven to successfully eradicate the bacteria 70% of the time in those for whom first-line treatment fails. If this treatment fails, a third antibi- otic can be used that is very expensive but is 60% successful in eradicating difficult-to-treat bacterial infections. Within a health plan, 10% of patients tend to contract the bacterial infection each year. How many patients out of a health plan of 1 million lives qualify for second-line treatment each year?
Patients with bacterial infection/
first-line antibiotic
Bacteria not eradicated/
second-line antibiotic Prob(success) = 0.8
Prob(fail) = 0.2
Successfully eradicated bacteria
Bacteria not eradicated/
third-line antibiotic Prob(success) = 0.7
Prob(fail) = 0.3
Successfully eradicated bacteria
Bacteria not eradicated Prob(success) = 0.6
Prob(fail) = 0.4
Successfully eradicated bacteria
Proportion of patients diagnosed
Number in health plan 1,000,000
with bacterial infection x 0.1
First line Prob(fail) x 0.2
Number of people needing
20,000 second line
Treatment pathway for estimating number of patients on second-line treatment of a bacterial infection
48