with management of the neovascular entity [1– 15]. These costs are considerable, especially since there are approximately 1,65,000 new cases of neovascular (“wet”) AMD in the USA annually. Nonetheless, the costs associated with the 1.7 million annual new cases of atrophic (“dry”) AMD are even more substantial [14, 15].

The direct ophthalmic medical costs associated with the management of neovascular AMD are the most visible, especially to those who allocate healthcare resources. Numerous papers have dealt with the subject of therapies for the neovascular variant, including laser photocoagulation [1, 2], intravitreal pegaptanib injections [3, 4], photodynamic therapy with verteporfin [3, 5, 6] and intravitreal VEGF-A inhibitors [7–13].

The costs associated with a disease, however, include not only the very apparent direct medical costs, but also the direct nonmedical costs and the indirect medical costs [13]. It appears that these direct nonmedical costs and indirect costs for neovascular AMD actually exceed the direct medical costs several times over [13].

Background information on healthcare economic analysis is helpful in understanding the current economics associated with AMD. Included herein are the four major variants of healthcare economic analysis, as well as a more extensive discussion on the most sophisticated form, costutility analysis, which integrates clinical outcomes with their associated costs. Value-Based Medicine®, the standardization of input variables and outcomes associated with cost-utility analysis, is also discussed in detail [15–23]. Lastly, the macroeconomic aspects (economics on a national level) associated with AMD are addressed.

There are essentially four major variants of healthcare economic analysis [23–25]. No matter the variant, however, a healthcare economic analysis is more robust if it is based upon the highest level of evidence-based medicine, a discussion of which follows.

Evidence-Based Medicine

In 1972, Archie Cochrane advocated the randomized clinical trial to promote the highest quality, most reproducible, medical evidence [26]. But, it was

not until the last decade of the twentieth century that the term “evidence-based medicine” was popularized [27, 28].

Interventional Evidence

There are basically five levels of medical interventional evidence [29]. The higher the level of evidence (Levels 1 and 2), the greater confidence the clinician can have that the data from a study are reproducible (reliable). A summary of the levels of interventional evidence is shown in Table 10.1 [29].

Level 1 interventional evidence is typically provided by a randomized clinical trial with low type 1 and type 2 errors.

•Type I error. The type 1 error, or “false positive” error, in a Level 1 clinical trial is typically associated with a probability, the p-value, designated by the letter a(alpha), of <0.05 for detecting significantly different outcomes. This indicates that a false positive result will occur in <5% of instances. The a(alpha) is analogous to the probability of a jury finding an innocent person guilty.

•Type II error. The type II error, or “false negative” error, is typically associated with a probability, designated by the letter b(beta), <0.20, meaning that the chance of missing a significant outcome is less than or equal to 20%. The b(beta) is the probability of failing to detect a significant association, analogous to the probability of a jury finding a guilty person innocent.

•Power. The power of a clinical trial is calculated by subtracting the type II error, or b(beta), from 1.0. For level 1 interventional evidence, a power of (1.0–0.20 =) 80% or greater is generally required to detect a predetermined outcome [29].

Table 10.1 Levels of interventional evidence [29]

Level 1 – randomized clinical trial with type 1 error <0.05 and type 2 error <0.20

Level 2 – randomized clinical trial with higher type 1 and/or type 2 error

Level 3 – nonrandomized clinical trial

Level 4 – case series

Level 5 – case report

•The negative study. Knowing the type 2 error is not as relevant for a clinical trial that has a significant outcome (p < 0.05) as it is when no significant difference is found. With a negative clinical trial, one that does not demonstrate a significant difference, it is important to know whether sufficient numbers of participants participated in the study to effectively assess a given outcome. For example, a 70% beneficial outcome when n=10 is not nearly as relevant as a 70% beneficial outcome when n=1,000. Knowing the power is most important to assess whether there are sufficient participants in a study to make it clinically relevant.

•Meta-analysis. In instances in which Level 1 interventional evidence is not available, a good meta-analysis can provide Level 1 evidence by combining two or more clinical trials, which may be underpowered. While much less costly than a clinical trial, this type of analysis can suffer from the reality that participating cohorts, treatments, and treatment outcomes in different trials may not be exactly the same.

Level 2 interventional evidence is supplied by the randomized clinical trial with a type 1 error >0.05 and/or a type 2 error >0.20.

Level 3 interventional evidence occurs with a nonrandomized clinical trial.

Level 4 interventional evidence is supplied by a case series.

Level 5 interventional evidence or anecdotal evidence is gleaned from a case report.

Of course, Level 1 clinical evidence is not available for all interventions. It is very expensive to obtain, and often requires long time periods to ideally assess an outcome. The other levels of interventional evidence, however, can provide important information as well, especially concerning the demographic features of a cohort, life expectancy, and the incidence of adverse events.

Masking

The clinician should also be aware of whether a study is appropriately masked [23, 28]. Masking can occur in the form of a single-blind study, in

which a participant typically does not know which treatment is given, a double-blind study, in which neither the researcher nor the participant knows which treatment is administered, or a triple-blind study, in which neither the researcher, the participant nor the person evaluating the treatment response knows who is receiving which treatment. A tripleor double-blind study is preferable, but may not always be possible. Masking greatly reduces the possibility of bias. An openlabel clinical trial is one in which there is no masking; it is commonly encountered with drug trials following an initial period of masking.

Dropout Rate

A dropout rate of <5% is considered excellent, while a rate of 5–15% is still acceptable. A rate greater than 20%, however, casts suspicion upon the validity of a study [28]. For example, a 10% event rate can be misleading if there is a 30% dropout rate and half of those who dropped out (15% of all participants) also had the event. Consequently, the true event rate might have been 25%, or 250% higher than the 10% event rate reported in the study with the 30% dropout rate. An intent to treat study keeps in data from all patients who are randomized, thereby allowing a clinician know the dropout rate.

Pearl

A dropout rate of >20% makes the results of a clinical trial suspect.

Validity

There are two basic types of validity: criterion validity and construct validity.

Criterion validity measures how well an intervention measures up to the criterion, or “gold standard” in the field. For the treatment of neovascular AMD, one might assess how well a new drug measures up to the current criterion of the VEGF-A inhibitor ranibizumab [1–3, 5–7].