Let's hope Pfizer (NYSE:PFE) was in the mood for a really crappy day on or about April 13th, because that's what NICE (UK's National Institute for Health and Clinical Excellence) delivered to them in the form of a technology evaluation report for the recently approved inhaled insulin, Exubera. The nice thing for readers of this blog is that this month I'll review the NICE decision to see whether it was at least fair, if not nice, and what other companies might take away from it.
One of the points that should strike any reader of the assessment report at the very outset is that Pfizer spends more than any pharma on R&D, and they didn't slouch on the development of Exubera. Indeed, the development program was more than adequate for regulatory approval and was clearly designed with pricing and reimbursement in mind. Yet, Pfizer failed to make a positive impression with the hard-to-please NICE reviewers: Corri Black, Ewen Cummins, Pam Royle, Sam Philip and Norman Waugh.
Skeptics will argue that NICE's decision was pre-ordained and that they fudged the numbers in their evaluation to make the data fit their preferred outcome. I usually count myself among the skpetics, but I will instead argue that one of NICE's mandates is to place the burden of proof for incremental cost effectiveness [ICE] squarely on the shoulders of the sponsor; they have no mandate to give sponsors the benefit of the doubt in their independent assessment.
To save time and space, I'm not going to summarize the entire NICE report. Instead, I'll point you to the brief overview document and ask you to read that prior to continuing. I'm also not going to dwell on details of the economic arguments made by either side. Rather, I'm going to focus on reasons NICE used to accept or reject the assumptions and data from the sponsor for economic modeling, because these assumptions and data drive the modeling results and thus the final decision.
Sponsors have control over the design, conduct, analysis and reporting of economic-supporting studies. They do not have control over the assumptions NICE or other independent agents use for their ICE models, but they do have the ability to adequately justify and support the choice of their own assumptions. So, did Pfizer spend their money wisely? Where could they have done more or better during study design and execution? What might they have done differently during reporting? Most importantly, is it feasible to demonstrate the ICE of new drugs prior to first launch, or should sponsors instead focus on approval and enough ICE evidence to get high co-pay (e.g. Tier 3) coverage in the U.S. and limited to no coverage in Europe?
NICE Conclusions and Arguments
Background Information from NICE of Special Importance to NICE Opinion
1. As Type 2 diabetes [N.B. Type 2 diabetes, because of its much higher prevalence than Type 1, accounts for roughly 80% of all insulin users in the U.S. and U.K. and an even higher proportion of insulin doses administered] severity (as reflected by HbA1c) changes, the relative significance of fasting and pre-prandial versus post-prandial glucose levels likewise changes. At low HbA1c levels, post-prandial glucose accounts for up to 70% of HbA1c elevation; at HbA1c levels above 8.4, fasting and pre-prandial glucose account for the majority of HbA1c elevation. This is important for Exubera, because it is a short-acting insulin that primarily affects post-prandial glucose.
2. Type 2 diabetes progresses, and most patients will eventually require exogenous insulin. In the UKPDS (the largest prospective, controlled study of Type 2 diabetes), only a small proportion of patients who required insulin refused it.
3. Exubera has a short-acting profile and would not, therefore, remove the need for a long-acting insulin injection for nearly all patients who might require insulin (assuming a stepwise approach to insulin therapy, where insulin is a therapy of last-resort).
4. Use of Exubera would not obviate the need for self-monitoring of blood glucose.
5. Given above, NICE limited its review to just three clinical scenarios where the use of Exubera would be considered “reasonable”: as an alternative to short-acting injected insulin in Type 1 diabetes; as an alternative to injected insulin regimens in Type 2 patients failing maximal oral therapy; as an alternative to intensifying the insulin regimen in Type 2 patients failing once-daily long-acting insulin.
Safety and Effectiveness of Inhaled Insulin
1. Inhaled insulin is clinically effective, and is as good as short-acting soluble (i.e. regular) insulin in controlling blood glucose.
2. The frequency of hypoglycaemia is similar to regular insulin.
3. Long-term studies are needed to rule out uncommon but serious adverse effects of inhaled insulin on the lung.
4. None of the published trials have compared Exubera with short-acting injected analogs.
5. Clinical trial subjects used combinations of short-acting, and either long or intermediate-acting insulin, and both were changed, making it more difficult to assess the change from soluble (i.e. injected) to inhaled.
1. Most patients preferred inhaled to injected short-acting, but bias due to trial design could have impacted this finding
2. The sponsor provided no evidence to support the assumption that a switch to insulin would be more acceptable if inhaled insulin was an option, yet used this assumption in their models.
1. Sponsor’s model is high-quality, but their key assumptions (QOL improvement size and acceptability of switch to insulin—see #2 under “Patient Preference) are not supported by the evidence; the ICE results depend more on the assumptions than the model.
2. The cost of inhaled insulin will be much higher than injected.
3. Inhaled insulin is unlikely to be cost-effective.
Pfizer was provided the opportunity to respond to the draft assessement. They summarized their main concerns as follows:
1. The subjective and non evidence-based assumptions made regarding the current rates of initiation of insulin therapy in treated patients uncontrolled on oral anti-diabetic drugs and intensification of insulin therapy in patients uncontrolled on their current insulin regimen.
2. The decision by the Assessment Group to calculate the cost-effectiveness of only two of the six patient groups used in the Pfizer submission.
3. The misleading use of evidence to minimise the concerns patients have with injecting insulin.
4. The significant differences in a number of key assumptions used when modelling the cost-effectiveness of Exubera
In turn, the assessment group responded to Pfizer's comments, although not to their re-modelling exercise specifically. They seemed willing to re-visit Pfizer's Point #1 and admitted that initiation of long-acting insulin (e.g. Glargine) has facilitated initiation of injected basal insulin earlier in the course of the disease and for HbA1c values above 7.5%. The assessment group took issue with Pfizer's contention that the group misleadingly used evidence to minimize patient concerns about a switch to insulin. They argued, essentially, that common sense precludes use of the Pfizer estimates of QoL improvement from switching, given that they are comparable to estimates of the QoL gain from avoiding hemodialysis.
Okay, so that's the crux of the debate between NICE and Pfizer so far. What might Pfizer have done differently to bolster their case with NICE? First, NICE was clearly annoyed that Pfizer did not do direct comparator studies with the latest short-acting analogs, instead using regular insulin as the comparator. Exubera's pharmacokinetic profile is more similar to the analogs than to regular insulin, so a direct comparison with analogs would have been reasonable and would have avoided this annoyance with lack of data. Pfizer might have been concerned that Exubera's efficacy would have looked less robust versus the analogs, but that is a risk they needed to take to convince NICE that the appropriate comparator was tested. After all, many patients who will consider using Exubera will be using analogs prior to a switch.
Pfizer did a reasonably good job with observational data showing that HbA1c control is currently inadequate and that a switch to insulin probably occurs too late in many Type 2 patients. However, they failed to demonstrate in a clinical study(ies) that Exubera would improve the situation. The studies that were performed were enough to demonstrate convincingly that Exubera is effective and safe (at least in the short-term), but Pfizer lacked the "real-world" QoL and efficacy comparisons to be convincing for ICE demonstration.
What Pfizer seemingly needed to convince NICE of Exubera's ICE was a large, randomized, controlled study, drawing from an unselected Type 2 population showing an improvement in therapy compliance and blood glucose in patients who are treated relatively early (i.e. with HbA1c as low as 7.5%) with basal insulin plus Exubera versus basal-bolus regimens of injected insulin. It's a risk for Pfizer, sure, but if their drug has a high compliance rate, as argued, they should come out ahead in terms of efficacy alone. Right? Adding QoL assessments to the mix should make Exubera the clear winner. So, why doesn't Pfizer just do this? Why didn't they do it prior to approval?
The relevant question for planners is whether the type of study alluded to above is feasible and desirable. The two major risks of such a study are: failure to demonstrate non-inferior efficacy and failure to demonstrate improved compliance and QoL. The efficacy risk as viewed today seems small, given the other studies of the drug in Type 2 diabetes, but at the time such a study might have been initially contemplated (i.e. at the start of Phase 3 or earlier), these reassuring data would not have been available. Likewise, compliance is always a risky proposition, as it is difficult to determine during a clinical study, which is an artifical environment that inflates compliance (adherence) rates for all therapies. Therefore, Pfizer would correctly have little confidence that Exubera would come out ahead in this measure.
What about QoL? NICE criticizes Pfizer for biased estimates of QoL (measured as time trade-offs, TTO) that artificially inflate the QoL increment for Exubera over injections. NICE would be unlikely to criticize the TTO increment if seen in a study of the type I've suggested above. But look at Pfizer's risk: both groups would be taking injections chronically (one as basal-bolus, one as basal-Exubera), both groups still must endure frequent sticks for blood glucose determintion. In order to show superiority in QoL, Exubera must be superior to bolus injections in specific domains related to self-injections, especially injections occurring during the day, when they might impose upon activities of daily life the most. An incremental improvement doesn't seem like such a "slam-dunk" from this perspective, does it?
What's a planner to do? I think Pfizer acted appropriately given the properties of the drug, its delivery mechanism and the alternative treatment available to patients. It's always a tall order to demonstrate ICE for a new therapy with similar average efficacy as older therapies. Toss in the fact that the new therapy must cost two to three times as much as older therapies in order to meet profit margin requirements, and economists are faced with a nearly untenable situation. It's rarely in the company's best interests today to take extraordinary risks to demonstrate ICE convincingly prior to first launch if the risks aren't necessary to gain marketing approval and at least Tier 3 coverage in the U.S., understanding that reimbursement in countries like UK is unlikely.
Looking into the crystal ball a bit, it's not difficult to envision a future where such risks become worthy of serious consideration in an environment of required ICE demonstrations prior to acceptance of a newly approved therapy on Medicare formularies in the U.S. The reality today is that Medicare can't restrict access to new therapies after FDA approval. I don't see that situation continuing indefinitely.