Novel oral iron therapy for iron deficiency anaemia: How to value safety in a new drug?

Novel oral iron supplements may be associated with a reduced incidence of adverse drug reactions compared to standard treatments of iron deficiency anaemia. The aim was to establish their value‐based price under conditions of uncertainty surrounding their tolerability.

the true cost of prescribing "inexpensive" iron salts may be significantly higher than the cost of the product itself, owing to associated increases in the utilisation of health care services. In turn, this has prompted the development of novel oral iron forms that seek to reduce ADRs or a switch away from oral iron to intravenous strategies. 5,6 Both alternative approaches are markedly more expensive than oral iron salts but the key issue concerns the pricing at which they would become cost-effective.
A meta-analysis of 20 randomised controlled trials identified a relative risk of 1.59 (95% confidence interval [CI] 1.43-1.76) for gastrointestinal ADRs to ferrous sulfate versus placebo. 4 Therefore, given that approximately one third of patients taking oral iron experience ADRs (29.9% for ferrous gluconate, 30.2% for ferrous sulfate and 43.4% for ferrous fumarate), 7 an absolute $12% of all patients prescribed common iron salts might avoid gastrointestinal ADRs if an ideal treatment were available. Indeed, quite a number of alternative iron preparations, either on the market or in the pipeline, claim to have superiority over ferrous iron salts in terms of acute, gastrointestinal ADRs. Gastrointestinal intolerance is generally considered to relate to the oxidation of ferrous iron in the gut lumen following iron ingestion, and the generation of damaging reactive oxygen species. To avoid this, iron may be chelated, potentially in its ferrous form (e.g., iron bis-glycinate 8 ) or, most commonly, as ferric iron (e.g., ferric maltol, ferric citrate, ferric EDTA 9,10 ). This not only helps to maintain iron in a soluble form but also opposes the drive for iron redox cycling in the intestine.
An alternative approach involves oral delivery of iron as a nanoparticle (e.g., iron hydroxide adipate tartrate 11 ) which also prevents luminal redox activity and delivers a bolus dose to the enterocyte lysosome, which is a safe house for iron dissolution, recycling and systemic absorption. Finally, an approach that is a hybrid of the two mechanisms described above has also been proposed with sucrosomial iron, which consists of ferric pyrophosphate coated with a lecithin and sucrose esters. This is absorbed as sucrose ester conjugates and particles or vesicle-like structures. 9 With all these potential therapeutic options, decision-making for the prescriber is clearly challenging.
A key determinant for informing decisions concerning the prescribing of medicines is their cost-effectiveness. The National Health Services (NHS) in the UK operate a threshold in the range of £20 000 to £30 000 per quality-adjusted life year (QALY) gained, below which medicines are considered to represent good value for money. 12 It follows that the value-based price of new medicines may be established from knowing the health gains-such as reduction in ADRs-that can be achieved. A medicine priced up to its value-based price would consequently be cost-effective. 13 The aim of this study was to estimate the value-based price, centred on the cost per QALY as a measure of value to the NHS, of oral iron therapies seeking to replace simple oral iron salts.

| CPRD and HES data
The analysis used the CPRD-GOLD database of anonymised, longitudinal, primary care clinical data contributed by general practices from across the UK. CPRD had a coverage of over 11.3 million patients from 674 practices in the UK in 2013, and has been validated to be representative of the UK population for age, sex and ethnicity. 14 The CPRD allows access to linked HES data from NHS Digital. This administrative dataset records all NHS England hospital inpatient admissions, including combined day case and ordinary elective spells.
The protocol for this study (reference number 14_201) was approved scientifically and ethically by the CPRD Independent Scientific Advisory Committee.

| Study population and treatments
The CPRD was accessed to identify female patients, aged between 18 and 45 years (i.e., typically pre-menopausal) who were prescribed,

What is already known about this subject
• Novel oral iron supplements may be associated with a reduced incidence of adverse gastrointestinal reactions but they might not be cost-effective compared with inexpensive iron salts.

What this study adds
• This analysis provides a framework for setting prices according to incidence of adverse drug reactions, and suggests that novel oral iron supplements could be costeffective at prices that are several-fold higher than existing oral iron salts.
• Estimates of the value-based prices of drugs in development provide a basis for assessing commercial viability, while also assisting the healthcare payers in their horizon scanning activities.
between January 2000 and October 2014, at least one of the top six iron products (which were all ferrous salts) by dispensing volume from the prescription cost analysis database. 15 The product codes for these in CPRD are: dried ferrous sulfate tablets 200 mg (33); ferrous fumarate tablets 210 mg (3035); ferrous gluconate tablets 300 mg (712); ferrous fumarate tablets 322 mg (2915, 3151); ferrous fumarate capsules 305 mg (5045, 6052); and dried ferrous sulfate MR tablets 325 mg (1745, 5582). A course of treatment was defined by a new prescription of oral iron distinct from the immediate previous oral iron prescription (by type or by dose), or a prescription of the same oral iron dated more than 2 months after the previous prescription date.

| Health states
CPRD and HES data were used to estimate the probabilities of patients residing in each of six mutually exclusive health states, defined by their response and tolerance to treatment ( Table 1). The Patients were considered intolerant of iron supplementation if their primary or secondary medical records indicated treatment cessation coinciding with a potential ADR (using Read codes for symptoms such as constipation, diarrhoea, nausea, heartburn), a reduction in daily dose or a change in product class. Treatment intolerance was ascertained for each course of treatment and was therefore timeinvariant within a given course of treatment.

| Transition probabilities
A previously described discrete-time Markov model with multinomial logistic regression was used to estimate transition probabilities between health states 16 (see Appendix in the Supporting Information). This was simplified for the economic analysis by averaging the probabilities across doses and treatment courses to provide a single matrix of time-dependent state occupancy probabilities ( Table 1). The Markov model is depicted schematically in Figure 1.

| Health state utilities
Health state utilities were derived from a purposive review of the literature (see Appendix in the Supporting Information).
For the "no improvement & tolerant" health state, utilities were assumed to be represented by data from a study of iron treatment for a population of 236 women with heavy menstrual bleeding (mean baseline Hb 11.0 g/dL). 17 This study reported a mean baseline EQ-5D utility score of 0.76. In the absence of any specific evidence, this utility value was also applied to the "improvement & intolerant" health state, on the assumption that the health effects of an improvement in iron deficiency is offset by the adverse reaction to the iron supplementation.
Peuranpää et al. 17 reported an improvement in Hb at 12 months of 1.2 g/dL following iron supplement intake. The corresponding utility (0.85) was assumed for the "improvement & tolerant" health state.
The disutility associated with being treatment-intolerant was determined from a trial of adult patients with iron-deficiency anaemia and who had failed, or were intolerant to, oral iron therapy. 18 Summary responses to the SF-36 were converted to EQ-5D-3L utilities, 19 which necessitated data to be extracted by digitising figures using WebPlotDigitizer. 20 The difference in EQ-5D-3L utility of 0.04 between baseline (mean Hb of 8.9 g/dL) and the end of follow-up (mean Hb of 11.7 g/dL), was subtracted from 0.76 to estimate the utility of the "no improvement & intolerant" health state, as 0.72.
For the "hospital state" utility value, International Classification of Diseases (ICD)-10 codes obtained from HES data were converted to ICD-9 in order to estimate the marginal disutility. 21 This was based on UK preference scores applied to EQ-5D descriptive questionnaire responses in relation to 135 chronic, ICD-9 coded conditions in the US-based Medical Expenditure Panel Survey, and controlled for gender, age and ethnicity. The mean disutility score for iron deficiency anaemia of À0.036 was subtracted from the "no improvement & tolerant" health state utility score, resulting in a mean utility score for the "hospital state" of 0.72.

| Cost analysis
The total cost per patient was calculated by summing the cost of prescribed iron preparations, the cost of blood tests, GP consultations and hospital stays. A cost multiplier was introduced, such that the cost of "new iron" could range between 1 (= cost of standard iron salts) and 10 times the cost of existing oral iron preparations.
Given the large sample size, the central limit theorem was assumed to apply, and health state costs were obtained from the coefficients of variables included in a linear (ordinary least squares) regression, specified with total costs as the dependent variable, health states and time on treatment (in days) as explanatory variables: where X = health state and ε is the error term.

| Base-case analysis
Treatments received by patients in the CPRD sample were assumed to reflect standard care; while the simulated effect of the intervention

| Sensitivity analysis
One-way sensitivity analyses were conducted to assess the impact of varying each health state utility by ± 0.1 on value-based prices under the base-case scenario of 30% and 40% reduction in the probabilities of treatment intolerance. These were depicted as tornado plots. The health utility corresponding to the state of anaemia being resolved was not varied as this was based on population norms. A two-way sensitivity analysis was performed to estimate the incremental cost-effectiveness ratio (ICER) for "new iron" versus standard treatment for different combinations of prices of "new iron" and probability of being treatment intolerant.
The improved safety of "new iron" was modelled as reductions in the relative risk of being treatment intolerant (in 5% increments up to 50%) and higher probabilities of being in corresponding treatmenttolerant states. For instance, a 5% relative decrease in the probability of being in the "improvement & intolerant" health state required a corresponding increase in the probability of "improvement & tolerant".
Similar changes were modelled for the "no improvement & tolerant" and "no improvement & tolerant" health states. In order for probabilities to sum to 1, they were normalised by sharing any small residual difference across the six health states.
Threshold analyses were undertaken to test the sensitivity of the value-based price, assuming a 30% and 40% decrease in the probability of treatment intolerance, to changes in health state utilities, costs and transition probabilities.

| Probabilistic sensitivity analyses
A probabilistic sensitivity analysis was performed to consider the joint uncertainty in costs and QALYs. Correlation between cost parameters in the regression model was preserved using the Cholesky decomposition of the variance-covariance matrix. Monte Carlo simulation was conducted to obtain 10 000 correlated draws from a multivariate normal distribution. 26 For utilities, a fixed standard deviation of 0.2 was assumed, and 10 000 draws made from independent beta distributions fitted using the method of moments.
Cost-effectiveness acceptability curves were constructed for two combinations each of the price of "new iron" (5 and 10 times the price of iron salts) and the probability of treatment intolerance (reduced by 30% and 40%). 27

| Scenario analysis
A scenario analysis was conducted to estimate the ICERs for "new iron" versus standard treatment for different prices of "new iron" and reduced probabilities of being treatment intolerant in the "hospitalisation" health states in addition to the "no improvement & intolerant" and "improvement & intolerant" states. As described above, the probability of treatment tolerance was increased in the opposite correspondent health states "no improvement & tolerant", "improvement & tolerant" and "anaemia resolution". Probabilities were again normalised to sum to 1 by distributing small differences equally across the six health states.

| Incremental analysis and value-based pricing
The model reflecting current use of iron salts yielded a mean total cost to the NHS of £779.24 over 1 year, and 0.839 QALYs. Assumed 30% and 40% relative risk reductions in the likelihood of treatment intolerance with "new iron" resulted in 0.0064 and 0.0086 QALY gains in comparison with iron salts (Table 4). Threshold analyses indicated that at the £20 000 per QALY threshold willingness to pay, the price of "new iron" could increase to 7.30 and 9.44 times that of the basket of iron salts (£1.38 per 28 tablets) to remain cost-effective with 30% and 40% reductions in treatment intolerance. This is equivalent to £10.07 and £13.02 per 28 tablets (Figure 2). At the higher threshold of and £18.38 per 28 tablets) for 30% and 40% relative risk reductions in intolerance, respectively.

| Sensitivity analyses
The one-way sensitivity analyses indicated that the value-based price was most sensitive to utility values in the state of "improvement & tolerant" (Figure 3). An increase in utility by 0.1 resulted in higher valuebased prices of £27.60 and £31.05 for the scenarios of the "new iron" reducing the risk of intolerance by 30% and 40%, respectively. However, a decrease in utility by 0.1 in three health states led to negative value-based prices, reflecting the fact that the "new iron" would be in the south-west quadrant of the cost-effectiveness plane. F I G U R E 2 Value-based price (£ per 28 tablets) of a hypothetical new oral iron preparation, according to different relative reductions in the probability of treatment intolerance, and for the upper and lower bounds of the willingness to pay threshold range The two-way analysis ( Table 5)  would be expected to be, the ICERs indicated that higher prices F I G U R E 3 Tornado plots depicting the sensitivity of the value-based price (£ per 28 tablets) to changes in health state utility (± 0.1). Vertical lines indicate the basecase value-based prices at 30% (upper figure) and 40% (lower figure) reduction in intolerance T A B L E 5 Two-way sensitivity analysis for the base-case, illustrating the dependency of the incremental cost-effectiveness ratio on the price of "new iron" and effectiveness in terms of reduction in intolerance may be set for "new iron" treatments with lower probabilities of intolerance.
The results from the probabilistic sensitivity analysis, assuming "new iron" is priced at 5 times the average cost of iron salts, and associated with 40% reduced probability of treatment intolerance, indicated that "new iron" is likely to be cost-effective, with 0.94 and 0.98 probabilities of being cost effective at willingness to pay thresholds of £20 000 and £30 000 per QALY, respectively ( Figure S1 in the Supporting Information). At 10 times the cost of iron salts, and 30% relative reduction in intolerance, the probability of "new iron" being cost effective reduces to 0.21 and 0.65 for willingness to pay values of £20 000 and £30 000 per QALY, respectively.

| Scenario analysis
In the alternative scenario where "new iron" is assumed not only to reduce the probability of treatment intolerance, but also the probability of residing in the hospitalisation health state, there were more zones of cost-effectiveness for the same combination of reduction in intolerance and prices of "new iron" (  13 The value-based price should therefore be considered as the maximally acceptable price. The analysis benefited from using routine NHS data which provided accurate estimates of health state occupancy and healthcare resource utilisation associated with the management of iron deficiency anaemia. The CPRD includes patient electronic healthcare records (EHR) collected routinely in primary care 30 and is linked to patients' HES data for accurate determination of hospital care. 31 By applying unit cost to items of resource use, the analysis considered the actual costs of primary and secondary care services in the NHS.
There was, however, a limited evidence base relating to health state utilities requiring assumptions that may not be generalisable to the modelled population. Sensitivity analyses indicated that changes in health state utilities within plausible ranges led to variation in value-based prices. Further research on utilities in iron-deficiency anaemia is warranted. The model structure was also limited by the data available from EHR, and there were no randomised controlled trial data on the relationship between dose and ADR, 4 or the effectiveness of sequential courses of treatment. Our analytic time horizon was set to 12 months, which may not adequately capture all costs and consequences, although it is recommended that treatment with elemental iron should be limited to 3 months after iron deficiency is corrected, this being considered sufficient to allow stores to be replenished. 32 Finally, we note that, currently in the UK, iron can be administered parenterally when oral therapy is unsuccessful, for example if patients cannot tolerate oral iron, or do not take it reliably.
Our economic analysis-very conservatively-did not compare new, hypothetical oral iron preparations with parenteral iron, such as ferric carboxymaltose, or iron dextran, sucrose or isomaltoside 1000.
This was principally because the CPRD data extraction was limited to patients being prescribed oral iron supplementation. While there are several budget impact analyses of parenteral iron for this clinical indication, there are no economic analyses; had we considered parenteral products, a higher value-based price would likely have resulted.
In conclusion, a significant proportion of patients with iron deficiency anaemia are intolerant to oral iron salt preparations. This increases the risk of non-adherence and treatment failure as well as impairing patients' health-related quality of life. The prospect of novel oral iron preparations to reduce the incidence of ADRs warrants careful analysis of how they are to be priced in the context of inexpensive alternative generic iron salts. This value-based pricing analysis estimates that new treatments may be cost-effective at prices that are several-fold higher than existing oral iron salts, and which may be attractive for commercial development while proving to be costeffective to the NHS.