Do We Understand Unmet Need? A Proposal to Use Length-Of-Life Equivalent (LOLE) as a Patient-Centric Measure of Unmet Need

Abstract

Many decision-makers have emphasized the importance of leveraging patient experience data to measure unmet need. However, there is no standardized, patient-centric unmet need measure that formalizes how the value judgements inherent in such a measure should be made. Several initiatives have identified measuring unmet need as one of the primary uses of patient preference data. After reviewing how decision-makers define unmet need, this paper proposes that a thresholding method could be used to generate a standardized, patient-centric, disease-agnostic, quantitative unmet need estimate, length of life equivalent (LOLE). LOLE would address some of the limitations of current methods, including facilitating capture of the impact of disease beyond health-related quality of life, and being more sensitive to the impact of a disease on patients. However, the acceptability of LOLE raises questions for decision-makers, including: Is length of life equivalence the best common metric in which to express unmet need? Is it appropriate to rate a disease as having no unmet need if patients are unwilling to trade off life expectancy for improvements in their quality of life? Can LOLE be estimated for more complex disease profiles? Is thresholding the appropriate method to use to estimate LOLE? How should LOLE be integrated into decision-making, including the level of LOLE that defines different levels of unmet need? Further work could usefully address these questions with decision-makers.

Key Points for Decision Makers

There is currently no standardized, patient-centric unmet need measure that formalizes how the value judgements inherent in such a measure should be made.

The thresholding method could be used to estimate such a measure, expressing unmet need on a common scale, length of life equivalents (LOLE).

Further work by decision-makers is required to address the normative and practical questions posed by LOLE.

Background

Alignment of healthcare resources to the most critical unmet needs is a key element of optimizing the impact of health investments [1]. To guide industry investment and approval and reimbursement decisions to deliver this impact, it is necessary to have a measure of unmet need that draws on patients’ experience to generate a standardized assessment of relative unmet need across indications [1]. Recent efforts have emphasized the importance of collecting patient experience data (PED) to inform unmet need assessments [2–9]. Prominent among these efforts is the Food and Drug Administration’s (FDA) Patient Focused Drug Development (PFDD) program and guidance on the use of PED [10].

Patient preference (PP) data is one type of PED that has received much attention. The FDA defines PP data as “qualitative or quantitative assessments of the relative desirability or acceptability to patients of specified alternatives or choices among outcomes or other attributes that differ among alternative health interventions” [11]. Several initiatives have identified understanding unmet need as one of the primary uses of PP data in regulatory and health technology assessment (HTA) decisions [12–15]; however, despite the commitment to PFDD and available guidance on PED, there remains uncertainty about how PP data could be used to estimate unmet need and incorporate these insights into decisions [2, 5]. While PP methods are now well established, we know of no examples of effectively using PP to measure unmet need nor guidance on how PP data should be collected to estimate unmet need.

The goal of this paper is to offer one possible method for using PP data to generate a standardized patient-centric estimate of unmet need. To do this, the next section asks how decision-makers define unmet need. The subsequent section proposes a PP method for measuring unmet need. The final section considers the advantages of this approach and identifies the normative and practice questions raised by implementing such an approach.

How is Unmet Need Defined and Measured by Decision-Makers?

While addressing unmet need is a central goal of investing in new health technologies, the diversity of health systems, stakeholders, and notions of fairness mean that many different conceptions of unmet need are employed [16]. A recent review identified 16 different definitions of unmet need used by patient representatives, pharmaceutical and vaccine developers, regulators, and HTA bodies [17]. The authors grouped these into three categories: the availability of existing treatments, disease severity or burden, and the size of the patient population. The availability of existing treatments was referenced by all the identified definitions. Under this definition, need is unmet if there is either a lack of available or accessible treatment for a disease, or if existing treatments are considered inadequate. Just over one-third of definitions (38%) included reference to disease severity or burden. A small proportion (6%) referenced the size of the patient population, either the incidence or prevalence of a condition.

Building on that review, Zhang et al. [16] developed a framework that included three different dimensions of unmet need: need (including disease burden and whether satisfactory treatments are available); equality of access (including the existence of any, accessible treatment, equality of outcomes, which includes both the components of need and quality of access); and maximizing principles (which broaden the outcomes of interest to include caregiver and economic burdens).

Agencies deploy these dimensions in different ways to integrate unmet need into their decisions. Regulators assess unmet need against thresholds that inform the regulatory review mechanism that is relevant for a treatment [17] and how the treatment’s benefit–risk is assessed [18]. For instance, the European Medicines Agency references whether a treatment “fulfils an unmet need” as one of the criteria for qualification for priority medicines designation [19]. Similarly, the FDA’s Expedited Programs for Serious Diseases defines unmet medical need as a “condition whose treatment or diagnosis is not addressed adequately by available therapy” [20]. The FDA states that “for a drug intended to treat a serious disease with unmet needs, [it] may accept greater uncertainties about benefit or risk at the time of approval” ([18], p. 10).

Operationalizing these definitions of unmet need involves making value judgements. For instance, the EU’s proposed updates to its medicine’s regulatory framework define unmet need as existing where “there is no treatment—or where existing treatment is considered unsatisfactory—and [there is a] meaningful reduction in disease morbidity or mortality” [21]. Assessment of what is “unsatisfactory” and “meaningful,” and doing so across the multiple dimensions of mortality and morbidity, requires decisions about which stakeholders should make such judgements and how this should be performed. HTA agencies have tended to be more explicit about these matters than regulatory agencies. For instance, more prominent efforts to quantify unmet need by HTA agencies in Norway [22], the Netherlands [23], and the UK [24] define unmet need in terms of absolute shortfall (AS) and/or proportional shortfall (PS), where the above value judgments are based on general population preferences as elicited in the utility tariffs used to estimate quality-adjusted life years (QALYs).

AS is defined as the absolute QALYs lost due to the condition when compared with the remaining QALYs that would be expected to be gained over the remainder of their lifetime by the general population with the same age and sex distribution as those with the condition. PS is defined as the proportion of QALYs lost due to the condition. QALYs gained are weighted depending on the level of AS/PS. For instance, for National Institute for Health and Care Excellence (NICE) decisions: where the AS is less than 12 or the PS is less than 0.85, the QALY weight is 1; for an AS of 12–18 or a PS of 0.85–0.95, QALY gains are weighted 1.2 times; and where AS is more than 18 or PS is more than 0.95, QALY gains are weighted 1.7 times. For instance, if a patient group could expect only another 1 QALY without treatment while a member of the general population with the same age but without the disease could expect 30 QALYs, the AS would be 29 QALYs and the PS would be 0.97. Then, a treatment that generates 3 QALYs gained would be assessed as having generated 5.1 QALYs (3 QALYs × a weight of 1.7).

While AS and PS provide standard quantitative estimation of condition severity, with explicitly defined approaches for making the value judgements involved, the resulting metrics are also subject to the limitations of the QALY. These include that the resulting estimates only capture a narrow set of disease impacts, and they are insufficiently sensitive to the impact of a disease on patients. Other frameworks have attempted to address these concerns by adopting a broader assessment of disease impact and/or alternative standardized metrics. Among the most prominent is the Global Burden of Disease measure [25], which adopts the disability-adjusted life year to measure disease burden. While this provides an alternative method for weighting mortality and morbidity impacts, most applications leverage public or health practitioner preferences, rather than patient preferences [26]. A recent endeavor to establish a framework for measuring unmet need—the NEED project—aims to address some of these challenges [1]. To date, the NEED project has generated a comprehensive list of unmet need in three domains: health needs, healthcare needs, and social needs. The project’s next step is to determine how the relative importance of these domains should be established. Might a PP-based method address the limitation of the QALY and/or provide the relative importance measure being sought by the NEED framework?

Designing a Patient Preference Method to Estimate Absolute Shortfall from a Patient Perspective

This section introduces a PP-based unmet need metric—the “length-of-life equivalent” (LOLE). For any approach to estimating unmet need to have general applicability, it will need to accommodate the diversity of conceptions of unmet need across healthcare stakeholders. With the requirement that the attributes of unmet need should be meaningful to patients, a PP-based method can achieve this and translate different notions of unmet need onto a standard scale by translating them into the loss of utility experienced by the patient. That is, all elements of unmet need are translated into a common utility metric. This general concept is then operationalized by converting utility into an equivalence measure—LOLE. In its most general form, the notion of unmet need underlying the LOLE approach is presented in Table 1.

Table 1.

General LOLE unmet need framework

Attributes of unmet need	Without disease A	With disease A	LOLE
Impact on patient-relevant aspect(s) of quality of life	1. Quality of life without disease A	2. Quality of life with disease A	5. The improvement in length of life that gives the same utility as improving quality of life from 2 to 1
Impact on length of life	3. Length of life without disease A	4. Length of life with disease A	6. Improvement in length of life moving from 4 to 3
Total unmet need			5+6

LOLE length-of-life equivalent

To explain the general LOLE framework and illustrate its application, we start with the unmet need of a hypothetical condition, disease A, described in Table 2, which reduces life expectancy, is symptomatic, reduces a patient’s ability to do daily activities, work, and participate in social and leisure activities, and the available treatment is associated with a side effect and is burdensome due to the mode of administration.

Table 2.

Unmet need of hypothetical disease A

Attribute		Without disease A	With disease A
Remaining years of life		Age-, sex-matched for “healthy” person	Reduced
Quality of life	Symptom 1		Increased
	Ability to do daily activities		Reduced
	Ability to work		Reduced
	Ability to participate in social or leisure activities		Reduced
	Side effect 1	None	Increased
	Treatment burden	None	Mode of administration

The unmet need associated with disease A is multiattribute and known to the patient. A patient experiences disease A, and it is possible to measure the corresponding reduction in life expectancy and quality of life compared with an age- and sex-matched healthy person. As such, the research problem is to value a well-defined, simultaneous improvement across multiple attributes. It is not necessary to estimate separate preference functions for the attributes that define unmet need. Rather, the problem is to value the “bundle” of simultaneous improvement across multiple attributes. It is also not necessary to develop preference functions that allow exploration of preference for uncertainty levels on multiple attributes. Rather, the research objective is to estimate a single value—that of the known unmet need. This relatively simple valuation problem lends itself to a less complicated elicitation technique, such as thresholding [27].Thresholding would value unmet need as the change in one attribute that generates the same value as resolving the unmet need. Or, to put it another way, the improvement in one attribute that would compensate for the utility difference between being healthy and living with the disease. As such, we need to select a numeraire, or the metric in which we will express the value of unmet need. One option would be to select a numeraire that aligns closest with the notion of value currently accepted by many HTA agencies. When HTA agencies use time trade-off (TTO) or standard gamble (SG) methods to value health states, participants are asked to trade off changes in length of life or likelihood of death for changes in quality of life. Adopting the same concept of value, a thresholding exercise could be designed to estimate the reduction in length of life that would generate the same disutility as the reduced quality of life caused by the disease.

The required thresholding task is presented in Table 3. In the first choice task, patients would be asked to choose between their current disease experience (option B) and the quality of life of an age- and sex-matched “healthy” person experienced over their current life expectancy (option A). If unmet need exists, option A will dominate option B, and all patients who are appropriately attending to the choice task will choose option A. The remainder of the choice tasks hold constant all the cells shaded blue and reduce the life expectancy shown for option A until the patient is indifferent between option A and option B.

Table 3.

Example thresholding choice task to estimate LOLE

LOLE length-of-life equivalent

The result of the thresholding exercise is the identification of length of life X as seen in Fig. 1. This is the length of life in a “without disease” quality of life at which patients are indifferent between that health profile and their current “with disease” health profile. At this point, we can say that, from the perspective of the patient, the utility of living with their disease (C + D) for length of life Y is equal to the utility of living without the disease for X years (A + C). Put another way, the utility lost due to the lower quality of life associated with the disease (A + B) is the same as the utility lost from losing (Y − X) years at the quality of life of someone without the disease (B + D).

Fig. 1.

Indifference point identified in the thresholding exercise. Y, length of life with disease; Z, length of life without disease. Y − X is the life years equivalent of the decrease in quality of life caused by living with the disease and Z − X is the life years equivalent of the unmet need

By comparing Fig. 1 with Table 1, we can also see that the thresholding exercise only measured part of the unmet need (A + B); however, it is now possible to express the whole unmet need (A + B + E + F) on the same scale—length of life at the quality of life associated with not having the disease. Z − Y is the reduction in life years attributable to the disease without any adjustment for quality of life. This value can be ≥ 0, but it cannot, by definition, be negative. The overall unmet need associated with disease A can therefore be expressed as being the equivalent of the disutility associated with a Z − X reduction in length of life of an age- and sex-matched “healthy” person.

Applying LOLE to Measure Inequality of Access and Outcomes

The application of LOLE described in the previous section measures Zhang’s concept of need. To appreciate how LOLE can also be used to measure Zhang’s concept of inequality of access and outcomes, consider applying the thresholding exercise with a sample of patients. The thresholding exercise generates an estimate of need for each patient in the sample and the variation in this estimate between patients will reflect both their different experiences of the disease and differences in their preferences. If the thresholding exercise also captures relevant patient characteristics, such as access to treatment or social determinants of health, then a multivariate model can be built to estimate how patients’ LOLE estimate of varies with these characteristics.

Applying LOLE in this manner would require a large enough sample of patients, stratified to allow the power to estimate how LOLE varies between patient groups. In turn, this requires an understanding of how the perception of unmet need varies between patients. Once the LOLE model is built, a treatment-specific estimate of LOLE can be estimated, controlling for the characteristics of the patients who will be able to access the treatment.

Figure 2 illustrates how LOLE could be applied by decision-makers to operationalize a standardized, quantitative assessment of unmet need that leverages patients’ preferences to weigh the different components of unmet need and capture both the “need” and “inequality of access and outcomes” dimensions of unmet need.

Fig. 2.

Illustration of the application of LOLE, example of HTA decision-making. HTA health technology assessment, LOLE length-of-life equivalent, MOA mode of administration, QALY quality-adjusted life year, QoL quality of life, SAE serious adverse event

Discussion

This paper outlines an approach to develop a standard, quantitative estimate of unmet need that considers both disease burden and differential access and outcomes elements of the definition of unmet need; it also looks to the patient to determine what is meaningful or dissatisfactory about unmet need. Specifically, it is proposed that a thresholding method could be used to estimate the LOLE. Table 4 describes the LOLE approach using the six questions identified by Dolan [28] as being required to define a method for measuring health outcomes.

Table 4.

Characteristics of the LOLE method

Methodological question	Characteristics of LOLE
What is to be valued?	The impact of a disease on quality and length of life
How is it to be described?	Description of the entire health profile of a disease. The dimensions of disease impact to be included in the description, or how this is arrived at, should be determined by the agency, contingent on their definition of unmet need. We would advocate that patients be engaged to understand unmet need
How is it to be valued?	Thresholding technique estimates the reduction in length of life that would generate the same disutility as the disease profile
Who is to value it?	Patients
How are values for all health states to be generated?	The goal is to value a single disease profile from the perspective of a single patient group. Valuing other disease profiles requires the exercise to be repeated
How are valuations to be aggregated?	Thresholding provides an estimate of LOLE for each participant. Depending on the agency’s needs, this data could be aggregated in several ways. Most obviously would be to estimate the mean LOLE for the target patient group. But, contingent on the appropriate sample, multivariate analysis could estimate how LOLE varies between patient groups

LOLE length-of-life equivalent

Overview of the Benefits of and Questions Raised by LOLE

The benefits of and questions raised by LOLE are summarized in Table 5. While LOLE can provide unmet need input into many health decisions, just as with measures of AS currently adopted by HTA agencies, LOLE expresses unmet need in units of disutility generated by a disease. LOLE would mimic the strengths of the QALY-based AS metrics adopted by several HTA agencies, consisting of a common, comparable, disease agnostic, quantitative measure of unmet need; however, LOLE additionally helps to address some of the limitations of QALY-based metrics. First, QALYs only capture a narrow set of impacts of a disease on health-related quality of life (HRQL). LOLEs could be used to capture a much broader set of disease impacts, as relevant to an agency’s normative objectives. LOLE would only be limited by whether the concept was relevant to patients, allowing for the inclusion of a broader conception of HRQL (e.g., capturing the tolerability of treatments) and even a perspective wider than HRQL (capturing treatment burdens such as mode of administration).

Table 5.

Overview of the benefits of and questions raised by LOLE

Benefits of LOLE

Questions raised by LOLE

Quantitative estimate, comparable across diseases Leverages the patient’s perspective to determine what is meaningful and weigh the different components on unmet need Can accommodate both disease severity and inequality of access/outcome dimensions of unmet need Can accommodate different components of disease severity if they are relevant to the patient An accurate representation of preferences for health states including capturing of time preference

Is the patient’s perspective the right one from which to assess unmet need? Is LOLE the right common metric in which to express unmet need? What LOLE threshold defines different levels of unmet need? Is thresholding the best method for estimating LOLE? How can LOLE be estimated for more complex disease profiles? Can the implementation of LOLE be made more efficient by standardizing the implementation of LOLE across decision makers and conceiving of more generalizable ways to estimate LOLE?

LOLE length-of-life equivalent

Second, LOLE is more sensitive to the impact of a disease on patients. While regulatory agencies acknowledge the role of PED in understanding unmet need, HTA agencies use standard HRQL metrics and general population preferences. The attributes of a PP study could include the HRQL domains that matter to patients. This could be either a condition-specific HRQL instrument or standard measures of the concepts that matter to patients, such as those from PROMIS [29]. The use of the latter would further enhance the cross-patient comparability of LOLE. Further, the use of PP, rather than the general population preference set used in the utility tariffs on which QALYs are calculated, ensures that LOLE reflects the patients’ experience of the disease. In addition, the characteristics of the patient’s experience in the threshold exercise could be provided to the patient and capture the mean or expected characteristics of living with the disease. It could also be elicited from the patient such that the patient is expressly comparing their own experience with the likely experience for an age- and sex-matched healthy person.

The proposal to use LOLE rather than an QALY-based estimate of AS evokes the 1990s debate over the relative merits of the QALY and Healthy Years Equivalent (HYE). While the QALY approach involves the valuation and aggregation of discrete health states, like LOLE, HYE involves the valuation of an entire health profile [30]. Arguments for HYE included that it better represented preferences for health states, including capturing time preferences and avoiding restrictive assumptions, such as additive separability—permitting the rate of trade-off between life years and quality of life to depend on the life span—and incorporating attitudes toward risk [30–33]. Rejection of the claim that HYE incorporates attitudes toward risk and the intractability of separately valuing all health state sequences [34–36] caused health economists to lean toward the greater generalizability of the QALY approach. Recent efforts to breathe new life into HYE have focused on justifying the greater investment required to implement HYE by substantiating its ability to generate better preference data [37]. However, this tradeoff between quality of preference data and the effort of separately valuing health states is less relevant to the consideration of LOLE for measuring unmet need. As the focus is on patients’ valuations of their unmet need, LOLE would necessitate a bespoke analysis of each health state with the relevant patient group, and the desire for greater generalizability becomes redundant.

The acceptability of LOLE as a measure of unmet need raises normative questions that HTA and regulatory agencies will need to consider. First, who should determine whether unmet need is meaningful? LOLE turns to the patient for this insight, which is consistent with efforts to implement PFDD; however, HTA agencies have conventionally looked to the public to support value judgements. This rests on the principle that the public are both potential recipients and funders of healthcare [38] and that patients adapt to their condition [39] which would result in an underestimation of unmet need. Arguments for instead leveraging a patient’s perspective point to the public being unable to conceptualize some health states [38–41] and, contrary to concerns that patients downplay their needs, there is evidence that in some situations, patients put greater weight on their needs than the public [42].

Second, is length of life an appropriate numeraire? Length of life was selected as the common metric as this aligns with the approach adopted by many HTA agencies in their application of TTO. However, this will result in patients’ valuation of losses in quality of life being a function of their life expectancy. The shorter a patient’s life expectancy, the more value they will put on a unit increase in their life expectancy [43]. Thus, any given reduction in symptom burden would generate fewer LOLE for a patient with a shorter life expectancy than a patient with a longer life expectancy—as they attach more importance to improving life expectancy, they are less willing to give up life expectancy for reducing symptom burden. It would be important for agencies to consider the normative acceptability of this feature of the LOLE approach.

Third, LOLE can capture a comprehensive set of patient-relevant disease impacts, but it cannot capture some of the more societal-level disease impacts included in some frameworks, such as disease prevalence or environmental impacts [1].

The LOLE approach also raises practical questions, the answer to which will depend on each agency’s goals for estimating unmet need. First, which unmet need profiles are amenable to estimating LOLE with thresholding? The illustrations of unmet need used in this paper will be simpler than those experienced by most patients, assuming that patients’ quality of life decrement is constant over time. Further work could usefully review the experience of vignette-based preference elicitation, including that associated with HYE, to understand patients’ cognitive limitations when conducting such exercises. Further work should also assess patients’ own understanding of their unmet need profile and test the extent to which patients require supporting their disease profile with detailed descriptions.

Second, to ensure LOLE is genuinely patient centric, we recommend that patients inform the dimensions of unmet need that are included in the disease profile. Leveraging existing good practice guidance [10, 14], further guidance could usefully clarify good practice in eliciting patients’ input into disease profile generation.

Third, is thresholding the appropriate method to use to estimate LOLE? HTA agencies conventionally turn to TTO or SG methods to elicit similar trade-offs. Several features of LOLE could be replicated with the TTO, including the estimation of X in Fig. 1, the use of patients’ preferences, the incorporation of non-health benefits, and the personalization of the health states been assessed. However, there are differences, including that LOLE uses a choice method while TTO is a matching method, and that the tariff generated by TTO requires constraining between 0 and 1. Further work on the relative merits of these methods should be undertaken.

Fourth, would patients be willing to accept tradeoffs of life expectancy for improvements in quality of life? What should be done with patients who are unwilling to make this tradeoff? Would it be appropriate to assess their unmet need as being zero, or would this phenomenon necessitate the selection of an alternative numeraire?

Fifth, operationalizing LOLE will require agencies to select the LOLE thresholds that define different levels of unmet need. Further, where unmet need estimates are used to determine cost-effectiveness thresholds, agencies will need to define the relationship between LOLE and willingness to pay for health gain.

Finally, operationalizing LOLE for each decision will require resources, most likely those of sponsors. While LOLE would replace a variety of other efforts to estimate unmet need, further work could usefully consider whether it is possible to implement LOLE more efficiently by standardizing the use of LOLE across decision makers or developing more generalizable estimates of LOLE.

Among other uses, LOLE is an alternative way to estimate AS used by HTA agencies. However, agency use of AS to weight QALYs gained has been criticized for reducing the weighted population health [44] by inflating the cost-effectiveness thresholds further over supply-side thresholds. Solving this problem would require agencies to either adopt a net-benefit approach in which weights are applied to both health gains and expected health forgone, or that the policy threshold be adjusted such that the difference between it and the supply-side threshold is constrained by the ratio of weights for health gained and health forgone. If agencies would be willing to update their decision-making framework accordingly, the implication for LOLE would be that it would need to be estimated for both the health gain and the health forgone, with the practical obstacle of needing to survey the patients in receipt of the technologies that would be disinvested. However, supporting the use of unmet need as a cost-effectiveness threshold weight is only one of the applications for a standardized patient-centric estimate of unmet need. As there are many other uses of unmet need across healthcare decisions, such as supporting the incorporation of unmet need into regulatory decisions, so there are many applications of LOLE.

Conclusions

This paper proposes that a thresholding method could be used to generate a standardized, PP-based, quantitative unmet need estimate, LOLE. LOLE has several advantages over the current QALY-based measures of AS, but it also raises normative and practical questions that need to be considered by decision makers. Further research should be undertaken to compare LOLE with existing unmet need measures, assess the validity of LOLE collected under different circumstances, such as varying complexities of disease profile and varying levels of experimental design detail, and test decision-maker’s acceptance of LOLE.

Declarations

Funding

No funding was received.

Conflicts of interest

K.M. is an employee of Evidera. R.R. and L.N. are employees and shareholders of GSK. B.H. and S.W. are employees of Pfizer. O.A.E. is an employee of the National Health Council. The National Health Council is a not-for-profit, membership organization. It is supported through membership dues and sponsorship funds. The complete list of members and sponsors is located on our website at www.nationalhealthcouncil.org.

Availability of data and material

Not applicable.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication:

Not applicable.

Code availability

Not applicable.

Authors’ contributions

All authors were involved in conceptualizing the manuscript and its writing.