Value of monitoring negative emotional bias in primary care in England for personalised antidepressant treatment: a modelling study

Abstract

Background

Depressed patients often focus on negative life events. Effective antidepressant therapy reverses this negative emotional bias (NEB) within 1 week. Clinical therapeutic effect usually requires 4–6 weeks. The value of implementing NEB monitoring for the personalisation of antidepressant therapy is unknown.

Objective

To estimate the likely outcome and cost consequences of adopting the P1vital Oxford Emotional Test Battery (ETB) for this purpose in routine primary care in England.

Methods

A hybrid decision analytic model (decision tree plus Markov model) was developed to estimate the cost-effectiveness of ETB monitoring versus no ETB over 52 weeks using quality-adjusted life years (QALYs). Differences in depression severity, episode type and analytical perspectives were considered. Input data were derived from relevant guidelines, literature, national databases, expert opinion and the developers for the year 2013. Multiple sensitivity analyses addressed uncertainty.

Findings

The mean number of ETB tests is 2.162 per newly diagnosed patient and 2.166 per patient with recurrent depression. The incremental cost-effectiveness of ETB versus ‘no ETB’ is £4355/QALY from the healthcare perspective. From the broader societal perspective, ETB is more effective and cost saving.

Conclusions

Monitoring negative emotional bias in primary care in England for personalised antidepressant treatment using ETB seems as an effective and cost-effective option under all considered scenarios (including worst case). Its main economic value seems to lie in reduced productivity loss as opposed to healthcare savings.

Clinical implications

The test supports accelerated application of evidence-based depression care. Further optimisation and implementation in the ongoing European PReDicT trial is ongoing.

Background

Disease and economic burden of depression

Depression refers to a range of mental conditions characterised by persistent low mood, absence of positive affect (loss of interest and enjoyment in ordinary things and experiences), and a range of associated emotional, cognitive, physical and behavioural symptoms. Symptoms occur on a continuum of severity, and day-to-day functioning is often impaired. Major depressive episodes may be mild, moderate, or severe.

Depression is the most common psychiatric disorder and produces greater disease burden than other common chronic diseases such as cardiovascular disease, arthritis, asthma or diabetes.¹ Depression was ranked as the single largest contributor to global years lived with disability in 2015 (7.5%). Depression is also the major contributor to suicide deaths.² In England, the Psychiatric Morbidity Survey 2014 reported a prevalence rate of 39 per 1000 adults for depression or anxiety with somewhat higher rates for women compared with men.³

Due to its high prevalence and treatment costs, its role as an important risk factor for suicide and its impact on workplace productivity, depression also presents an enormous burden on the healthcare system and the wider economy. For the year 2000, the economic burden of adult depression in England was estimated at over £9 billion. Direct treatment costs were estimated at £370 million, of which 84% was attributable to antidepressant medication.⁴ The lost productivity costs of depression were estimated to be far greater than its direct treatment costs, with costs reaching £8 billion due to morbidity and £562 million due to mortality. Another study estimated the number of people suffering from depression rising to 1.45 million by 2026 in England. The relevant total cost of services for depression in England was projected to be £3 billion in direct treatment costs and £12.2 billion in lost productivity costs for the year 2026.⁵ In comparison with 36 million items prescribed in 2008, the continuous increase in the number of prescriptions dispensed for antidepressants over time with numbers reaching almost 71 million in the year 2018 also supports the growing prevalence of treated depression in the UK.⁶

Antidepressant treatment response and negative emotional bias

Within the STAR-D study, the largest existing real-world study looking at sequenced treatment options to relieve depression, overall 20% of the patients had no response to the first antidepressant medication (citalopram) including a cycle of dose increase when needed due to initial no response. Respectively, 33% of the patients achieved full remission, whereas 47% responded to treatment but did not achieve full remission.⁷ These data confirm that many patients treated with antidepressants for major depression may require two or more cycles of antidepressant therapy before successful treatment, thus effective treatment can be delayed for months. The STAR-D study also showed that while 93% of patients with a positive treatment outcome continued antidepressant therapy beyond 8 weeks, treatment discontinuation was much higher (36%) for patients with no response. Discontinuation of the given antidepressant treatment due to drug intolerance during the acute treatment phase was 11% for patients with no response, but less than 2% for patients with a positive treatment outcome.⁷ Similarly low rates of treatment discontinuation (2%) were found at 12 weeks for patients with positive antidepressant treatment outcomes in a trial by Romera and colleagues.⁸

It is expected that stratification of patients with earlier matching to effective antidepressant treatment significantly reduces the length of treatment cycles, the potential side effects experienced by patients, and the time to achieve relief from depression enabling patients to return to a normal quality of life and productivity much earlier. Depressed patients often focus on negative life events that maintain a depressed mood known as negative emotional bias (NEB). Effective antidepressant therapy reverses NEB within 1 week, whereas the clinical therapeutic effect, improvement of mood, usually requires 4–6 weeks of treatment.^{9 10} The P1vital Oxford Emotional Test Battery (ETB) is a suite of cognitive tests specifically developed to detect early changes (after 1 week) in NEB that are indicative of successful treatments for depression.

Objective

Implementing NEB monitoring for the personalisation of antidepressant therapy for major depressive disorder (MDD) requires robust a priori evidence on its clinical and cost-effectiveness. This study aimed to estimate the likely outcome and cost consequences and overall value of adopting the ETB in routine primary care in England using modelling techniques.

Methods

To achieve the above objectives, a decision analytic model was developed for the year 2013. In the lack of any other similar monitoring tool, ETB implementation was benchmarked against the standard therapeutic pathway of depression in NHS primary care, based on relevant clinical guideline recommendations supported by expert information via interviews from two psychiatrists (GMG and CJH), 12 general practitioners (GPs) and an online GP survey (n=50).¹⁰ For the latter, all participating GPs had to have been qualified for at least 6 years and spend 70% or more of their time clinically managing patients. All GPs were (1) personally treating or actively managing a minimum of 50 patients with depression per month (of which a minimum of 10% were patients newly presenting with depression), (2) based in England, (3) sampled from GP practices with a minimum of five full time equivalent GPs and (4) with equal regional distribution.

Model structure

The usual antidepressant care pathway of adults presenting with a new episode of MDD and starting antidepressant treatment in NHS primary care in England was modelled over 52 weeks with and without ETB monitoring using Microsoft Excel for base year 2012/2013. Only primary care patients with mild, moderate or severe depression at no increased risk of death due to depression were considered as relevant for the model. The model was split into two sections. The initial acute antidepressant treatment pathway was illustrated in the form of a decision tree model (figure 1). Following best practice recommendations, the model simulated up to a maximum of two different lines of antidepressants given initially in a starting and then in an increased dose when necessary allowing a maximum of four, 4-weekly cycles of antidepressant therapy before referral of treatment-resistant patients to secondary care. In case of no response to the first-line antidepressant in a starting dose, patients underwent a step of dose increase before switching to a second-line antidepressant. Adjunctive drug therapy in primary care was not considered.¹⁰ Clinical response was achieved after 4 weeks of a given treatment with full or partial remission after 8 weeks of treatment. Patients with full or partial remission entered a Markov simulation model. In a Markov model, patients are classified into a number of different health states, each state associated with a certain cost and outcome. Over a defined time period called the Markov cycle, patients can move between different health states according to a set of transition probabilities.^{11 12} The Markov model had weekly cycles and followed patients through a series of health states corresponding to full remission with continuation treatment, full remission without continuation treatment, partial remission with continuation treatment, partial remission without continuation treatment, and relapse. Due to the given patient cohort and time horizon of the model, death was not modelled as a separate outcome.

Figure 1.

Simplified model structure for (A) acute antidepressant treatment (decision tree) and (B) continuation antidepressant therapy (state transition Markov model). AD, antidepressant.

Model parameters and assumptions

Intervention

The ETB was modelled as a stand-alone monitoring test to detect early change in NEB in adults newly presenting with MDD and starting antidepressant therapy in primary care. It was assumed that all eligible patients undergo an initial ETB assessment before starting antidepressant therapy and have a follow-up ETB assessment after 1 week. In case of a positive ETB test result (predicted response to treatment), the given antidepressant regime continues with final treatment response being judged clinically as part of a routine GP follow-up visit, and no further ETB tests would be carried out. In case of a negative ETB test result indicating ‘no response’ to the first antidepressant regime, an increase in dose is implemented followed by a second follow-up ETB test after 1 week. In case of a second negative ETB test result, the patient is expected to be changed to a second-line antidepressant. Antidepressant therapy changes following a negative ETB test result required additional visits. For second-line antidepressants, no further ETB tests would be carried out. This limit on the use of the ETB test use was implemented based on expert advice to acknowledge the specificity limitations of the ETB and account for some uptake limitations. The sensitivity and specificity of the ETB were based on the analysis of existing experimental data provided by the developers of ETB. All follow-up ETB tests were assumed to have the same prognostic characteristics. In the base case, it was assumed that only the initial ETB assessment would require the presence and help of a GP nurse (on average 10 min per patient), while follow-up ETB tests would be fully self-administered. The impact of this assumption as opposed to nurse help with all ETB tests was tested in a sensitivity analysis. Although no ETB tests were assumed for relapse during continuation treatment, patients newly presenting with recurrent depression were considered the same as newly diagnosed patients in terms of eligibility for ETB tests.

Effects

As far as available, all effect parameters were based on published evidence. Input data were synthesised from published model-based economic evaluations for depression as reported in the systematic literature review by Haji Ali Afzali et al supplemented with rapid focused reviews in PubMed where necessary.¹³ Effect studies were selected based on their robustness, relevance of applied methods and generalisability to the current context. Where effect estimates could not be identified from the literature, expert opinions were used or relevant information was collected as part of the online GP survey.

Response rates to the different cycles of antidepressant treatment were derived separately for mild, moderate and severe depression from the online GP survey. A set of alternative values for mild and moderate depression were explored in the sensitivity analysis using data from the STAR-D study^{7 14} and a study by Nordström and colleagues.¹⁵ Relevant response rates for severe depression were reduced to 60%.¹⁶ Secondary care treatment response rate was based on expert opinion.

In the case of the most commonly used antidepressants, treatment discontinuation due to intolerance was deemed negligible and was not modelled. Discontinuation due to non-response was not assessed separately either, but patients who did not respond to a given cycle of antidepressant treatment automatically entered the next cycle of treatment. Longer-term treatment discontinuation by non-depressed patients was, however, of major concern. Following full or partial remission and an initial 4-week continuation treatment, relapse and treatment discontinuation was simulated in the Markov model on a weekly basis. The risk of treatment discontinuation was not expected to be different between full and partial remission; therefore, the same transition probability was applied for both patient cohorts. The relevant estimate was taken from the literature and transformed to weekly transition probabilities for the Markov simulation.¹⁷ Patients who discontinued treatment remained in the relevant health state unless relapse happened.

There is evidence that the risk of relapse for patients with full remission is lower than the risk of relapse for patients with partial remission.¹⁶ Similarly, the risk of relapse is anticipated being lower for patients on continuation treatment than for patients who previously discontinued their antidepressant medication. Therefore, relevant relapse risk estimates were sought separately from the literature and transformed to weekly transition probabilities to match the cycle length of the Markov model as needed.^{18 19} The Markov model did not allow relapsed patients to remit. In the lack of any relevant published evidence, the severity of relapse was kept the same as the severity of the initial presentation of depression. This was deemed as a conservative estimate by our experts preventing the overestimation of the impact of ETB.

The cost-effectiveness analysis used the quality-adjusted life years (QALYs) as its main outcome measure. To be in line with the National Institute for Health and Clinical Excellence’s (NICE) recommendation, utility values derived from the EuroQoL EQ-5D, a generic, health-related quality-of-life questionnaire, were preferred.^20–22 All three included studies were conducted in primary care in Europe.^{16 23 24}

Costs

Resource use parameters were either based on the literature or on expert opinion and valued using national-level unit costs taken from national sources (eg, British National Formulary, Personal Social Services Research Unit’s collection of ‘Unit Costs for Health and Social Care’).²⁵ The costs of the first-line and second-line antidepressant regime in the model were based on the average of the most frequently prescribed antidepressants in England for the base year.²⁶ The cost of secondary care antidepressant therapy was based on venlafaxine. Generic unit costs were used in all cases. The frequency of GP visits and antidepressant prescriptions were driven by expert advice and the relevant NICE depression guideline recommendations.¹⁰ It was assumed that the initial assessment visit for antidepressant therapy would last on average 17 min. Antidepressants would be prescribed by GPs in monthly instalments with routine follow-up visits of 11 min scheduled every 4 weeks. Secondary care antidepressant therapy involved an initial assessment visit carried out by a consultant psychiatrist over 45–60 min followed by 4-weekly visits by a specialist registrar over 15–30 min (psychiatrist expert opinion). The impact of secondary care costs within these ranges was subject to sensitivity analysis. Annual estimates for work days lost due to depression were taken from the GP survey and transformed to weekly rates. In the lack of a market value, the price of ETB test was set at £0.

The cost of a work day lost was estimated according to the average wage rate in England.²⁷ An alternative minimum wage rate approach was explored in the sensitivity analysis.²⁸

The model structure, assumptions and parameters were extensively discussed and agreed within the team and validated by external clinical experts in a face-to-face meeting. All input parameters, their base case and sensitivity analysis values and source references are listed in table 1.

Table 1.

Model input parameters

Parameters	Values			References
	Base case	Minimum	Maximum
Probabilities
ETB sensitivity	0.957	0.870	1.000	P1vital
ETB specificity	0.400	0.610	0.190	P1vital
Response to first-line standard dose AD, moderate depression	0.530	0.600		GP survey7 14
Response to first-line standard dose AD, severe depression	0.670	0.600		GP survey7 14
Response to first-line standard dose AD, mild depression	0.350	0.360		GP survey; 60% of mild/moderate response rate16
Response to first-line increased dose AD, moderate depression	0.460	0.500		GP survey7
Response to first-line increased dose AD, severe depression	0.610	0.500		GP survey7
Response to first-line increased dose AD, mild depression	0.320	0.300		GP survey; 60% of mild/moderate response rate16
Response to second-line standard dose AD, moderate depression	0.510	0.410		GP survey15
Response to second-line standard dose AD, severe depression	0.630	0.410		GP survey15
Response to second-line standard dose AD, mild depression	0.380	0.246		GP survey; 60% of mild/moderate response rate16
Response to second-line increased dose AD, moderate depression	0.460	0.342		Identical base value to first-line AD (expert opinion)
Response to second-line increased dose AD, severe depression	0.610	0.342		Identical base value to first-line AD (expert opinion)
Response to second-line increased dose AD, mild depression	0.320	0.205		GP survey; 60% of mild/moderate response rate16
Response to secondary care AD therapy	0.400			Expert opinion
Full remission after response	0.410			7
Relapse, full remission no therapy (weekly)	0.017			18
Relapse, partial remission no therapy (weekly)	0.023			19
Relapse, full remission AD therapy (weekly)	0.007			18
Relapse, partial remission AD therapy (weekly)	0.009			19
Discontinuation, full remission (weekly)	0.044			17
Discontinuation, partial remission (weekly)	0.044			Same as for full remission (expert opinion)
Costs (£, 2011/2012)
First-line AD, standard dose (per week)	0.51			British National Formulary (BNF)
First- line AD, increased dose (per week)	0.76			BNF
Second-line AD, standard dose (per week)	0.51			BNF
Second-line AD, increased dose (per week)	0.76			BNF
Secondary care AD (per week)	0.71		0.94	BNF
Initial GP visit	44.00			25
Follow-up GP visit	30.00			25
Practice nurse, initial ETB test	7.17	0.00	7.17	25
Practice nurse, follow-up ETB test	0.00	0.00	7.17	25
Initial secondary care visit	264.75		353.00	25
Follow-up secondary care visit	45.75		91.50	25
Relapse treated in primary care (per week)	8.01		8.26	25
Relapse treated in secondary care (per week)	12.15		23.81	25
Utilities
Full remission, antidepressant therapy	0.810	0.680	0.940	16
Full remission, no therapy	0.850	0.720	0.980	23
Partial remission, antidepressant therapy	0.720	0.520	0.920	23
Partial remission, no therapy	0.760	0.560	0.960	Same improvement as in the case of full remission23
Mild depression	0.600	0.540	0.650	24
Moderate depression	0.460	0.300	0.480	24
Severe depression	0.270	0.210	0.340	24
Relapse	0.600	0.540	0.650	Same QoL impact as initial episode (expert opinion)
No of work days lost (per week)
Moderate depression	0.5		1.5	GP survey
Mild depression	0.2		0.6	GP survey
Severe depression	1.7		4.8	GP survey
Full remission, antidepressant therapy	0			Expert opinion
Full remission, no therapy	0			Expert opinion
Partial remission, antidepressant therapy	0			Expert opinion
Partial remission, no therapy	0			Expert opinion
Value of a work day saved (£)	119	48.64		27 28
Willingness to pay threshold (£/QALY)	30 000	20 000	30 000	20 22

AD, antidepressant; ETB, P1vital Oxford Emotional Test Battery; GP, general practitioner; QALY, quality-adjusted life year; QoL, quality of life.

Cost-effectiveness analyses

An incremental cost-utility analysis was carried out. Results were expressed in the form of incremental cost-effectiveness ratios (ICERs) defined as the difference in total costs divided by the difference in QALYs between the current antidepressant treatment pathway (no ETB) and the expected antidepressant pathway with ETB. Using the current UK cost-effectiveness threshold of £20 000–30 000£/QALY,^{20 22} results were also presented according to the net benefit framework. Costs were calculated in British pounds using national-level unit costs for the year 2011/2012. Discounting of costs and/or outcomes was not necessary. The base-case analysis was carried out from the healthcare’s perspective. The likely impact of ETB on productivity losses due to absenteeism from work, however, was also assessed and relevant cost-effectiveness estimates were presented from a broader societal perspective in an alternative scenario. The model was run separately for patient cohorts with mild, moderate and severe depressions. Results were then combined into a joint net benefit estimate separately for newly diagnosed and recurrent patients according to their relevant severity distributions derived from the online GP survey. An overall cost-effectiveness estimate for these two cohorts was also calculated using a 50%:50% ratio between newly diagnosed and recurrent patients (GP survey estimate). An additional threshold price analysis was also carried out.

Uncertainty in the cost-effectiveness results due to key parameters was investigated in extensive sensitivity analyses and within a best-case–worst-case scenario analysis. Investigated parameters included the sensitivity and specificity of ETB, response rates to different antidepressant therapy cycles, applied utility estimates, the level of nurse input to ETB test provision, the cost of secondary care antidepressant treatment, and the value and number of work days saved (see online supplementary table 1).

Findings

Model structure

The general model structure is presented in figure 1 including the Markov state transition diagram. The established initial primary care pathway and treatment probabilities of patients presenting with newly diagnosed or recurrent depression are shown in figure 2.

Figure 2.

Initial primary care pathway for patients presenting with (A) newly diagnosed depression and (B) recurrent depression.

Model parameters and assumptions

Based on the available experimental data, the sensitivity of the ETB was estimated at 0.96 (95% CI 0.87 to 1), while the specificity of ETB was 0.40 (95% CI 0.19 to 0.61). The model results indicated that the mean number of ETB tests per patient was 2.12 for mild depression, 2.16 for moderate depression and 2.23 for severe depression, resulting in an average 2.162 ETB tests per newly diagnosed patient and 2.166 ETB tests per patient with recurrent depression. The five most commonly prescribed antidepressants in England were citalopram, amitriptyline, fluoxetine, mirtazapine and sertraline in the base year.²⁶ From these, amitriptyline was deemed not relevant anymore for newly initiated antidepressant treatment in NHS primary care according to guideline recommendations. Therefore, the average generic price of the other four drugs was used to calculate the weekly cost of first-line and second-line antidepressant medication in primary care (standard dose, £0.51; increased dose, £0.76). The weekly cost of secondary care antidepressant medication was estimated at £0.71 in standard dose and £0.94 in increased dose.

Cost-effectiveness analyses

According to the base-case scenario, the incremental cost-effectiveness of ETB testing versus ‘no ETB’ for the monitoring of antidepressant therapy in NHS primary care is £4355/QALY from the healthcare’s perspective (£4463/QALY for newly diagnosed patients and £4254/QALY for recurrent patients). From the broader societal perspective, ETB is both more effective and cost saving, and therefore dominates the ‘no ETB’ option. Further results by depression severity expressed both as ICERs and net benefits (at a £30 000/QALY willingness-to-pay threshold) can be found in table 2.

Table 2.

Base case cost, effect and cost-effectiveness results

Patient cohort	Alternative	QALY	Difference	WDL	Difference	Costs	Difference	ICER1	NB1	ICER2	NB2
Mild depression	ETB	0.7235	0.0008	2.8326	−0.0456	287.57	8.87	10 531	16	4092	22
	No ETB	0.7226		2.8782		278.70
Moderate depression	ETB	0.6742	0.0021	8.0341	−0.1675	320.69	9.99	4661	54	ETB dom	74
	No ETB	0.6721		8.2016		310.70
Severe depression	ETB	0.5769	0.0046	34.1694	−0.7836	394.83	12.03	2596	127	ETB dom	220
	No ETB	0.5723		34.9531		382.80
New diagnosis	ETB	0.6693	0.0023	11.8236	−0.2570	325.76	10.06	4463	58	ETB dom	88
	No ETB	0.6671		12.0806		315.70
Recurrent episode	ETB	0.6640	0.0024	12.9020	−0.2824	329.62	10.18	4254	62	ETB dom	95
	No ETB	0.6616		13.1844		319.44
Overall	ETB	0.6667	0.0023	12.3628	−0.2697	327.69
	No ETB	0.6644		12.6325		317.57	10.12	4355	60	ETB dom	92

ETB dom, ETB dominates the ‘no ETB’ alternative; ETB, P1vital Oxford Emotional Test Battery; ICER1, incremental cost-effectiveness ratio from the healthcare’s perspective; ICER2, incremental cost-effectiveness ratio from the societal perspective; NB1, net benefit from the healthcare’s perspective; NB2, net benefit from the societal perspective;QALY, quality-adjusted life year; WDL, work days lost.

Results of the sensitivity analysis are shown in online supplementary table 1. Using the most attractive estimates (best-case scenario), the cost-effectiveness of ETB improved to £3217/QALY from the healthcare’s perspective and was even more dominant from the societal perspective. Applying the least attractive estimates (worst-case scenario), the cost-effectiveness of ETB reduced to £19 311/QALY and £12 378/QALY, respectively, but did not reach the current lower cost-effectiveness threshold of £20 000/QALY in the UK. For the online ETB, the worst-case scenario estimates were £10 184/QALY and £3250/QALY, respectively. One-way sensitivity analysis revealed that the cost-effectiveness estimates are most sensitive to (1) the assumed level of nurse input to the provision of ETB testing (£7945/QALY), (2) the prognostic characteristics of ETB (£6646/QALY), (3) the treatment response estimates (£5684/QALY) and (4) the applied utility values (£3217/QALY–£5217/QALY).

Discussion

The current analysis looked at the likely outcome and cost impacts of ETB implementation as a monitoring test of antidepressant therapy in NHS primary care in England. The 1-year time horizon of the analysis reflects that the ETB is not expected to have any direct impact on the natural course of depression. The impact of ETB is rather seen as a change in the initial antidepressant treatment pathway and the consequent reduction in time spent on ineffective antidepressant therapy, an accelerated response and remission process, and gains in the quality of life and work productivity of patients. Although the implementation of the ETB is also likely to result in an accelerated referral of non-responder patients to secondary care treatment, this may be seen as another positive change in patient care. Further positive effects may include the reduction of initial treatment discontinuation due to intolerance and non-response. In the lack of any supporting data, this later aspect was not built into the current model. Instead, the structure, assumptions and input values of the model were chosen to be conservative and worst-case/best-case sensitivity analysis was carried out.

The results promote the ETB as a potentially good-value-for-money monitoring test with cost-effectiveness estimates ranging from £3217/QALY from the healthcare’s perspective as best-case scenario (dominating the ‘no ETB’ option from the societal perspective) to £19 311/QALY as worst-case scenario (£3606/QALY from the societal perspective) at a price of £0 per ETB test. The model highlights that the main economic value of such personalised antidepressant therapy rather lies in earlier return to work and reduced productivity loss as opposed to savings in the healthcare system. Depending on the level of staff input required to ETB testing, the analytical perspective and the societal willingness-to-pay per additional QALY gained, the economic threshold price of an ETB test was shown to vary between £12.5 and £41.5.

Limitations

Model parameters were based on published evidence as much as possible and all assumptions and input parameters were externally validated by experts. Full external validation of the model structure by building an independent comparative model, however, was not feasible. Due to the complexity of the model and the available resources, a systematic review of the literature for all model parameters, however, was not feasible. The model is based on cost information for the year 2011/2012. Although updating this information would result in some numerical changes in the final cost and cost-effectiveness estimates, no impact on the overall results and conclusions is expected since none of the input unit costs have changed considerably since. Despite the extensive sensitivity analysis carried out on individual parameters, considerable uncertainty still remains about the sensitivity and specificity of repeated ETB testing by depression severity, the feasibility and acceptability of ETB by patients and doctors, and the broader impact of ETB testing on demand for therapist provided psychosocial interventions. Prospectively collected data on relevant utility values, treatment discontinuation, response rates and productivity data would further increase the robustness of the value estimates. As the model was based on depression care and data validated for England, results should be treated with caution when trying to extrapolate beyond the current context for example to other countries or to other disease areas such as bipolar disorders.

Clinical implications and conclusions

The analysis revealed that the implementation of the ETB could gear clinical practice towards accelerated application of current evidence-based depression care. Following the modelling study, the ETB has been further optimised into the P1vital PReDicT Test, a CE-marked Class I medical device. The test uses a proprietary machine learning algorithm to provide a rapid measure of a patient’s response to an antidepressant and guide potentially necessary medication changes to accelerate their recovery from the illness. To address further the above uncertainties, a large-scale European study, the PReDicT Project (Predicting Response to Depression Treatment), is currently ongoing in five countries (UK, Germany, Spain, France and the Netherlands) to assess the clinical and cost-effectiveness and acceptability of using the PReDicT Test to guide the antidepressant treatment of depressed patients in primary care.²⁹ Cost-effectiveness results over 6-month follow-up are expected to be available from the study by the end of 2019 and will be used to cross-validate the findings of our modelling study.

As demonstrated by the current analysis, to achieve large-scale impact and the required cost-effectiveness threshold, the PReDicT test needs to have high acceptance rates both by the patients and the healthcare professionals. Furthermore, since the main economic benefits of such a personalised accelerated pathway for antidepressant therapy are likely to fall on the employment sector, the study highlights the importance of conducting and considering economic value assessments from a broad societal perspective including potential intersectoral costs and benefits.