Cost-effectiveness of the cell-based quadrivalent versus the standard egg-based quadrivalent influenza vaccine in Germany

Abstract

Background

Standard influenza vaccines are produced using egg-based manufacturing methods. Through the process, the resulting egg-adapted viral strains may differ from the selected vaccine strain. Cell-derived influenza vaccine manufacturing prevents egg-adaptation of the antigen which can improve vaccine effectiveness. We evaluated the cost-effectiveness of quadrivalent cell-derived influenza vaccine (QIVc) versus an egg-based quadrivalent influenza vaccine (QIVe) in preventing seasonal influenza from German societal and payer perspectives.

Methods

Adapted version of the individual-based dynamic 4Flu transmission model was combined with a decision-tree to calculate the impact of QIVc versus QIVe on influenza over 20 seasons in Germany. Egg-adaptation, resulting in lower effectiveness of QIVe versus QIVc towards the H3N2 influenza strain, is sourced from a US retrospective study and assumed in 100% (base case) or 55% (conservative scenario) of years. Influenza-related probabilities of outpatient visits, hospitalizations, productivity loss, and mortality, with associated (dis)utilities/costs, were extracted from literature. Costs and outcomes were discounted 3.0%/year.

Results

Replacing QIVe with QIVc in subjects aged ≥ 9 years can annually prevent 167,265 symptomatic cases, 51,114 outpatient visits, 2,091 hospitalizations, and 103 deaths in Germany. The annual number of quality-adjusted life-years (QALYs) increased by 1,628 and healthcare costs decreased by €178 M from societal perspective. From payer perspective, the incremental cost-effectiveness ratio was €2,285 per QALY. Scenario analyses confirmed results robustness.

Conclusions

The use of QIVc compared to QIVe, in the German Immunization Program, could significantly prevent outpatient visits and hospitalizations and would enable substantial savings from a societal perspective.

Keywords:

• Cost-effectiveness
• influenza
• vaccination
• cell-derived
• vaccine effectiveness

JEL CLASSIFICATION CODES:

• C50
• C51
• C53

Introduction

There are two main types of influenza viruses that commonly cause seasonal influenza: Types A and B. Influenza A viruses are further classified into subtypes H1N1 and H3N2. Influenza B viruses are broken down into lineages B/Yamagata and B/Victoria¹. The most common symptoms of influenza are chills, fever, sore throat, muscle pains, headache (often severe), coughing, weakness/fatigue, and general discomfort¹. Annually, an estimated three to five million influenza-related hospitalizations and about 250,000–500,000 influenza-related deaths are reported globally¹. In Germany, 3.8 million general practitioner (GP) visits and 18,000 hospitalizations were reported in the 2018–2019 season due to influenza². Influenza can occasionally lead to pneumonia, and several rare but severe complications such as myocarditis, myocardial infarction, stroke, and gastrointestinal bleeding³. All age groups can be affected by influenza; however, elderly people (>65 years of age), young children (<2 years of age), and patients with comorbidities such as chronic lung, heart, and renal diseases, and immunodeficiency are more prone to develop complications and are considered as high risk groups⁴. In Germany, quadrivalent influenza vaccine (QIV) is recommended for all people aged ≥60 years and for high risk groups⁵.

The influenza virus needs to bind to a cellular receptor in order to infect a cell. The influenza viruses grow in humans as well as in other mammals. However, nearly all seasonal influenza vaccines are produced using candidate vaccine viruses (CVVs) grown in chicken eggs⁶. Avian cells have different receptors than those that are on the surface of human or other mammalian cells. For a human influenza virus to grow well in avian cells, it needs to adapt to bind the avian receptor in a process known as egg-adaptation⁷. This adaptation during the isolation and propagation steps of egg-based vaccines antigen may lead to a mismatch with the World Health Organization (WHO) selected influenza vaccine strain^8,9. As a result, the standard egg-adapted trivalent and quadrivalent influenza vaccines (QIVe) have shown lower vaccine effectiveness (VE) in recent years, especially against circulating type A (H3N2) viruses^8,10. A manufacturing process that uses cell-culture technology (Madin-Darby Canine Kidney cells) and does not involve the use of fertilized eggs for influenza vaccine production was approved by the United States Food and Drug Administration (US FDA) in 2012⁷. Cell-derived manufacture prevents changes in the antigenic structures due to egg-adaptation and can minimize the chances of altering the influenza vaccine strain, thereby potentially increasing the VE¹¹. Flucelvax Tetra is a cell-culture based quadrivalent influenza vaccine (QIVc) developed by Seqirus and it is the first QIVc approved by the European Medicine Agency in 2018^12,13.

Previous studies on cost-effectiveness analysis (CEA) of different seasonal influenza vaccines have been reported in the US and in European countries like Italy, Spain, and the UK^14–17. However, at the moment that the German health authorities are considering the reimbursement of QIVc, a CEA on the impact of QIVc in Germany is still lacking. As epidemiological data, vaccine coverage, and costs vary from country to country, CEA results from other European countries cannot be used for Germany and thus a German-specific model is necessary. Moreover, different vaccine effectiveness was used in the aforementioned CEA studies as uncertainties exist around the vaccine effectiveness. The study explores conservative scenarios of the vaccine effectiveness of QIVc. This study aims to evaluate cost-effectiveness, from the societal and payer perspectives, for the projected health and economic outcomes associated with the use of QIVc versus standard QIVe in individuals aged ≥9 years for preventing seasonal influenza in Germany.

Methods

Model framework

Transmission model

We used an extended version of the 4Flu model developed by Epimos GmbH and GlaxoSmithKline GmbH & Co. KG¹⁸. The model simulates the concomitant and independent transmission of four different influenza viruses, i.e. A(H1N1), A(H3N2), B/Victoria, and B/Yamagata. Infected individuals pass through a latent period of 2 days and then become infectious for 2 days (adults) or 4 days (children)^18,19. During this infectious period, they can infect any susceptible person within their contact network. Keeping 2019 as the base year, the model predicts the number of infected individuals per 100,000 persons by age, year, and risk group (normal and high risk) over a time horizon of 20 years. These outputs are used to feed the cost-effectiveness model (CEM). For a more detailed description of the simulation tool, see Supplementary Online Supporting Material.

Cost-effectiveness model

This CEM was developed in Microsoft Excel based on a previously published decision-tree model by Dolk et al.²⁰. The structure of the CEM shown in Supplementary Figure S2 represents the evolution of patients infected with influenza. The model considers that some of the infected patients will become symptomatic, while others will not develop clinical signs. A proportion of the symptomatic patients would seek medical attention while another proportion of patients with symptoms would be expected to develop complications like otitis, respiratory problems (bronchitis, pneumonia, upper respiratory tract infections), cardiac complications (myocarditis, myocardial infarction), neurological issues, or gastrointestinal bleeding. Each of these complications may be treated at home or in hospital. Hospitalized complications have a probability of being fatal. These complications were included as per Dolk et al.²⁰, along with the addition of myocardial infarction (MI) and stroke. The complications considered in the model are fully detailed in Supplementary Table S1. We have performed the necessary quality assurance steps to ensure the model produces predictable and accurate outputs. The quality assurance process involves an independent review of the model by a senior health economist.

Model perspective and patient populations

As mentioned, societal and payer perspectives were considered for this CEA. Following the German Standing Committee on Vaccination (STIKO) guidelines, the societal perspective was considered as the base case²¹, wherein the vaccination costs, GP visits, treatment for symptomatic influenza, complication costs, transportation costs, patient co-payments, absenteeism costs, and costs borne by child sickness benefit were included. In addition, costs associated with the loss of productivity for the caregiver due to influenza were included. Costs data from the payer perspective focused only on the vaccination costs, GP visits, treatment for symptomatic influenza, complication costs, and costs borne by child sickness benefit.

The simulations were performed on a cohort of 100,000 individuals and the outcome (i.e. number of infections) per treatment arm was analyzed in the CEM. Results were projected by age-distribution to the entire German population (82.5 million), as the clinical and health economic outcomes accumulated over seasons for a period of 20 years (time horizon of the 4Flu model).

The time horizon for the CEM was also 20 years¹⁸. The costs and quality-adjusted life-years (QALYs) were discounted using a rate of 3% per annum after year one in compliance with the German guidelines²².

Vaccination with QIVe was compared to vaccination with QIVc in individuals aged ≥9 years and to QIVe in individuals under 9 years. A cut-off point of 9 years was used as QIVc is indicated for individuals aged ≥9 years per the label granted by the European Medicine Agency (EMA)²³.

Transmission model inputs

Vaccination for a given season primarily depends on the age and risk status of the individual. A re-vaccination factor of 4.25 was used in the model, meaning that previously vaccinated individuals are more than 4-times as likely to be vaccinated again as previously unvaccinated individuals. The re-vaccination factor was calculated from reported data on people who were vaccinated at least once in three subsequent seasons from 2004/5 to 2006/7 in Germany²⁴. Vaccination coverage depends on the individuals’ age. Vaccine coverage in the 4Flu model is shown in Supplementary Table S2. The VE depends on the age of the individual and on the type of vaccine being used. Relative VE (i.e. rVE) describes the difference between QIVc and QIVe. It was assumed that, in years without egg-adaptation, the VE of QIVc against A(H3N2) was the same as that of QIVe. For years with egg-adaptation, it was assumed that the rVE of QIVc against A(H3N2) is 36.2% higher than that of QIVe (rVE: 36.2%; 95% confidence interval [CI]: 26.1–44.9)²⁵. For a more conservative scenario, the sensitivity analysis result of the same study was used, showing an rVE of 19.3% (95% CI: 9.5–28.0)²⁵. Absolute VE (i.e. aVE) per arm is based on the rVE and on previous findings reported by Belongia et al.²⁶, resulting in a mean VE (33%) of QIVe against A(H3N2) when averaged over all years (years with and without egg-adaptation).

The transmission model used different VE values depending on the base case assumption that egg-adaptation either occurs in 100% (Setting A) or in 55% (Setting B) of all years. The 55% assumption was derived from a study of Rajaram et al.²⁷ in which they reported that there was minimal or no antigenic similarity between type A(H3N2) circulating viruses and egg-adapted reference viruses in 16 of the 29 seasons analyzed. For the other influenza strains and lineages, the same VE was used for QIVc and QIVe in all years^28–30. The age-dependent VE against A(H3N2) with and without egg-adaptation for the base case analyses is shown in Table 1.

Table 1. Vaccine effectiveness for Settings A and B in the 4Flu model.

	Setting A (e.g. g-adaptation occurs in 100% of years)					Setting B (e.g. g-adaptation occurs in 55% of years)
Age group (years)	VE against A(H3N2)			VE against other strains		VE against A(H3N2)			VE against other strains
	QIVc	QIVe		QIVc*	QIVe* (%)	QIVc	QIVe		QIVc*	QIVe* (%)
		Normal year (%)	Egg-adapted change year (%)				Normal year (%)	Egg-adapted change year (%)
0–8	NA	44.95	33.00	NA	69	NA	38.65	28.38	NA	69
9–17	44.95%	44.95	33.00	69%	69	38.65%	38.65	28.38	69%	69
18–64	44.95%	44.95	33.00	73%	73	38.65%	38.65	28.38	73%	73
65+	44.95%	44.95	33.00	62%	62	38.65%	38.65	28.38	62%	62

Abbreviations. NA, not applicable; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine; VE, vaccine effectiveness.

*Values apply to all simulations.

A mean VE (33%) of QIVe against A(H3N2) was averaged over all years (years with and without egg-adaptation) and is based on the rVE and on previous findings reported by Belongia et al.²⁶. For the other influenza strains and lineages, the same VE was used for QIVc and QIVe in all years sourced by papers of Jefferson et al.^28–30.

Cost-effectiveness model inputs

Clinical inputs

As reported by Carrat et al.³¹, we assumed that 66.9% of the infected patients developed influenza symptoms. Age-specific probabilities of seeking medical attention by symptomatic cases are shown in Supplementary Table S3. The probability of receiving antivirals was 14.7% and 19.6% in normal and high-risk groups, respectively²⁰. Probabilities of most likely influenza attributable complications included in this model were taken from Dolk et al.²⁰. The model additionally includes MI and stroke, and the probabilities were based on published literature^32–35. Emergency room (ER) visits could be required for patients with complications. These patients can be hospitalized or sent home. It was assumed that all patients that were hospitalized were admitted via ER. The percentage of patients being hospitalized and discharged from ER is shown in Supplementary Tables S4 and S5, respectively. Fatality rates with complications are shown in Supplementary Table S6.

Health-related quality-of-life inputs

The baseline utilities were taken from Dolk et al.²⁰ and are shown in Supplementary Table S7. Disutility associated with clinical influenza cases were obtained from averaging utility decrements during an influenza episode and were multiplied with the duration of influenza symptoms. Utility decrements were applied to symptomatic influenza, hospitalization related to complicated influenza, and for outpatient treatment for complicated influenza. The disutility and duration of symptomatic and complicated influenza were based on Dolk et al.²⁰, except for MI and stroke³⁵. Disutility due to symptomatic influenza was applied to both uncomplicated and complicated cases. To account for longer disease course and recovery time of hospitalized influenza patients compared to outpatient cases³⁶, disutilities for three extra ambulatory days were added to hospitalization duration. Tables 2 and 3 depict the calculation of disutility values used in the model.

Table 2. Annualized disutility due to symptomatic influenza (calculated*).

Age group (years)	Disutility influenza (no antiviral treatment)	Disutility influenza (antiviral treatment)
0–4	−0.006	−0.005
5–17	−0.006	−0.005
18–49	−0.007	−0.006
50–64	−0.007	−0.006
65–69	−0.009	−0.008
70–74	−0.009	−0.008
75–79	−0.009	−0.008
80–84	−0.009	−0.008
85+	−0.009	−0.008

*Average utility decrements (365.25 days) during an influenza episode were multiplied by duration of influenza symptoms to obtain the disutility associated with clinical influenza cases.

The annual disutility and duration of influenza were taken from Dolk et al.²⁰. The event disutilities reported here were calculated with the formula annual disutility*duration/365.25.

Table 3. Disutility due to complications (calculated*).

	Bronchitis	Pneumonia	URTI	Myocarditis	MI	Renal	CNS	Stroke	Otitis media	GI bleeding
Disutility hospitalized patients	−0.0090	−0.0116	−0.0073	−0.0085	−0.3800	−0.0120	−0.0088	−0.3800	−0.0036	−0.0053
∟Disutility hospitalization	−0.0070	−0.0096	−0.0062	−0.0074	−0.3800	−0.0105	−0.0067	−0.3800	−0.0034	−0.0044
∟Disutility of three extra ambulatory days	−0.0021	−0.0021	−0.0010	−0.0011	NA	−0.0014	−0.0022	NA	−0.0001	−0.0010
Disutility ambulatory patients	−0.0046	−0.0063	−0.0021	−0.0025	−0.1980	−0.0049	−0.0046	−0.2870	−0.0001	−0.0014

Abbreviations. CNS, central nervous system; GI, gastrointestinal bleeding; MI, myocardial infarction; NA, not applicable as we assumed one-off disutility for stroke and MI; URTI, unspecified upper tract respiratory infection.

*Average utility decrements (365.25 days) during an influenza episode were multiplied by duration of influenza symptoms to obtain the disutility associated with clinical influenza cases.

The disutilities of stroke and MI were taken from Ara et al.³⁵ and are event disutilities. The disutility and duration of the other complications were derived from Dolk et al.²⁰. The values reported were calculated using the disutility*duration/365.25²⁰.

Costs and resource inputs

All costs were updated to 2019, using an inflation rate of 1.74% and 1.85% for the years 2017 and 2018, respectively³⁷, and expressed in Euros (€). An estimated average price of €11.6 per dose was used for all QIVe³⁸. Vaccination cost for QIVc was €12.5 per dose. Other costs, taken from Dolk et al.²⁰, included vaccine administration, GP visits of symptomatic patients, antivirals, and complication treatment costs (except for MI and stroke in the ambulatory, ER and hospital settings, costs borne by child sickness benefit, transportation costs, and premature death costs).

The hospitalization costs, including patient co-payment, for MI and stroke, were derived from official German diagnosis-related groups (DRG) listings, weighted by the number of patients per disease severity³⁹. Ambulatory reimbursements and patient co-payments of MI and stroke management were sourced from the German uniform standard of evaluation (EBM, 2018), according to which outpatient services are billed⁴⁰.

Absenteeism costs were calculated based on the working days lost and the average net production per day, which was estimated with the labor costs⁴¹ and the employment rates⁴². The working days lost were assumed to be the same as the disease duration in Dolk et al.²⁰. The working days lost for stroke (34 days) and MI (36 days) were sourced from the literature⁴³. For individuals aged under 18 years, the working days lost were assumed to happen to their caregiver (only one parent was assumed to be affected).

Cost inputs are shown in Supplementary Table S8.

Cost-effectiveness analysis

Results are reported per 100,000 individuals and also projected to the German population, as explained earlier. The number of symptomatic cases, GP visits, complications, hospitalizations, and deaths were the reported clinical outcomes, whereas QALYs and incremental cost-effectiveness ratio (ICER) expressed as cost per QALY gained and net monetary benefit (NMB) were the considered economic outcomes. There is no official willingness-to-pay (WTP) threshold in Germany for CEA. In our analysis, a threshold of €30,000 per QALY gained was used in line with other published cost-effectiveness vaccines studies conducted earlier in Germany⁴⁴.

Scenario and sensitivity analyses

As already explained in earlier sections, two base case analyses were conducted (Setting A and B). To assess the impact of uncertainty in 4Flu and cost-effective model parameters on the base case results, scenario and sensitivity analyses were conducted. From a payer perspective, a scenario in which rVE was set to 19.3% instead of 36.2% was considered²⁵. Univariate deterministic sensitivity analyses (DSA) were conducted from the societal perspective only. Published lower and upper values, 95% CIs, or ±10% variations around base case were tested for epidemiological and health economic parameters (grouped and varied together per category level (e.g. costs of complication hospitalizations, ER discharges, sickness benefit, transportation, absenteeism and premature death, baseline utilities, and disutility). Second-order Monte Carlo probabilistic sensitivity analyses (PSAs) were performed. Beta distributions were used for baseline utilities, lognormal distributions were used for disutilities, and direct medical costs and normal distributions were used for indirect costs. As the CEM used epidemiological outcomes from the 4Flu model, PSAs of the 4Flu and the CEM models needed to be combined. For each iteration of the PSA (1,000 iterations in total), the CEM model used an average of 100 PSA random simulation outcomes of the transmission model as input, which is similar to the approach already deployed by Dolk et al.²⁰.

Results

Base case: Societal perspective

Setting A and B results

If 100% of the years are assumed to be egg-adapted years (Setting A), the use of QIVc instead of QIVe in subjects aged ≥9 years can annually prevent 4,054 symptomatic cases, 1,239 outpatient or GP visits, 419 complications, 51 hospitalizations, and two deaths in a cohort of 100,000 individuals (Table 4). QIVc was found to be a dominant strategy, with an annual gain of 39 QALYs and cost savings of €4,328,713; this resulted in a NMB of €5,512,075 (Table 5). If 55% of the years are assumed egg-adapted years (Setting B), the use of QIVc can annually prevent 2,092 symptomatic cases, 640 outpatient or GP visits, 216 complications, 26 hospitalizations, and one death (Table 4). QIVc was a dominant strategy in Setting B, with an annual gain of 21 QALYs per 100,000 individuals and cost savings of €2,080,721, thereby resulting in a NMB of €2,697,974 (Table 5).

Table 4. The annual clinical outcomes for Settings A and B in the cohort size of 100,000 individuals.

	Setting A (e.g. g-adaptation occurs in 100% of years)				Setting B (e.g. g-adaptation occurs in 55% of years)
	QIVe	QIVe + QIVc	Increment	% Incremental	QIVe	QIVe + QIVc	Increment	% Incremental
Number of symptomatic influenza cases (Total)	151,610	147,556	−4,054	−3	154,369	152,277	−2,092	−1
└normal risk	123,756	120,706	−3,051	−2	125,959	124,392	−1,567	−1
└high risk	27,854	26,851	−1,003	−4	28,410	27,885	−525	−2
Number of GP visits for symptomatic influenza cases (Total)	45,389	44,150	−1,239	−3	46,214	45,574	−640	−1
└normal risk	34,949	34,087	−863	−2	35,566	35,123	−443	−1
└high risk	10,440	10,064	−376	−4	10,648	10,451	−197	−2
Number of complications (Total)	16,111	15,691	−419	−3	16,403	16,186	−216	−1
└normal risk	12,274	11,991	−283	−2	12,490	12,345	−145	−1
└high risk	3,836	3,700	−136	−4	3,913	3,841	−71	−2
Number of hospitalized complications (Total)	1,574	1,523	−51	−3	1,605	1,579	−26	−2
└normal risk	987	959	−28	−3	1,005	991	−15	−1
└high risk	587	565	−23	−4	600	588	−12	−2
Number of deaths (Total)	66	64	−2	−3	68	66	−1	−1
└normal risk	31	30	−1	−3	32	31	−1	−3
└high risk	35	34	−1	−3	36	35	−1	−3

Abbreviations. GP, general practitioner; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine.

Table 5. The annual health economic outcomes for Settings A and B in the cohort size of 100,000 individuals.

	Setting A (e.g. g-adaptation occurs in 100% of years)				Setting B (e.g. g-adaptation occurs in 55% of years)
	Total QALY (total)	Costs (grand total)	ICER (total)	NMB (total)	Total QALY (total)	Costs (grand total)	ICER (total)	NMB (total)
Total population (normal risk + high risk)
QIVe	1,353,698	198,246,715 €	Dominant	5,512,075 €	1,353,674	201,473,808 €	Dominant	2,697,974 €
QIVe + QIVc*	1,353,738	193,918,002 €			1,353,695	199,393,087 €
Incremental	39	−4,328,713 €			21	−2,080,721 €
Normal risk population
QIVe	1,063,509	169,995,098 €	Dominant	4,384,296 €	1,063,492	172,788,358 €	Dominant	2,161,388 €
QIVe + QIVc*	1,063,534	166,369,318 €			1,063,505	171,020,225 €
Incremental	25	−3,625,781 €			13	−1,768,132 €
High risk population
QIVe	290,190	28,251,617 €	Dominant	1,127,778 €	290,182	28,685,450 €	Dominant	536,586 €
QIVe + QIVc*	290,204	27,548,685 €			290,189	28,372,861 €
Incremental	14	−702,932 €			7	−312,588 €

Abbreviations. GP, general practitioner; ICER, incremental cost-effectiveness ratio; NMB, net monetary benefit; €, Euros; QALYs, quality-adjusted life-years; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine.

*QIVc vaccination was provided in population aged 9 years + and QIVe vaccination was provided in population aged < 9 years.

The annual clinical and health economic outcomes for Settings A and B are depicted in Tables 4 and 5, respectively.

Extrapolated results to the entire German population

Results of the base cases (Settings A and B) were extrapolated to annual outcomes for the entire German population (82.5 million). For Setting A, QIVc can annually prevent 167,265 symptomatic cases, 51,114 outpatient or GP visits, 17,299 complications, 2,091 hospitalizations, and 103 deaths (Supplementary Table S9). This would amount to an annual QALY gain of 1,628 years and cost savings of €178,606,268. For Setting B, the annual QALY gain is 849 years and cost savings are €85,852,259. The NMBs were €227,432,752 and €111,320,638 for base case Settings A and B, respectively. QIVc was a dominant strategy with both base case settings.

Extrapolated clinical and health economic results for base case Settings A and B are summarized in Supplementary Tables S9 and S10, respectively.

Scenario and sensitivity analyses

Payer perspective

From a payer perspective with rVE of 36.2%, the ICERs per QALY were €2,285 and €8,984 for Settings A and B, respectively (Table 6). Similar ICER values were reported when Settings A and B results were extrapolated to the entire German population (82.5 million; Table 7).

Table 6. Payer perspective scenario analysis annual health economic outcomes for Settings A (egg-adaptation occurs in 100% of all years) and B (egg-adaptation occurs in 55% of all years) in the cohort size of 100,000 individuals.

	Scenario: Setting A (payer perspective^)				Scenario: Setting B (payer perspective^)
	Total QALY (Total)	Costs (Grand total)	ICER (Total)	NMB (Total)	Total QALY (Total)	Costs (Grand total)	ICER (Total)	NMB (Total)
Total population (normal risk + high risk)
QIVe	1,353,698	15,669,610 €	2,285 €	1,093,209 €	1,353,674	15,801,809 €	8,984 €	432,411 €
QIVe + QIVc*	1,353,738	15,759,763 €			1,353,695	15,986,652 €
Incremental	39	90,152 €			21	184,842 €
Normal risk population
QIVe	1,063,509	10,777,386 €	2,227 €	702,209 €	1,063,492	10,868,217 €	8,823 €	277,597 €
QIVe + QIVc*	1,063,534	10,833,692 €			1,063,505	10,983,876 €
Incremental	25	56,306 €			13	115,659 €
High risk population
QIVe	290,190	4,892,225 €	2,389 €	391,000 €	290,182	4,933,592 €	9,266 €	154,814 €
QIVe + QIVc*	290,204	4,926,071 €			290,189	5,002,775 €
Incremental	14	33,846 €			7	69,183 €

Abbreviations. GP, general practitioner; ICER, incremental cost-effectiveness ratio; NMB, net monetary benefit; QALY, quality-adjusted life-year; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine; rVE, relative vaccine effectiveness; €, Euros.

*QIVc vaccination was provided in population aged 9 years + and QIVe vaccination was provided in population aged < 9 years.

^From a payer perspective, instead of rVE of 36.2%, a scenario with rVE of 19.3% (for both Settings A and B) was considered based on Constantina Boikos et al. publication²⁵.

Table 7. Scenario analysis annual health economic outcomes for Settings A (egg-adaptation occurs in 100% of all years) and B (egg-adaptation occurs in 55% of all years) in the entire German population.

Setting			Setting A (e.g. egg-adaptation occurs in 100% of years)			Setting B (e.g. egg-adaptation occurs in 55% of years)
			Incremental QALYs	Incremental Costs, €	ICER (€/QALY gained)	Incremental QALYs	Incremental Costs, €	ICER (€/QALY gained)
Base case		Societal and 36.2%	1,628	−178,606,268	Dominant	849	−85,852,259	Dominant
Scenario 1	rVE	Societal and 19.3%	895	−93,264,396	Dominant	427	−36,017,877	Dominant
Scenario 2	Perspective	Payer and 36.2%	1,628	3,719,764	2,285	849	7,626,753	8,984
Scenario 3	Perspective and rVE	Payer and 19.3%	895	7,340,493	8,199	427	9,758,197	22,845

Abbreviations. ICER, incremental cost-effectiveness ratio; QALYs, quality-adjusted life-years; rVE, relative vaccine effectiveness.

When a payer perspective was applied with an rVE of 19.3%, the ICERs were €8,199 per QALY and €22,845 per QALY for Settings A and B, respectively (Table 6).

Deterministic sensitivity analyses (DSAs)

Results of the DSAs are presented in Figure 1. This figure highlights the ten parameters with the highest impact on the NMB of QIVc vs. QIVe for Settings A and B. For Setting A, the uncertainties around rVE (both 19.3% [9.5–28.0%] and 36.2% [26.1–44.9%]), aVE (33% [26–39%]) of QIVe, and the probabilities of being symptomatic were the major drivers of the NMB. For Setting B, rVE (19.3% [9.5–28.0%]) was the most important driver of the NMB, followed by the frequency of mismatched years (55% [35.69–73.55%]). Other parameters used in the model have a minor impact on the outcomes.

Figure 1. Deterministic sensitivity analysis: impact of the 10 key drivers on the NMB (Settings A [egg-adaptation occurs in 100% of all years] and B [egg-adaptation occurs in 55% of all years], societal perspective, 100,000 individuals). (a) Impact of the 10 key drivers on the NMB for Setting A. (b) Impact of the 10 key drivers on the NMB for Setting B. Abbreviations. aVE, absolute vaccine effectiveness; NMB, net monetary benefit; OSWA, one-way sensitivity analyses; rVE, relative vaccine effectiveness. Setting A: egg-adaptation occurs in 100% of all years. Setting B: egg-adaptation occurs in 55% of all years. % of death and % of hospitalizations are due to the following complications: bronchitis, pneumonia, unspecified upper tract respiratory infection, myocarditis, myocardial infarction, renal, central nervous system, stroke, otitis media, gastrointestinal bleedings where the base case values were changed with ±10%; disutilities due to influenza symptoms, baseline utilities, and disutilities due to influenza hospitalization changed with ±10%.

Probabilistic sensitivity analyses (PSAs)

From a societal perspective, the cost-effectiveness acceptability curves (CEAC) showed that QIVc has a high likelihood of being a dominant vaccine strategy. PSA outcomes suggested that QIVc was cost-effective against QIVe in 100% and 99% of cases in Settings A and B, respectively (Figures 2 and 3) at a WTP threshold of €30,000 per QALY. From the Payer perspective, the probability of QIVc being cost-effective was almost 100% and 85% of the analyses for Settings A and B, respectively, using the same WTP threshold (Supplementary Figure S3).

Figure 2. Scatter plot and cost-effectiveness acceptability curve for PSA with 1,000 iterations for Setting A (egg-adaptation occurs in 100% of years), societal perspective. (a) Scatter plot of PSA for Setting A. (b) Cost-effectiveness acceptability curve (CEAC) for Setting A. Abbreviations. CEAC, cost-effectiveness acceptability curve; PSA, probabilistic sensitivity analyses; QALYs, quality-adjusted life-years; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine; €, Euros. Setting A: egg-adaptation occurs in 100% of all years.

Figure 3. Scatter plot and cost-effectiveness acceptability curve for PSA with 1,000 iterations for Setting B (egg-adaptation occurs in 55% of years), societal perspective. (a) Scatter plot of PSA for Setting B. (b) Cost-effectiveness acceptability curve (CEAC) for Setting B. Abbreviations. CEAC, cost-effectiveness acceptability curve; PSA, probabilistic sensitivity analyses; QALYs, quality-adjusted life-years; QIVc, cell-based quadrivalent influenza vaccine; QIVe, egg-based quadrivalent influenza vaccine; €, Euros. Setting B: egg-adaptation occurs in 55% of all years.

Discussion

The QIV vaccines are commonly used seasonal influenza vaccines in Germany and most of them are produced using CVVs grown in eggs. However, lower VE has been reported with QIVe due to the adaptation of the vaccine antigen during the manufacturing process, that can create a mismatch with the selected influenza vaccine strain, particularly the A(H3N2) virus^8,10. The production of vaccines using cell-derived influenza viruses circumvents viral mutations occurring in the vaccine antigen and can minimize viral antigenic dissimilarity with the vaccine virus strains. Potential increase in the VE has been reported with the use of cell-derived isolation and manufacturing¹¹. Flucelvax Tetra (QIVc) is currently approved for prophylaxis of influenza in adults and children aged ≥9 years¹².

In Germany, the previously mentioned CEA conducted by Dolk et al.²⁰ compared QIVe with egg-derived trivalent influenza vaccine (TIVe), and showed that QIVe was cost-effective (ICER: €14,461) from the payer perspective and a dominant vaccine strategy from the societal perspective. As QIVe is a common comparator to the analysis of Dolk et al. and our current analysis, we can compare the number of predicted cases and predicted costs (taking into account inflation of costs which is minor) between both analyses. In Dolk et al., annual costs of €1.3 million (€169 per symptomatic influenza case) were calculated from the payer perspective, while annual costs of €6.4 million (€794 per symptomatic influenza) from the societal perspective were accrued for 8,129 symptomatic influenza cases per 100,000 individuals, vaccinated with QIVe vaccine according the German vaccine coverage²⁰.

In the current analyses, we combined the 4Flu transmission model with a similar CEM. Over the time horizon of 20 years, €15.7 million from the payer perspective (Setting A) and €198.2 million from the societal perspective (Setting A) were accrued for 151,610 symptomatic influenza cases for a population of 100,000 individuals, vaccinated with QIVe vaccine according to the German vaccine coverage. This is equivalent to €103 and €1,307 per symptomatic influenza individual from the payer and societal perspective, respectively. These costs are in line with costs that were reported by Dolk et al.²⁰. However, there are certain differences in costs between both analyses due to differing assumptions on hospitalizations and deaths in the model. In our analysis, the probability to be hospitalized or to die is complication-specific and, where possible, age-dependent. In comparison, Dolk et al. assumed that hospitalization and mortality solely depend on age and not on the complication type.

Results from this study are in line with similar analyses conducted on the comparison of QIVc versus QIVe in the UK, Italy, Spain, and in the US^14–17. In all countries, the obtained ICERs for QIVc versus QIVe are far below the WTP threshold from the payer perspective, and QIVc is a dominant vaccination strategy from the societal perspective. Compared to our study, a different rVE was used in studies in the UK and Spain due to age restriction of QIVc in these two countries. In the analysis conducted in the UK population, a rVE (26.8%) was used since the use of QIVc is recommended only in those under 65 years in the UK^15,45. Without these age restrictions, rVE in our German CEM was assumed to be 36.2% in the base case and 19.3% in the scenario analysis. In the UK-based study, it was assumed that 100% of the years are egg-adapted, and the resulting ICER was £598 from the payer perspective¹⁵. In the analysis conducted in Spain, rVE was also assumed to be 26.8% due to a similar age restriction as that recommended in the UK. The reported ICER (QIVc vs. QIVe) was €12,852 from the payer perspective and QIVc was dominant from the societal perspective¹⁷. Similar to our study, rVE was assumed to be 36.2% in the base case and 19.3% in scenario analyses for Italy¹⁴. The QIVc was cost-effective compared to QIVe from the payer perspective with ICER of €2,714 for the base case and €8,021 for a scenario analysis, and was reported as a dominant vaccine strategy in both the base case and scenario analyses¹⁴. The study conducted in the US shows that using QIVc instead of QIVe can reduce annual outpatient visits and hospitalizations by three million and 84,185, respectively. Using QIVc can be a cost saving alternative vaccination strategy for the US Health Care System¹⁶. Pro-rated to the German population size, using QIVc instead of QIVe results in a more modest reduction of outpatient visits and hospitalizations by 51,114 and 2,091 (for Setting A) and by 26,410 and 1,091 for Setting B, and in substantial (€170.8 million for Setting A and for €78.1 million for Setting B) savings from a societal perspective.

Note that there is also a study that does not show a significant better effectiveness of QIVc⁴⁶. In a case-control study, in patients aged < 65 years, statistically significant protection against influenza hospitalization of QIVc compared to QIVe was not observed⁴⁶. However, it should be noted that comorbidities were typically more present among recipients of QIVc than recipients of QIVe, especially among patients ages < 65 years in the study. Selection bias could occur if QIVc vaccinated patients, who had more comorbidities, were more likely to be hospitalized and tested for influenza than QIVe vaccinated individuals. Moreover, the population size of the QIVc group (75 patients < 65 years and 157 patients ≥ 65 years) is very small compared to the QIVe group (1,741 patients < 65 years and 3,498 patients ≥ 65 years). Furthermore, there may be other factors that affect VE, such as the immune response to prior influenza infection/vaccination and host characteristics, but that are not considered in the study.

This CEA has several strengths such as the flexibility to allow input per age. The model and analyses are based on a simple, transparent, and published model structure and the main source of inputs in our model came from one published model. The model also included stroke and MI as possible outcomes which were not included in the analyses of Dolk et al.²⁰ performed in a German population earlier. Our study was conducted from societal as well as payer perspectives. Therefore, study findings may provide support for reimbursement of QIVc in Germany. Uncertainty in model parameters was tested by performing a number of sensitivity analyses. The rVE and aVE of QIVe, probabilities of being symptomatic and the number of mismatched years were the main drivers of the NMB. All uncertainty testing demonstrated robustness of the model.

However, results from this study must be interpreted in light of a few limitations. The main limitation for this CEM was related to costs. Firstly, costs data used within CEM were mainly based on Dolk et al.²⁰ and were inflated to the current year. Hence, we assume that those cost sources are still valid in the current scenario. Secondly, absenteeism cost, which is a major cost item in this analysis, was estimated based on the assumption that the number of working days lost is the same as the duration of the disease. However, people may take fewer (or more) days off in reality and parents might not take time off to be with their children during sickness.

Furthermore, changes in medical management can impact medical resource use over time, which are not reflected in this CEM. Non-availability of absolute effectiveness levels for the different vaccines in matched and mismatched years can also be considered as a limitation for this study because the same relative effectiveness leads to different absolute differences depending on the baseline effectiveness value.

Conclusion

This study suggests that influenza vaccination with QIVc is cost-effective compared with QIVe in Germany. From a societal perspective, the introduction of QIVc can help in avoiding a significant number of influenza cases and complications while generating cost savings (dominant strategy).

Transparency

Declarations of funding

This work was funded by Seqirus Inc.

Declaration of financial/other interests

RC, LG, and ML are full-time employees of IQVIA. IQVIA received consulting fees from Seqirus for the development of the cost-effectiveness model. ME is a shareholder of Epimos GmbH, which has received funding from Seqirus to complete the work disclosed here as well as consultancy from other pharmaceutical companies. MS is a shareholder of ExploSYS GmbH, which has received funding from Epimos GmbH to complete the work disclosed here. SR and JFM-Q are employees of Seqirus and shareholders of CSL Limited.

JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

Results were presented as a poster at the Virtual European Ispor Conference 2020. Cai R, Eichner M, Schwehm M, Rajaram S, Mould-Quevedo J, Lamotte M. Cost-effectiveness of the CELL-based quadrivalent versus the standard egg-based quadrivalent influenza vaccine in Germany. Results of a dynamic transmission modeling approach. ISPOR Europe 2020 Virtual conference, Milano, November 16–19, Abstract ID# 105463.

Acknowledgements

Chandresh Kumar and Paranjoy Saharia are employees of IQVIA India, who have provided writing support for this manuscript.