A 21-winter seasons retrospective study of antibody response after influenza vaccination in elderly (60–85 years old) and very elderly (>85 years old) institutionalized subjects

ABSTRACT

Influenza vaccination is considered the best mean for preventing the higher rates of mortality associated with influenza virus infection in the elderly as compared with younger people. Since the number of very elderly subjects, aged >85 years, is rapidly increasing, and some authors reported increments in influenza-associated mortality with age, the aim of this study was to increase the limited information available on the immunogenicity of the influenza vaccines in this age group. This was a retrospective study which analyzed the antibody response induced by commercially available trivalent inactivated influenza vaccines in 1491 elderly subjects (60–85 years old) and 1139 very elderly subjects (>85 years old) during 21 winter seasons included between 1993–1994 and 2014–2015. The antibody response of the two age groups was, in most instances, acceptable according to the Committee for Medical Products for Human Use and comparable. In accordance with previous data obtained in the elderly, the use of MF59-adjuvanted or intradermal administered vaccines (enhanced vaccines) was found to be preferable as compared with conventional formulations (split or subunit vaccines). Vaccines containing new strains induced higher antibody response as compared with vaccines with the same antigenic composition of the previous years. These results suggest that the current recommendation for use of enhanced influenza vaccines for the elderly is appropriate, but that efforts to improve the effectiveness of the present prophylactic measures against influenza are needed, especially in the years with vaccines with the same antigenic composition of the previous winter season.

KEYWORDS:

• aging
• influenza vaccine
• vaccine immunogenicity
• vaccine strain changes
• vaccine formulation

Introduction

Respiratory infections, and in particular influenza virus infections, are considered to be one of the major cause of the severe morbidity and mortality which disproportionally affects elderly although attack rates for influenza are higher in children. Each year in the United States, there are 300,000 hospitalizations and 23,000 deaths associated with influenza among those aged ≥65 years¹ and for this reason influenza control in the elderly through the use of vaccines is considered to be a public health priority in all the world.

However, currently available influenza vaccines are known to induce reduced responses in elderly as compared with younger adults and age-specific influenza vaccine effects were previously found in older individuals.² Govaert et al.³ examining results obtained in the only prospective, randomized, double-blind placebo-controlled trial of influenza vaccination conducted in people aged >60 years, demonstrated an efficacy of 58% against serologically confirmed clinical influenza, but persons aged >70 years appeared to be less protected. The antibody responses to influenza vaccination in older adults, examined in a quantitative review by Goodwin et al.,⁴ evidenced that the responses were especially impaired in subjects >75 years of age relative to those observed in younger persons.

Results obtained in our laboratory examining influenza vaccine immunogenicity in 27 consecutive seasons in elderly institutionalized people evidenced a different tendency. In fact, comparing subjects above and below 75 years of age, in many instances the values observed in people aged >75 years were slightly higher as compared with those found in people aged ≤75 years. The differences were particularly evident against the A/H1N1 strain.⁵

Since the number of elderly and very elderly (>75 or >85 years) is rapidly increasing⁶ and since Thompson et al.¹ demonstrated increases with age in influenza-associated mortality among elderly people, those older than 85 years of age were found to be 16-times more likely to die of influenza-related illness and 32-times more likely to die of influenza-associated pneumonia than those between 65 and 69 years of age, it is important to have additional information about the effect of age on influenza vaccine immunogenicity in elderly people.

The aim of the present work was to examine in a more restricted way data previously published obtained in our laboratory.⁵ The present results are reported as crude mean antibody responses induced after influenza vaccination and were obtained studying 2630 elderly volunteers in 21 different winter seasons.

The 2630 subjects were elderly frail institutionalized volunteers vaccinated with trivalent inactivated vaccines (with 15 µg of haemagglutinin (HA)/strain) for which we had available demographic and health status information. We performed a subgroup analysis of elderly subjects considering 1491 younger (60–85 years) and 1139 older (>85 years) persons. Moreover we investigated on the effects of the introduction of a new antigen in the vaccine composition and on the use of different influenza vaccine formulations.

Results

Study population and influenza vaccine composition

Table 1 shows the number and the mean age of volunteers, vaccinated each studied influenza season, the type of vaccine and the influenza strains included in the seasonal vaccine used for the winter examined.

Table 1. Characteristics of studied population, type and composition of influenza vaccines used in the 21 winter seasons covered by the study.

Season (N. of subjects)	Mean age (range)	Conventional vaccines^a	Enhanced vaccines^b	Vaccine composition
				A/H1N1	A/H3N2	B
1993–94 (298)	76 (60–99)	298	–	A/Singapore/6/86	A/Beijing/32/92	B/Panama/45/90
1994–95 (131)	78 (60–100)	131	—	A/Singapore/6/86	A/Shandong/9/93	B/Panama/45/90
1995–96 (95)	77 (60–100)	95	—	A/Singapore/6/86	A/Johannesburg/33/94	B/Beijing/184/93
1996–97 (66)	80 (60–99)	66	—	A/Singapore/6/86	A/Wuhan/359/95	B/Beijing/184/93
1998–99 (96)	74 (60–102)	96	—	A/Beijing/262/95	A/Sydney/5/97	B/Beijing/184/93
1999–00 (138)	83 (60–103)	123	15	A/Beijing/262/95	A/Sydney/5/97	B/Beijing/184/93
2000–01 (112)	83 (60–103)	112	—	A/New Caledonia/20/99	A/Moscow/10/99	B/Beijing/184/93
2001–02 (91)	82 (60–104)	91	—	A/New Caledonia/20/99	A/Moscow/10/99	B/Sichuan/379/99
2002–03 (105)	82 (60–105)	105	—	A/New Caledonia/20/99	A/Moscow/10/99	B/Hong Kong/330/01
2003–04 (124)	83 (60–101)	33	91	A/New Caledonia/20/99	A/Moscow/10/99	B/Hong Kong/330/01
2004–05 (147)	82 (60–99)	26	121	A/New Caledonia/20/99	A/Fujian/411/02	B/Shanghai/361/02
2005–06 (71)	83 (60–99)	19	52	A/New Caledonia/20/99	A/California/7/04	B/Shanghai/361/02
2006–07 (74)	83 (60–98)	—	74	A/New Caledonia/20/99	A/Wisconsin/67/05	B/Malaysia/2506/04
2007–08 (65)	84 (61–102)	—	65	A/Solomon Islands/3/06	A/Wisconsin/67/05	B/Malaysia/2506/04
2008–09 (112)	83 (60–103)	—	112	A/Brisbane/59/07	A/Brisbane/10/07	B/Florida/4/06
2009–10 (64)	83 (65–98)	—	64	A/California/7/09	A/Brisbane/10/07	B/Brisbane/60/08
2010–11 (112)	85 (64–101)	—	112	A/California/7/09	A/Perth/16/09	B/Brisbane/60/08
2011–12 (139)	84 (64–102)	—	139	A/California/7/09	A/Perth/16/09	B/Brisbane/60/08
2012–13 (252)	85 (60–103)	26	226	A/California/7/09	A/Victoria/361/11	B/Wisconsin/1/10
2013–14 (159)	86 (60–105)	—	159	A/California/7/09	A/Texas/50/12	B/Massachusetts/2/12
2014–15 (179)	84 (60–104)	—	179	A/California/7/09	A/Texas/50/12	B/Massachusetts/2/12

^a conventional influenza vaccines = whole, split and subunit

^b enhanced influenza vaccines = adjuvanted with MF59, intradermal

The study population consisted of 2630 elderly volunteers aged ≥60 years (mean age 82.2 years, median 83, range 60–104), 1491 of them (57%), classified as younger elderly, were 60–85 years old (mean age 76.1, median 77, range 60–85) and 1139 (43%), classified as very elderly, were aged >85 years (mean age 90.1, median 89, range 86–104). The majority of volunteers were females (80% of the total population, 72% of the young elderly and 89% of the very elderly subjects). Each year different cohorts of volunteers were examined except for a low percentage (20%) found to be present for one or more consecutive seasons. Almost all of the 2630 volunteers had been previously vaccinated (90–100%) before the enrollment in the study, although the number and the year of previous vaccinations was unavailable. The majority of the subjects (87%) was living in nursing homes located in central Italy, more than 80% presented underlying diseases or risk factors for influenza and, as a consequence, used chronically drugs. The most frequent chronic diseases were cardiovascular, respiratory diseases and diabetes. The most frequent drugs used were antihypertensive/inotropic drugs and benzodiazepines.

During the 21 winter seasons covered by the study, different formulations of influenza vaccines were used. We indicated as conventional vaccines (whole, split, i.e. viruses disrupted by a detergent, and subunit, i.e. purified surface viral glycoprotein) vaccines recommended for all ages and as enhanced vaccines (MF59 adjuvanted vaccines and vaccines administered intradermally) vaccines recommended for elderly people, since previously found to induce in this age group higher responses as compared with conventional formulations.⁷

Table 1 reports the number of subjects receiving a conventional vaccine, i.e. conventional (N = 1221), and those who received a potentiated vaccine (N = 1409). In some seasons all the people studied received a conventional (8 seasons) or an enhanced vaccine (8 seasons), in the others both types of vaccines were used.

Finally since the antigenic composition of influenza vaccines was updated each year to match the strains with the highest probability of circulation,⁸ Table 1 reports the antigenic composition of the trivalent vaccines used each year formulated according to the recommendation of “Ministero della Salute (Italy)” and WHO (Northern Hemisphere).

Vaccine immunogenicity

We examined in a more restricted way data previously published⁵ obtained in our laboratory. The data were obtained studying 2630 volunteers in 21 different winter seasons. We evaluated the effect on the vaccine immunogenicity of age (people above or below 85 years), changes in vaccine components, vaccine formulation (conventional and enhanced) and sex.

Since ANCOVA test executed for evaluating differences on the vaccine response of groups reveals that the slopes of the regression lines were not significantly different, all post-vaccination GMTs were corrected for the average pre-vaccination imbalance estimated on the whole population. Thus, the results were reported as crude mean antibody responses induced after influenza vaccination.

The results of the regression model, calculated as described in materials and methods, are reported in Table 2.

Table 2. Linear model of the regression analysis (OLS method) for the three antigens predicting T_post by T_pre (up), and including sex, age, vaccine type and change (down).

	H3N2		H1N1		B
Independent variables	Estimates of regression models
	A (y-intercept)	1.98^**	A (y-intercept)	1.77^**	A (y-intercept)	1.76^**
Tpre	Bpre	0.79^**	Bpre	0.78^**	Bpre	0.74^**
	R^2	0.47	R^2	0.42	R^2	0.43
	A (y-intercept)	2.14^**	A (y-intercept)	1.80^**	A (y-intercept)	1.66^**
Tpre	Bpre	0.76^**	Bpre	0.78^**	Bpre	0.76^**
Gsex	Bsex_M	−0.25^**	Bsex_M	−0.37^**	Bsex_M	−0.23^**
Gage	Bage_> = 85	−0.12	Bage_> = 85	−0.01	Bage_> = 85	−0.17^**
GvaccType	BvacType_Conv	−0.51^**	BvacType_Conv	0.07	BvacType_Conv	0.23
Gch	Bch_NonCh	0.48^**	Bch_NonCh	0.38^**	Bch_NonCh	0.08^**
	R^2	0.50	R^2	0.43	R^2	0.44

^** : p < 0.01 comparing the estimate with respect to 0

Vaccine immunogenicity in the overall population (≥60 years), in younger elderly (60–85 years) and very elderly (>85 years) volunteers

Table 3 reports the results obtained examining haemagglutination inhibiting (HI) crude mean titres found in sera collected before and one month after vaccination against the 3 strain antigens (A/H1N1, A/H3N2 and B) present in the vaccines used.

Table 3. HI antibody response to vaccine antigens of overall population and of the two age groups (younger elderly, 65–85 years, and very elderly, >85 years).

		Seroprotection rate [CI 95%]		GMT (GMTR) [CI 95%]
	Groups	Pre-vacc.	Post-vacc.	Pre-vacc.	Post-vacc.	Seroconversion rate [CI 95%]	CHMP criteria	Post-GMT corrected [CI 95%]
A/H3N2	Overall	38 [36–40]	70^** [68–72]	21.7 [20.7–22.7]	62.6^** (2.9) [59.3–66.0]	34 [33–36]	3/3	19.8 [19.0–20.6]
	65–85	33 [31–36]	67^** [64–69]	19.2 [18.1–20.3]	56.8^** (3.0) [52.9–61.0]	35 [33–38]	3/3	19.9 [18.8–20.8]
	>85	43^A [40–46]	74^**^A [71–76]	25.5^A [23.6–27.4]	71.1^**^A (2.8) [65.5–77.1]	33 [31–36]	3/3	19.7 [18.6–21.0]
A/H1N1	Overall	27 [25–28]	59^** [57–61]	15.5 [14.6–16.3]	41.2^** (2.7) [38.8–45.3]	29 [27–30]	1/3	17.1 [16.9–17.2]
	65–85	26 [24–28]	58^** [56–61]	15.4 [14.6–16.3]	40.6^** (2.6) [38.0–43.4]	28 [26–30]	1/3	16.8 [16.0–17.7]
	>85	28 [25–30]	60^** [57–63]	15.5 [14.6–16.5]	41.9^** (2.7) [38.8–45.3]	30 [27–32]	3/3	17.3 [16.3–18.4]
B	Overall	30 [28–32]	62^** [60–64]	17.5 [16.9–18.2]	43.1^** (2.5) [41.2–45.2]	27 [25–29]	2/3	16.9 [16.3–17.5]
	65–85	28 [26–31]	62^** [60–65]	16.8 [16.0–17.7]	43.8^** (2.6) [41.3–46.5]	29 [26–31]	2/3	17.8^A [17.0–18.6]
	>85	32 [29–35]	63^** [60–65]	18.5 [17.4–19.7]	42.1^** (2.3) [39.3–45.1]	25 [23–38]	2/3	15.9 [15.1–16.7]

^** : p < 0.01 comparing pre and post-vaccination data

A: p < 0.01 comparing the two age groups

Because previous observations suggested the possibility of lower responses after influenza vaccination in the very elderly people,⁴ the results obtained were examined both in the overall population (≥60 years, N = 2630) and in volunteers divided in younger elderly (60–85 years, N = 1491) and very elderly (>85 years, N = 1139).

Significant increases were observed against all the three vaccine antigens comparing pre- and post-vaccination percentages of seroprotected people (HI titres ≥40, considered to be associated with protection from influenza infection),⁹ and geometric mean titre (GMT) values in the overall population and in people aged 60–85 or >85 years.

Moreover, results were evaluated according to the criteria of the Committee for Medicinal Products for Human Use (CHMP) for approval of influenza vaccines, which require that for individuals aged ≥60 years at least one of the following values must be met: seroprotection rate ≥60%, ratio of post-immunization to pre-immunization GMT (GMTR) ≥2, and seroconversion rate (subjects with a fourfold or greater increase in titre in pre-vaccination seropositive subjects or from <10 to ≥40 in seronegative volunteers) ≥30%.¹⁰

The 3 CHMP requirements were always reached against A/H3N2 antigen, whereas only 2 of the 3 were satisfied against the B antigen in the 3 groups. Considering the A/H1N1 strain, the 3 CHMP criteria were met only in the very elderly whereas only 1of the 3 was reached in the overall population and in people aged 60–85 years.

Comparing the pre- and post-vaccination HI antibody titres observed in the younger elderly and very elderly groups, we found similar values against the A/H1N1 and the B antigens. Instead, against the A/H3N2 antigen the pre- and post-vaccination seroprotection rates and the values of GMT of the very elderly were significantly higher than those of the younger elderly. However, examining post-vaccination GMT values corrected for the pre-vaccination status according to Beyer et al.,¹¹ in order to have more comparable data, the results showed that the antibody response of the two age groups was statistically similar both against the A/H1N1 and the A/H3N2 vaccine components, as confirmed also by the regression model shown in Table 2 and thus age cannot explain the variance in post-vaccination titres for these two vaccine components. The correction for the pre-vaccination status showed instead a statistical difference for the B vaccine component against which a better response was developed by the younger group (65–85 years).

HI antibody response in presence of novel antigen in the vaccine composition

Previous published data evidenced differences in the HI responses of subjects using influenza vaccines with unaltered or changed antigen strains composition.^12,13

As reported in Table 1, during the 21 winter seasons considered in our study the WHO recommended 12 A/H3N2, 5A/H1N1, and 9 B new influenza strains for inclusion in seasonal vaccines. To add information about the possibility that the presence of a novel antigen in the vaccine composition might influence the antibody response, we examined the responses observed in the volunteers according to the presence (changed group) or the absence (not-changed group) of a new vaccine antigen. The volunteers of the changed and not-changed groups were 1255 and 1375 for A/H3N2, 177 and 2453 for A/H1N1, and 1846 and 784 for B antigen, respectively.

Table 4 shows the HI antibody responses induced following influenza vaccination against the 3 different vaccine antigens in changed and not-changed groups of the overall population and of the 2 age groups (60–85 and >85 years). One month after vaccination, significant increases in the values of seroprotection and GMT values were always found against all the 3 different antigens. During the seasons characterized by a change of one or more antigens, the 3 CHMP criteria were always satisfied, on the contrary when the vaccine composition didn't change, only 2 or 1 of the 3 criteria were reached.

Table 4. HI antibody response to vaccine antigens of overall population and of the two age groups (younger elderly, 65–85 years, and very elderly, >85 years old) in presence (changed group) or not (no-changed group) of novel antigen in the vaccines composition.

			Seroprotection rate [CI 95%]		GMT (GMTR) [CI 95%]
	Groups		Pre-vacc.	Post-vacc.	Pre-vacc.	Post-vacc.	Seroconversion rate [CI 95%]	CHMP criteria	Post-GMT corrected [CI 95%]
A/H3N2	Overall	Changed group	38 [36–41]	74^**^A [72–77]	23.2^a [21.7–24.9]	79.2^**^A (3.4) [73.0–85.8]	42^A [40–45]	3/3	23.6^A [22.3–25.1]
		No-changed group	37 [34–40]	66^** [63–68]	20.3 [19.1–21.7]	50.5^** (2.5) [47.1–54.2]	27 [25–29]	2/3	16.8 [16.0–17.6]
	65–85	Changed group	34^B [31–38]	72^**^A [69–75]	20.7^B [19.0–22.6]	71.2^**^AB(3.4) [63.9–79.2]	43^A [39–47]	3/3	23.2^A [21.5–25.2]
		No-changed group	33^B [29–36]	62^**^B [58–65]	17.8 [16.4–19.3]	46.2^**(2.6) [42.2–50.6]	28^** [25–31]	2/3	17.0 [15.9–18.1]
	>85	Changed group	44 [40–48]	77^**^A [74–81]	27.1 [24.2–30.2]	91.1^**^A(3.4) [80.8–102.8]	42^A [38–46]	3/3	24.2^A [22.1–26.5]
		No-changed group	43 [39–47]	71^** [67–74]	24.1 [21.8–26.7]	56.8^**(2.3) [50.9–63.3]	26 [22–29]	2/3	16.5 [15.3–17.8]
A/H1N1	Overall	Changed group	31 [24–38]	75^**^A [67–85]	14.9 [12.7–17.3]	51.4^**^A (3.5) [43.2–61.2]	43^A [36–50]	3/3	22.0^A [18.8–25.7]
		No-changed group	26 [25–28]	58^** [56–60]	15.5 [14.9–16.2]	40.5^** (2.6) [38.4–42.7]	28 [26–29]	1/3	16.7 [16.1–17.4]
	65–85	Changed group	33 [23–42]	76^**^A [67–85]	15.8 [12.7–19.6]	53.9^**(3.4) [42.2–68.9]	41 [31–51]	3/3	22.0^A [17.5–27.7]
		No-changed group	25 [23–38]	57^** [54–59]	15.4 [14.6–16.3]	39.8^**(2.6) [37.2–42.6]	27 [35–29]	1/3	16.6 [15.7–17.4]
	>85	Changed group	29 [19–39]	74^** [65–84]	13.9 [11.1–17.3]	48.6^**(3.5) [37.7–62.7]	45 [34–56]	3/3	21.9 [17.7–27.2]
		No-changed group	28 [25–31]	59^** [56–62]	15.6 [14.6–16.7]	41.4^**(2.6) [38.2–44.9]	28 [26–31]	1/3	17.0 [16.0–18.1]
B	Overall	Changed group	27^A [25–29]	61^** [59–64]	16.0 [15.3–16.8]	42.3^** (2.6) [40.1–44.7]	31^A [29–33]	3/3	17.8^A [17.1–18.6]
		No-changed group	38 [35–42]	65^** [61–68]	21.7 [20.1–23.5]	44.9^** (2.1) [41.3–48.8]	18 [16–21]	2/3	15.0 [14.2–15.9]
	65–85	Changed group	25^A [22–27]	60^**^A [57–63]	15.3^A [14.5–16.3]	41.7^**^A(2.7) [38.9–41.7]	31^A [28–34]	3/3	18.1 [17.1–19.1]
		No-changed group	38 [34–43]	68^** [63–72]	21.7 [19.4–24.3]	50.4^**(2.3) [44.7–56.7]	22^b [18–26]	3/3	16.9^B [15.5–18.5]
	>85	Changed group	29^a [26–33]	63^** [60–67]	17.1^A [15.9–18.3]	43.3^**(2.5) [39.7–47.2]	31 [27–34]	3/3	17.4^A [16.3–18.6]
		No-changed group	38 [33–43]	62^** [57–66]	21.8 [19.5–24.3]	40.0^**(1.8) [35.6–44.9]	15^A [11–18]	1/3	13.4 [12.5–14.3]

^** : p < 0.01 comparing pre and post-vaccination data

A: p < 0.01; a: p < 0.05 comparing the changed and no-changed groups

B: p < 0.01; b: p < 0.05 comparing the two age groups for the changed or the no-changed groups

In particular, comparing the results found in the changed and not-changed groups, the values observed in the volunteers in the seasons with a new vaccine antigen were very frequently higher than values found in the years with no antigen vaccine change. In many instances the differences were found to be statistically significant. The comparison of GMT corrected for the pre-vaccination status¹¹ confirmed the higher average immunogenicity in the groups of volunteers vaccinated with new antigens for all the 3 vaccine components, and the regression model confirmed the result (Table 2) since the change predictor was significant for all antigens.

Comparing the age groups in case of change or not change of vaccine antigen (Table 4), the corrected post-GMT values found against A/H3N2 and A/H1N1 vaccine components were not significantly different and thus the change predictor was not necessary for explaining the regression model, as also confirmed by results in Table 2. On the other hand, for the B vaccine component the differences between age groups against B component were significant only when the vaccine component was not changed. This effect was peculiar for the B component and probably was due to the interaction between age and presence of a new B vaccine component.

HI antibody response after immunization with different formulations of influenza vaccines

Immunization with different formulations of influenza vaccine might influence the antibody response.⁴ Since conventional and enhanced inactivated trivalent vaccines were used in the 21 winter seasons (see Table 1), the antibody response induced by the two vaccine formulations was compared, in the overall population, in younger (60–85 years) and very elderly (>85 years) groups.

The results obtained are reported in Table 5. Significant increases in the values of seroprotection and of GMT were always found in the overall population and in volunteers divided in the 2 age groups against all the 3 different antigens one month after vaccination. Following vaccination with enhanced vaccines, the 3 CHMP criteria were in most instances reached against A/H3N2 and A/H1N1 antigens, whereas only 2 of the 3 criteria were met against the B antigen and in people aged >85 years against A/H1N1. Following vaccination with conventional vaccines the number of reached CHMP criteria was in most instances lower since only 1 or 2 criteria were satisfied against all the three vaccine antigens with the only exception of the enrichment of all the 3 criteria against B antigen in people aged >85 years.

Table 5. HI antibody response to vaccine antigens of overall population and of the two age groups (younger elderly, 65–85 years, and very elderly, >85 years) after vaccination with conventional (whole, split and subunit) or enhanced (adjuvanted with MF59 or intradermal) influenza vaccines.

			Seroprotection rate [CI 95%]		GMT (GMTR) [CI 95%]
	Groups		Pre-vacc.	Post-vacc.	Pre-vacc.	Post-vacc.	Seroconversion rate [CI 95%]	CHMP criteria	Post-GMT corrected [CI 95%]
A/H3N2	Overall	Convent.	31^A [28–33]	61^**^A [58–63]	18.5^A [17.3–19.8]	45.8^**^A (2.5) [42.6–49.3]	26^A [24–29]	2/3	16.3^A [15.5–17.2]
		Enhanced	44 [41–46]	78^** [75–80]	24.8 [23.3–26.5]	82.1^** (3.3) [76.2–88.5]	41 [39–44]	3/3	23.3 [22.0–24.6]
	65–85	Convent.	25^AB [22–28]	56^**^AB [53–60]	15.9^AB [14.7–17.1]	39.9^**^AB (2.5) [36.5–43.6]	26^A [24–29]	1/3	16.1^A [15.1–17.1]
		Enhanced	43 [39–47]	79^** [76–82]	23.9 [21.9–26.1]	86.0^**(3.4) [77.5–95.4]	45^**^B [42–49]	3/3	25.2 [23.2–27.3]
	>85	Convent.	41 [37–46]	69^** [65–74]	24.8 [22.1–27.9]	59.7^**^A(2.4) [52.7–67.7]	26^A [22–30]	2/3	16.9^A [15.5–18.4]
		Enhanced	45 [41–48]	77^** [73–80]	25.8 [23.4–28.4]	78.6^**(3.0) [70.7–87.5]	38 [34–41]	3/3	21.6 [20.0–23.4]
A/H1N1	Overall	Convent.	26 [24–29]	55^**^A [53–58]	15.8 [14.9–16.8]	41.7^** (2.6) [38.6–45.0]	27 [24–29]	1/3	17.0 [16.0–18.0]
		Enhanced	27 [25–30]	62^** [59–64]	15.2 [14.3–16.0]	40.7^** (2.7) [38.1–43.5]	30 [28–33]	3/3	17.3 [16.3–18.0]
	65–85	Convent.	26 [23–29]	53^**^A [50–57]	16.0 [14.8–17.3]	39.4^**(2.5) [35.8–43.3]	24^AB [21–27]	1/3	15.9^B [14.9–17.0]
		Enhanced	25 [22–28]	63^** [60–67]	14.8 [13.7–16.0]	42.0^**(2.8) [38.5–46.1]	33 [29–36]	3/3	18.0 [16.7–19.5]
	>85	Convent.	26 [22–30]	59^** [55–64]	15.5 [14.0–17.2]	46.5^**(3.0) [40.8–53.1]	32 [28–37]	2/3	19.2 [17.3–21.3]
		Enhanced	29 [26–32]	61^** [57–64]	15.5 [14.3–16.8]	39.5^**(2.5) [35.9–43.4]	28 [32–35]	2/3	16.3 [15.2–17.5]
B	Overall	Convent.	26 [24–29]	59^**^A [56–62]	16.5 [15.5–17.4]	43.0^** (2.6) [40.2–46.1]	29 [26–31]	1/3	17.7 [16.8–18.7]
		Enhanced	33 [31–36]	65^** [63–68]	18.5 [17.6–19.6]	43.1^** (2.3) [40.6–45.8]	26 [23–28]	2/3	16.2 [15.5–17.0]
	65–85	Convent.	24^A [21–27]	57^**^A [53–60]	15.4^AB [14.3–16.5]	40.9^**(2.6) [37.6–44.5]	28 [25–31]	1/3	17.7 [16.6–18.9]
		Enhanced	33 [30–37]	69^** [65–72]	18.7 [17.3–20.2]	47.6^**(2.5) [43.7–51.8]	29^B [26–33]	2/3	17.8 [16.6–19.1]
	>85	Convent.	31 [27–36]	64^** [59–69]	18.8 [17.0–20.8]	47.5^**(2.5) [42.1–53.7]	30 [25–35]	3/3	17.7 [16.2–19.4]
		Enhanced	33 [29–36]	62^** [58–66]	18.4 [17.0–19.9]	39.3^**(2.1) [36.0–42.7]	22 [19–25]	2/3	16.3 [15.2–17.5]

^** : p < 0.01 comparing pre and post-vaccination data

A: p < 0.01; a: p < 0.05 comparing the two vaccine types

B: p < 0.01; b: p < 0.05 comparing the two age groups for the same vaccine type

Comparing results obtained after immunization with conventional and enhanced vaccines, the A/H3N2 component showed that both post-GMT and post-GMT corrected values were always higher after enhanced vaccines independently from age, as also confirmed by the regression model in Table 2. The A/H1N1 and B strains responses showed that the vaccine type was not affecting the response since the corrected post-GMT were not significantly different (see also Table 2). The seroconversion and seroprotection rates of the A/H1N1 component were both significantly higher in the younger elderly when they were immunized with the enhanced vaccines.

Comparing the two age groups vaccinated with conventional or enhanced vaccine types, the results showed higher serconversion rates against the A/H3N2 and B vaccine components in the younger as compared with the very elderly after administration of enhanced vaccines. The corrected post-GMT and the seroconversion rate against A/H1N1 vaccine component were significantly higher in the younger elderly as compared the very elderly after conventional vaccines administration. This result was not explained by the regression model (Table 2) and might be due to cross-reactivity generated from a different exposure of the two age groups to A/H1N1 strains.

Discussion

This retrospective study describes the HI antibody response of 2630 elderly institutionalized frail subjects induced after immunization with influenza trivalent inactivated vaccines commercially available during 21 different years included between 1993–1994 and 2014–2015 winter seasons (Table 1). Previous observations evidenced rates of serious illness and death particularly high in elderly people probably do to immunosenescence, age-related dysregulation at multiple levels of the immune system, and presence of comorbidities as chronic medical conditions.^14,15 Moreover since age-specific effects on influenza vaccine immunogenicity were reported among persons aged ≥65 years,^3,4 we evaluated the response of all the volunteers (≥60 years) and of volunteers divided in younger elderly (60–85 years) and very elderly (>85 years). The present study increases the state of knowledge with respect to our previous published data.⁷ In fact, new information were added relative to influenza vaccine immunogenicity in people aged >85 years, a population not specifically analyzed in the previous work⁷ and found to have an higher morbidity and mortality for influenza as compared with younger elderly people (65–69 years of age).¹

In the present study a regression model was applied to the log₂-transformed pre- and post-vaccination vaccination titres to analyze the dependence of the vaccine immunogenicity by some factors (age, and vaccine components change, vaccine type and sex). Using a regression model based only on pre-vaccination titres as independent variable to predict the average post-vaccination titre, we found, for each antigen, a different predictive power of the model (R²) equal to 46.6% (A/H3N2 antigen), 41.6% (A/H1N1) and 42.7% (B) (upper section of Table 2). The adopted model (lower section of Table 2) added to R² 3% for the A/H3N2 component and 1% for both the H1N1 and B strains.

First results, in accordance with previous data,⁵ confirmed that influenza vaccine administration was able to induce statistically significant increases in HI titres evaluated as seroprotection rate and GMT values against all the 3 vaccine antigens both considering the overall old people or the two age groups (Table 2).

The antibody response of the very elderly people (>85 years) was not impaired as compared with younger elderly volunteers (60–85 years). No significant difference was found against A/H1N1 antigen comparing pre- and post-vaccination values, the HI antibody response of the two age groups were similar.

Considering the A/H3N2 antigen, both pre- and post-vaccination values of seroprotection and GMT in people aged >85 years were statistically higher as compared with values found in the 65–85 age group. However, the diversity of the post-vaccination GMT values corrected for the pre-vaccination status, according to Beyer¹¹, did not persist. The lack of differences between the antibody response of the two age groups, may be explained by results reported by Mosterín Höpping et al.,¹⁷ showing how responses to influenza vaccination in the elderly may be not uniquely influenced by immunosenescence do to aging of the immune system, but also by the different pre-study vaccination and infection histories. However we are not able to examine this aspect and possible differences against the 3 vaccine components since the number of previous vaccinations and natural infections were not available.

Against the B component, the results found that the two age groups were similar, however, considering the GMT values corrected for the pre-vaccination status, an higher response was observed in the younger elderly group as compared with very elderly.

According to these results, the regression model confirmed the best response of the youngest only against the B antigen (Table 2).

Our results, against A/H3N2 and A/H1N1 antigens and against B antigen considering only the data not corrected for the pre-vaccination status, manifested some differences with those reported by Goodwin et al.⁴ that evidenced an impairment in antibody responses following influenza vaccination in people >75 years old.

However, not all the studies included in the Goodwin's review found a lower response in the older vaccines group, furthermore, the elderly people studied by us were all resident in nursing homes instead in his review Goodwin included studies without regard to whether the elderly subjects were living (independently in the community, in institution depend on care or in a mixture of setting). Our previous observations evidenced an higher influenza vaccine immunogenicity in elderly institutionalized volunteers as compared with non institutionalized people.¹⁶

The information related to the immune response after vaccination in subjects over 85 years of age are limited. Those relating to antipenumococcal vaccination, as our data, showed a good response also in subjects over 85 years, although less durable (we did not examine this aspect).¹⁸ Results obtained investigating the relationship between the age and the function of B lymphocytes (measured by in vitro opsonophagocytosis assay) suggested that the quality seems to be affected by the patient's age rather than the quantity.¹⁹ Our results were obtained using HI test and in order to study the functionality of antibody it would be useful to analyze the sera by neutralization test. However neutralization test is more laborious than HI and does not allow for simultaneous analysis of a large number of samples.

The second point we studied was the effect of the introduction of a new antigen in the vaccine composition, since the antigenic composition of the influenza vaccines is reviewed annually and varies to match the strains most prevalent in the hemisphere. To this aim, the vaccine component change predictor that distinguishes between change and not change condition was considered in the regression model. The data in Table 4 and in the lower section of Table 2, showed that the HI responses against the 3 vaccine antigens and especially against A/H1N1 and A/H3N2 antigens, were in prevalence higher in volunteers vaccinated with a new strain as compared with volunteers receiving a vaccine with the same antigens of the previous season. No differences were found in this tendency considering younger and very elderly groups. These data confirm our previous published results supporting the possibility of a slight impairment of HI antibody response against unaltered influenza vaccine antigens, especially for influenza strains that have circulated for prolonged periods of time.²⁰ Other research groups reported a reduced antibody response^{12,13, 21,22} or a lower vaccine effectiveness after repeated vaccination.²³

Furthermore, some studies evidenced that after immunization with the same vaccine components used the previous year, the percentage of elderly subjects that exhibited an high increase (fourfold or more) in antibody titres was appreciably lower as compared with values found in young individuals.^24,25

Fonville et al.,²⁶ introduced a novel method for determining serum antibody titres toward different antigenic variants in response to influenza vaccination and found that vaccine updates improve vaccine efficacy in pre-exposed subjects.

Finally, because different formulations of inactivated trivalent vaccines were used in the 21 winter seasons, we compared the immunogenicity of conventional vaccines (whole, split and subunit vaccines) and of enhanced vaccines (administered intradermally or adjuvanted with MF59) specifically indicated for the vaccination of the elderly people.⁷

Comparison of the results obtained evidenced an higher immunogenicity of the enhanced vaccines, as compared with traditional preparations (Table 5) supporting previously published data.^27-30 In most instances the responses induced in elderly and very elderly individuals were similar. Finally the sex influence on antibody response to influenza vaccination, previously reported and showing a better response in female volunteers, was confirmed by analysis reported in Table 2.

In conclusion, when age, change of vaccine composition, vaccine formulation and sex are included in the model used to study influenza vaccine immunogenicity, it results that sex and change vaccine composition are highly significant for all antigens while the age is an highly significant factor only for the B antigen and vaccine type only for the A/H3N2 antigen.

The HI antibody responses of people aged >85 years were not impaired as compared with those observed in younger elderly with exception of responses against the B antigen evidenced after correction for the pre-vaccination status. These results might be probably due to differences in the pre-vaccination exposure (natural infections or vaccinations) to different antigenic strains of influenza viruses. Moreover, the very elderly subjects examined in our study might represent a selected group of elderly individuals with possible lower effects induced by immunosenescence on antibody responses. However, it was not possible from our data to exclude that the effects of immunosenescence induced impairment of immune responses different from antibody responses. Recently, in accordance with our data, Richardson et al.³¹ comparing a standard dose with a potentiated high-dose (4 times the amount of HA antigen) trivalent inactivated influenza vaccine, evidenced in elderly patients a lower rates of hospitalization for influenza and pneumonia only in the subgroup of patients ≥85 years vaccinated with high-dose vaccine as compared with standard dose vaccine.

All these results support the need to investigate better the immunosenescence phenomenon in order to use the new knowledge for the development of vaccines with improved effectiveness.

Materials and methods

Study design and vaccination

These retrospective study included a total of 2630 elderly people aged ≥60 years (mean age 80, range 60–104) enrolled in 21 seasons. After providing informed consent, all the subjects received 1 dose of trivalent inactivated commercially available influenza vaccine intramuscularly, in the deltoid or intradermally. Each dose of vaccine consisted of 15 µg of HA in a 0.5 ml dose (for vaccines administered intramuscularly) or in a 0.1 ml dose (for vaccines administered intradermally) of each of the three influenza strain antigens (A/H3N2, A/H1N1 and B influenza viruses). At time for recruitment of this study, demographic data, health status (evaluated using the data collected using the Italian VAOR (Valutazione Anziano Ospite di Residenza) schedule) and history of influenza vaccination of the preceding year were obtained for each subjects. Serum samples were obtained from the same subject before and 1 month after vaccination. Subjects were included in this study if they did not have a history of immediate hypersensitivity to eggs components. Subjects suffering from specific illnesses or chronic condition were not excluded. The study was conducted in Italy according to the Declaration of Helsinki and Good Clinical Practices. Since vaccines were assigned by local Health Authorities within the annual influenza campaign and sera were leftover sera from samples collected for clinical routine controls, the study did not need to be registered as a formal trial.

Vaccine immunogenicity and statistical analysis

HI antibody titers were determined using a standard microtiter method with 0.5% chicken (before 1996–1997) or turkey erythrocytes.³²

All sera were heat-inactivated at 56° C for 30 minutes and treated with potassium periodate and trypsin (before 1994) or with receptor-destroying enzyme (RDE) of Vibrio cholerae to remove non-specific inhibitors. The first dilution for antibody titration was 1:10. Pre- and post-vaccination sera from each of the vaccinees were frozen at −30°C until used and tested simultaneously for HI antibody titers using the same antigens as those in the vaccine.

Vaccine immunogenicity was assessed by comparing antibody titers in blood samples, collected before and 1 month after vaccination, as protection rate and GMT (any HI antibody titer <10 was considered equal to 5 for GMT calculation).

Statistics and statistical analysis considered in this paper were evaluated as described in Camilloni et al.⁵ Student's t test were used for comparing the means between groups and p values <0.01 were considered highly statistically significant, whereas p values <0.05 were regarded as marginally statistically significant.

Pre- and post-vaccination titers levels were divided by 5 and log₂-transformed (T_pre and T_post- values), and an ordinary least square (OLS) regression was applied using the fitlm function in Matlab® of MathWorks Inc. release 2014b. Groups means of the log-transform titers were then presented as geometric mean titers (GMTs) in Tables 2–4. The starting regression model was:(1) $T_{post}=A+B_{pre}*T_{pre},$(1) with A the y-axis intercept and B_pre the slope coefficient of the independent variable T_pre. This B_pre coefficient was adopted for correcting post-vaccination titers for the imbalance due to the pre-vaccination status and these compensated values were used for calculating the corrected post-GMT of each group in the paper according to Beyer¹¹.

In order to evaluate effects on immunogenicity of sex, age, vaccine type and vaccine components change, the previous linear regression model was improved by adding the sex group (G_sex: female = 0, male = 1), the age group (G_age: elderly = 0, very elderly = 1), the vaccine type group (G_vaccType: enanched = 0, conventional = 1) and the change group (G_ch: change = 0, non-change = 1) as independent variables thus obtaining:(2) $T_{post}=A+B_{pre}*T_{pre}+B_{sex}*G_{sex}+B_{age}*G_{age}+B_{vaccType}*G_{vaccType}+B_{ch}*G_{ch}.$(2)

The corresponding regression coefficients were indicated with B_pre, B_sex, B_age, B_vaccType, B_ch, respectively. All analyses were run for the three antigens separately. The ANCOVA test was executed for testing significance of differences in slopes between groups.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.