1 Background

HB vaccination in Romania was implemented through the National Immunization Program (NIP) in 1995 in newborns with 4 doses (ages 0-7 days with Engerix, 2 months, 6 months with DTP-HepB Tritanrix-HepB [http://www.cdep.ro/pls/legis/legis_pck.htp_act_text?idt=22076]). Subsequent vaccination campaigns were implemented in 3rd grade children (approx. age 9) between 1998-2008 and in 12th grade high-school students (approx. age 18) between 2004-2008. The resulting cohorts are further detailed in Methods.

2 Methods

2.1 Study design

We performed a retrospective study which included both hospitalized and ambulatory patients examined during 2014-2018 in the Teaching Hospital of Infectious Diseases, Cluj-Napoca, Romania. Data received from our hospital’s electronic database consisting of anti-AgHBs titres along with all demographic and clinical information recorded at the moment of examination. For the main part of the research, we excluded infants (aged ≤1 years old, analyzed separately). Then, we excluded patients with known hepatitis B history (previous diagnosis and / or positive AgHBc, AgHBe, viral load). Lastly, we excluded duplicated examinations, leaving only the first remaining anti-AgHBs measurement for each patient.

We reported separately the protection levels in several special groups: infants, declared medical students and staff, pergnant women, partners of confirmed HB cases and participants accidentally exposed to blood of nknown or unknown origin and HB status. As a proxy for HB incidence, we counted the number of hospitalized acute HB cases (ICM10 codes “B16”, “B18.0” and “B18.1”) which occured in our institution during the same period in cohorts demographically corresponding to the birth-cohorts used in the main study.

2.2 Anti-AgHBs titres

Our laboratory measured Anti-AgHBs by electrochemiluminescence (Vidas BioMerieax) and reported titres as either quantitative data (PEI/L) or qualitative data if measured values were outside detection limits (<5 and >300 during Jan-Mar 2014, <3 and >500 since Apr 2014 and <0.001 and >1000 for some 2018 samples taken from ambulatory participants). These thresholds did not impede calculations of the percentages of participants with protective titres (>10 PEI/L) or detectable titres (above the inferior detection threshold) but posed a challenge in measuring mean and median titres.

2.3 Cohorts formation and Lexis diagram

Based on vaccination program, several birth-cohorts were formed (Table 1). The percentage of patients with protective titres by years of examination and birth was superimposed as a heatmap on a Lexis diagram (a plot of age by calendar time, used for visualizing cohorts, see results section) [cartea cu diagrame Lexis] which made several patterns apparent, roughly reflecting cohorts formed by vaccination policy. Additionally, the Lexis diagram revealed a sharp separation line around birth-year 2005 within patients vaccinated at birth and we decided to separate these into birt-cohorts A1 (born ≥2005) and A2 (born 2001-2004, incl.). The separation was chosen based on the visual aspect of the Lexis chart and we confirmed it by demonstating that this treshold leads to the lowest Log OR of having a protective titre in cohorts A2 compared to A1 (Log OR = -1.94, 95% CI = -2.76 to -1.1). Tresholds 2006 and 2007 led to slightly higher Log OR with narrower 95% CIs and may also be considered appropriate tresholds.

Table 1: Birth-cohorts formed by vaccination campaigns and the hypothesized separation between cohorts A1 and A2.

Birth-cohort [years, incl.]	NIP vaccination status	Years of vaccination	Time since vaccination
E: ≤1985	not part of NIP; may have been independently vaccinated or have attained protection through acute infection
D: [1986-1990]	as 12th grade high-school students (approx. age 18)	2004-2008	6-14 years
C: [1991-1994]	as 3rd grade pupils (approx. age 9)	2000-2003	11-18 years
B: [1995-2000]	at birth OR catch-up as 3rd grade pupils (approx. age 9)	2004-2009*	13-23 years*
A2: [2001-2004]	at birth	2001-2004	10-17 years
A1: ≥2005	at birth	≥2005	2-12 years
* Cohort B was assumed te be vaccinated last time at birth.

2.4 Statistical analysis

Preliminary data cleaning and management was performed in spreadsheet and all subsequent statistical analysis were performed in R 3.6.2. We used basic descriptive (numeric and absolute frequencies, means ±standard deviations, medians and inter-quartile ranges, geometric means and geometric standard deviations for antibody titres) and comparative methods (Chi-squared or Fisher tests, T or Mann-Whitney tests, ANOVA or Kruskal-Wallis test). All multiple-comparisons corrections of p-values was performed adjusting for false-discovery rates [Benjamini Y, Hochberg Y: http://www.jstor.org/stable/2346101]. We chose to report results based on both quantitative data (means ±standard deviations, median with inter-quartile ranges and geometric means with geometric standard deviations) as well as based on both quantitative and qualitative data (% of patients with protective and detectable titres, median titres using interval-censored methods (inverse commulative proportion of patients with protective titres by cohort) [Akritas 1994, articol censored regression].

We studied the correlation between estimated time since last vaccination (cohort B assumed vaccinated at birth) and proportion of patients with protective titres, overall and separately for each NIP-vaccinated cohort. Linear trends and Pearson correlation coefficients were weighted by the number of patients within each integer duration and chohort. We computed logistic models for odds of having a protective titres predicted by NIP-vaccinated cohort (refferenced to cohort A1), estimated time since last vaccination (normalized to z-scores) and both predictors simultaneously. We did not adjust for age becasue it is coded by cohort and duration.

Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. 1995; Series B, 57, 289–300. http://www.jstor.org/stable/2346101.

Fay MP, Shaw PA. Exact and Asymptotic Weighted Logrank Tests for Interval Censored Data: The interval R package. J Stat Softw. 2010 Aug;36(2):i02. PMID: 25285054; PMCID: PMC4184046.

3 Analyses

3.1 Separation A1/A2

3.2 Separation A1/A2+B

3.3 Time since vaccination

3.4 Logistic models

3.5 Descriptive tables

3.6 Bar titre

3.7 Vio titre

3.8 KW test

3.9 Conover Iman test

3.10 Lexis

3.11 Censored statistcs

3.12 Special groups

4 Results

4.1 Demographic composition

We collected data from a total of 2963 unique participants (78.5% outpatient, 64.1% female, ages 2 to 97, median=25 years old) with significant demographic differences between cohorts (Table 2). Cohorts C and D were most represented, averaging 212 and 109 patients / birth-year respectively. The proportion of hospitalized patients was highest in cohorts E and A. Cohorts A1 and A2 had an almost sex ratio while older cohorts had a higher proportion of men.

Table 2: Summary of demographic parameters by cohort. μ ±SD = Mean (standard deviation); M (range) = Median (min:max); KW: Kruskal-Wallis test.

Cohort: birth-years, incl. N (%)		E: ≤1985	D: 1986-1990	C: 1991-1994	B: 1995-2000	A2: 2001-2004	A1: ≥2005	Overall
Cohort		1043 (35.2%)	543 (18.3%)	847 (28.6%)	337 (11.4%)	63 (2.1%)	130 (4.4%)	2963
Data origin	outpatient	679 (65.1%)	445 (82.0%)	794 (93.7%)	273 (81.0%)	40 (63.5%)	94 (72.3%)	2325 (78.5%)	V=0.29 (p<0.001)
Data origin	inpatient	364 (34.9%)	98 (18.0%)	53 (6.3%)	64 (19.0%)	23 (36.5%)	36 (27.7%)	638 (21.5%)	V=0.29 (p<0.001)
Age (years)	M (range)	44 (29:97)	27 (24:32)	24 (20:27)	21 (14:23)	14 (10:17)	6 (2:13)		Kruskal-Wallis: p<0.001
Age (years)	μ ±SD	46.43 ±12.1	26.92 ±2.19	23.92 ±1.17	20.17 ±2.12	14.03 ±1.87	6.68 ±2.99	31.00 ±14.2	Kruskal-Wallis: p<0.001
Sex	F	624 (59.8%)	363 (66.9%)	613 (72.4%)	208 (61.7%)	31 (49.2%)	61 (46.9%)	1900 (64.1%)	V=0.14 (p<0.001)
Sex	M	419 (40.2%)	180 (33.1%)	234 (27.6%)	129 (38.3%)	32 (50.8%)	69 (53.1%)	1063 (35.9%)	V=0.14 (p<0.001)
μ ±SD = Mean (standard deviation); M (range) = Median (min:max); V = Cramér V (p value from Chi² test);

4.2 Lexis chart

The visual aspect of the Lexis chart (Figure 1) revealed a relatively high proportion of protected patients within cohorts C, D and the youngest part of cohort A, while cohort B and the rest of cohort A appeared with a lower proportion. The separation lines between cohorts A1 / A2 and B / C were sharp and aligned with birth-years 2005 and 1995 while the other separation lines were less apparent. Cohort E appeared heterogeneous, with a slight decrease in the propotion of protected patients with age and blending gradually with cohort D.

Figure 1: Lexis diagram. Highlited areas and the separation lies between them were formed by the vccination policies applicable to the participants.

4.3 Anti-AgHBs titres

By all aggregate measures, cohort A2 (and to lesser extents, cohort B), had the lowest protection, cohorts C and D had the best protecton and cohorts A1 and E had intermediate protection levels. Only 19% of cohort A2, 45% of cohorts B and E, 62% of cohort A1, had protective titres, compared to 71% and 74% of cohorts C and D respectively. Post-hoc pairwise comparrisons (Supplementary table 6) found signifficant differences in the proportion of protected participants between all cohorts except B vs. E (45.4% and 44.8%), C vs. D (74.3% and 71.3%) and A1 vs. D (62.3% and 71.3%).

Taking into account censored data, we found median titres below the detection limit of 3 PEI/L in cohort A2 and below the detection limit of 5 PEI/L in cohort E. Cohort B also showed a median titre below the protective level of 10 PEI/L. None of the cohorts had median titres above the robust protection level of 100 PEI/L. Post-hoc pairwise comparrisons (Supplementary table 7) found signifficant differences in the cumultive titre distribution between all cohorts except A1 vs. E (medians 24 and 4).

Table 3: Summary data for the cohorts. Results from both quantitative and qualitative titre data (proportions of participatants with protective, detectable and >100 PEI/L titre levels, median titre) and quantitative data only (median and geometric mean of titres).

		Qualitative and quantitative titre data:				Quantitative titre data only:
Cohort	n=2963	% protective	% detectable	% >100	Median	Median	Geometric mean
E: ≤1985	1043	44.8%	58.1%	28.9%	4.0	23.0	6.7
D: [1986-1990]	543	71.3%	82.0%	49.5%	97.0	75.0	33.2
C: [1991-1994]	847	74.3%	87.8%	40.0%	55.0	48.9	40.8
B: [1995-2000]	337	45.4%	70.6%	18.1%	6.0	15.8	8.9
A2: [2001-2004]	63	19.0%	52.4%	7.9%	0.6	7.0	3.7
A1: ≥2005	130	62.3%	84.6%	22.3%	24.2	29.3	21.2

The cohors had significantly different (Asymptotic Logrank k-sample test: p<0.001) cumulative distribution curves for anti-HB antibody titres, taking into account censored measurements. Visually, cohort A2 had the lowest titres, followed by cohorts B and E (wich had a higher proportion of robust titres than B), while cohorts C, D and A1 (which had a lower proportion of robust titres) had the highest titres. Post-hoc pairwise comparrisons (Supplementary table 7) found signifficant differences in the cumultive titre distribution between all cohorts except A1 vs. E (medians 24 and 4).

Figure 2: Comparrison between chohorts, using both quantitative and qualitative data. Qualitative data was considered interval-censored.

4.4 Additional studies

4.5 Special groups

Almost 90% of the infants had protective titres, and most of them at high values. The majority of medical staff / students and pregnant women were protected while the majority of the participants exposed to blood (source status: confirmed, suspected or unknown) and sexual partners of known HP patients were not. In the case of sexual contacts, the median titre was below the detection limit of 5 PEI/L (Table 4, Figure 2).

Table 4: Summary data for special groups. Results from both sets of data: quantitative and qualitative, quantitative only.

		Qualitative and quantitative titre data:			Quantitative titre data only:
Group	n, Age	% protective	% detectable	Median	Median	Geometric mean
Infants*	45, 0.4 ±0.5	88.7%	88.7%	253.0	138.0	100.8
Pregnant women	15, 28.5 ±4.2	66.7%	73.3%	21.0	25.5	7.4
Medical staff and students	74, 27.6 ±10.1	75.7%	81.1%	52.5	54.0	56.4
Accidental exposure to blood	19, 36.7 ±16.9	47.4%	57.9%	6.0	23.9	11.5
Infected sexual partner	33, 32.6 ±11.7	39.4%	54.5%	4.0	18.0	14.9

Figure 3: Special groups, using both quantitative and qualitative data. Qualitative data was considered interval-censored.

4.6 Time since last vaccination

The proportion of participants with protective titres showed an oscilating downwards trend with the estimated time since latest NIP vaccination (overall weighted R = -0.51) and the patterns varried across cohorts. Patients vaccinated only at birth (cohorts A1 and A2) showed an important linear decrease from 2 to 17 years post-vaccination (weighted R = -0.87), patients vccinated at birth who may have received a catch-up dose at age 9 (cohort B) showed a linear ascending trend (weighted R=0.75) from 14 to 23 years after the dose received at birth and patients vaccinated only in school (cohorts C and D) showed relatively stable protection (weighted R = -0.31) from 6 to 17 years after the vaccine received at ages 9 or 18.

Figure 2: Correlation between the estimated time since last vaccination (assumed at birth for cohort B) and proportion of participants with protective titre. Weighted linear correlation estimates. Facets show cohorts A1 & A2, B, C & D and the overall sample having different patterns.

4.7 Logistic models

Logistic models showed odds of having protective tites significnatly decreasing with time since latest NIP vaccination (transformed to z-scores), both overall (OR = 0.74) as well as adjusting for birth-cohort (Adj-OR = 0.72). Compared to the youngest cohort (A1: ≥2005, vaccinated at brith), cohorts A2 (OR = 0.14) and B (OR = 0.50) showed significantly lower unadjusted odds of protective titres but only cohort A2 (Adj-OR = 0.24) had signficantly lower odds of protective titre adjusted for time since latest NIP vaccination. Both cohots C and D showed significantly higher odds of having protective titres both undjusted (OR = 1.75 and 1.50, respectively) and adjusted for time (Adj-OR = 3.1 and 1.76, respectively). Compared to A2 and A2+B cohorts, all other cohoths had significantly higher odds of protective titres, both unadusted and adjusted for duration sice latest NIP vaccination (Supplementary charts).

Figure 3: Logistic models’ ORs of having a protective titre by estimated time since last vaccination (z-scores) and cohort (A1 used as refference).

5 Supplementary results

The separation line between cohorts A1 and A2 at birth-year 2005 was based on the visual aspect of the Lexis chart and therfore subjective. A more formal was to find which birth-year optimimized the OR of having a protective titres between these birt-cohorts, while having a narrow 95% CI not including 1. We tred all possible tresholds within cohorts A1-A2 and found that year 2005 resulted in the lowest significant Log-OR (-1.94, 95% CI = -2.76 to -1.18). Years 2006 and 2007 also resulted significant ORs, of higher value but having narrower 95% CI and could also be cosidered possible separation tresholds. A parabolic fit line, weighted by 1/Log SE of OR had the inflection point closest to 2007, further reinforcing tha validity of a separation treshold at birth-year 2007. Eventually, we chose treshold 2005 because it was a closer match of the Lexis diagram and because it resulted in the most extreme significant OR.

Table 3: Possible separation tresholds between cohorts A2 and A1 by Log OR for having a protective titre in cohort A2 vs. A1. Participants born during the treshold year will be included in cohort A1.

Treshold (year of birth)	Log OR (95% CI)	p-value
2002	-0.90 (-2.41 to 0.39)	0.168
2003	-0.90 (-1.83 to -0.05)	0.034
2004	-1.47 (-2.32 to -0.68)	<0.001
2005	-1.94 (-2.76 to -1.18)	<0.001
2006	-1.78 (-2.50 to -1.10)	<0.001
2007	-1.78 (-2.47 to -1.12)	<0.001
2008	-1.50 (-2.16 to -0.86)	<0.001
2009	-1.25 (-1.90 to -0.62)	<0.001
2010	-1.06 (-1.72 to -0.42)	<0.001
2011	-1.37 (-2.12 to -0.66)	<0.001
2012	-1.00 (-1.83 to -0.22)	0.008
2013	-0.42 (-1.32 to 0.46)	0.316

Figure 4: Possible separation tresholds between cohorts A2 and A1. Participants born during the treshold year will be included in cohort A1. Log OR were for having a protective titre in cohort A2 vs. A1. Filled circles show significant ORs (Fisher test). A. Log OR with 95% CI by year; size shows SE of Log OR; Second order polynomial fit line. B. Each year’s Log SE for Log OR.

5.1 Separation A1/A2+B

The visual aspect of the Lexis chart was indistinguishable between cohorts A2 and B but different from cohort A1, therefore we also tried the same method on the cumulated cohorts A2 & B compared to A1. This way, we found two significant local minima of OR at years 2007 and 2011. Althow year 2011 resulted in a lower Log-OR, it was less supported by the Lexis chart and had a larger SE than the 2007 treshold, which was also closer to the infection point of the weighed polynomial fit at year 2008. We chose not to pursue this approach since cohorts A2 and B were heterogeneous both demographically and in the relation between time since last vaccination and proportion of protected participants and joining these cohorts was not justified.

Table 4: Possible separation tresholds between cohorts A2+B and A1 by Log OR for having a protective titre in cohort A2+B vs. A1. Participants born during the treshold year will be included in cohort A1.

Treshold (year of birth)	Log OR (95% CI)	p-value
1996	1.59 (1.06 to 2.16)	<0.001
1997	0.96 (0.57 to 1.34)	<0.001
1998	0.46 (0.10 to 0.82)	0.011
1999	0.21 (-0.14 to 0.57)	0.223
2000	-0.04 (-0.40 to 0.32)	0.860
2001	-0.11 (-0.48 to 0.26)	0.587
2002	-0.20 (-0.58 to 0.18)	0.311
2003	-0.31 (-0.70 to 0.08)	0.107
2004	-0.56 (-0.97 to -0.16)	0.005
2005	-0.85 (-1.29 to -0.43)	<0.001
2006	-0.95 (-1.41 to -0.51)	<0.001
2007	-1.06 (-1.54 to -0.59)	<0.001
2008	-0.97 (-1.46 to -0.49)	<0.001
2009	-0.91 (-1.43 to -0.41)	<0.001
2010	-0.85 (-1.41 to -0.31)	<0.001
2011	-1.18 (-1.85 to -0.54)	<0.001
2012	-0.94 (-1.71 to -0.22)	0.008
2013	-0.45 (-1.32 to 0.38)	0.250

Figure 5: Possible separation tresholds between cohorts A+B2 and A1. Participants born during the treshold year will be included in cohort A1. Log OR were for having a protective titre in cohort A2 vs. A1. Filled circles show significant ORs (Fisher test). A. Log OR with 95% CI by year; size shows SE of Log OR; Second order polynomial fit line. B. Each year’s Log SE for Log OR.

5.2 Lexis diagrams by other measures

The aspect of the Lexis chart was similar with all other aggregate measures: mean, median and geometric mean of anti-AgHBs titre on quantitative data, median anti-AgHBs titre on both quantitative and censored data and proportion of patients with detectable titres.

The majority of participants showed detectable titres in cohorts A1, C and D but not in cohorts A2, B and E. There were sharp separation lines beween cohorts A1 / A2 and B / C and a more grdual transition between cohorts D / E.

Figure 6: Lexis diagram.

Few participants showed titres >100 PEI/L in all cohorts, with relatively higher proportion in cohort D and, to a lesser extent, C. A1, Cohort E appeared heterogeneous with a slight tendency towards a lower proportion of participants with robust titres with age. There were relatively sharp separation lines beween cohorts B / C and C / E and a grdual transition between cohorts D / E. Cohorts A1, A2 and B were indistinguishable.

Figure 7: Lexis diagram.

Taking into account censored values, the median titres were above the protective level of 10 PEI/L in cohorts A1, C, D and E but not in cohorts A2 and B. There were sharp separation lines beween cohorts B / C and a grdual transition between cohorts A1 / A2 and A2 /B. Cohorts D and E were indistinguishable.

Figure 8: Lexis diagram.

Taking into account only numerical data, the median titres were above the protective level of 10 PEI/L in cohorts A1, C, D and E but not in cohorts A2 and B. There were sharp separation lines beween cohorts B / C and a grdual transition between cohorts A1 / A2. Cohorts A2 & B, and A1, C, D & E were indistinguishable. Numerical data throughout 2019 included lower values due to a the use of a lower detection limit (0.001 PEI/L) in some outpatient participants.

Figure 9: Lexis diagram.

Taking into account only numerical data, the geometric mean titres were above the protective level of 10 PEI/L in cohorts A1, C, D and E but not in cohorts A2 and B. There were sharp separation lines beween cohorts B / C and a grdual transition between cohorts A1 / A2. Cohorts A2 & B, and A1, C, D & E were indistinguishable. Numerical data throughout 2019 included lower values due to a the use of a lower detection limit (0.001 PEI/L) in some outpatient participants.

Figure 10: Lexis diagram.

Taking into account only numerical data, the mean titres were above the protective level of 10 PEI/L in cohorts A1, C, and D but not in cohorts A2, B and E. There were sharp separation lines beween cohorts A1 / A2, B / C and D / E. Cohorts A2 & B, and A1, C & D were indistinguishable. Numerical data throughout 2019 included lower values due to a the use of a lower detection limit (0.001 PEI/L) in some outpatient participants.

Figure 11: Lexis diagram.

5.3 Pairwise comparrisons

Table 5: Pairwise comparrisons between all chohorts, using both quantitative and qualitative data: Chi-square test for % with protective titres by cohort. P-values adjusted by controlling for false discovery rate, according to Benjamini Y and Hochberg Y (1995).

Pairwise	%	E: ≤1985	D: [1986-1990]	C: [1991-1994]	B: [1995-2000]	A2: [2001-2004]	A1: ≥2005
E: ≤1985	44.8%		p<0.001	p<0.001	p=0.890	p<0.001	p<0.001
D: [1986-1990]	71.3%	p<0.001		p=0.261	p<0.001	p<0.001	p=0.068
C: [1991-1994]	74.3%	p<0.001	p=0.261		p<0.001	p<0.001	p=0.008
B: [1995-2000]	45.4%	p=0.890	p<0.001	p<0.001		p<0.001	p=0.002
A2: [2001-2004]	19.0%	p<0.001	p<0.001	p<0.001	p<0.001		p<0.001
A1: ≥2005	62.3%	p<0.001	p=0.068	p=0.008	p=0.002	p<0.001
Overall Chi-squared test: p<0.001

Table 6: Pairwise comparrisons between all chohorts, using both quantitative and qualitative data. Qualitative data was considered interval-censored, therefore appropriate logrank methods were performed according to Fay MP and Shaw PA (2010). P-values adjusted by controlling for false discovery rate, according to Benjamini Y and Hochberg Y (1995).

Pairwise	E: ≤1985	D: [1986-1990]	C: [1991-1994]	B: [1995-2000]	A2: [2001-2004]	A1: ≥2005
E: ≤1985		p<0.001	p<0.001	p=0.022	p<0.001	p=0.902
D: [1986-1990]	p<0.001		p=0.043	p<0.001	p<0.001	p<0.001
C: [1991-1994]	p<0.001	p=0.043		p<0.001	p<0.001	p<0.001
B: [1995-2000]	p=0.022	p<0.001	p<0.001		p=0.003	p=0.027
A2: [2001-2004]	p<0.001	p<0.001	p<0.001	p=0.003		p<0.001
A1: ≥2005	p=0.902	p<0.001	p<0.001	p=0.027	p<0.001
Asymptotic Logrank k-sample test (permutation form): p<0.001

Table 7: Pairwise comparrisons between all chohorts, using only quantitative data. P-values adjusted by controlling for false discovery rate, according to Benjamini Y and Hochberg Y (1995).

Pairwise	E: ≤1985	D: [1986-1990]	C: [1991-1994]	B: [1995-2000]	A2: [2001-2004]
D: [1986-1990]	p<0.001
C: [1991-1994]	p<0.001	p=0.161
B: [1995-2000]	p=0.411	p<0.001	p<0.001
A2: [2001-2004]	p=0.142	p<0.001	p<0.001	p=0.339
A1: ≥2005	p=0.760	p=0.001	p=0.009	p=0.024	p=0.007
Kruskal-Wallis rank sum test: p<0.001

Table 8: Pairwise comparrisons between all chohorts, using only quantitative data. P-values from Conover-Iman post-hoc test, adjusted by controlling for false discovery rate, according to Benjamini Y and Hochberg Y (1995).

Pairwise	E: ≤1985	D: [1986-1990]	C: [1991-1994]	B: [1995-2000]	A2: [2001-2004]	A1: ≥2005
E: ≤1985		p<0.001	p<0.001	p=0.036	p=0.013	p=0.584
D: [1986-1990]	p<0.001		p=0.222	p<0.001	p<0.001	p<0.001
C: [1991-1994]	p<0.001	p=0.222		p<0.001	p<0.001	p=0.010
B: [1995-2000]	p=0.036	p<0.001	p<0.001		p=0.197	p=0.033
A2: [2001-2004]	p=0.013	p<0.001	p<0.001	p=0.197		p=0.008
A1: ≥2005	p=0.584	p<0.001	p=0.010	p=0.033	p=0.008
Kruskal-Wallis rank sum test: p<0.001

5.4 Time since last vaccination

The proportion of participants with protective titres showed an oscilating downwards trend with the estimated time since latest NIP vaccination (overall weighted R = -0.51) and the patterns varried across cohorts. Among patients vaccinated at birth, cohort A1 showed a linear decrease from 2 to 12 years post-vaccination (weighted R = -0.58), while patients cohorts A2 (vaccinated only at birth) and B (vaccinated at birth but may have received a catch-up dose at age 9) showed ascending linear trends (weighted R = 0.85 and 0.75, respectively) from 10 to 17 and 13 to 23 years after the dose received at birth. Patients vaccinated only in school at age 9 (cohort C) showed relatively stable protection (weighted R = -0.51) from 11 to 18 years after vaccination and patients vaccinated only in school at age 18 showed a downwards linear trend (weighted R = -0.93) from 6 to 14 years after vaccination.

The proportion of participants with protective titres showed an oscilating downwards trend with the estimated time since latest NIP vaccination (overall weighted R = -0.51) and the patterns varried across cohorts. Patients vaccinated at birth (cohorts A1, A2 and B) showed little linear decrease from 2 to 23 years post-vaccination (weighted R = -0.17) but a “U”-shaped parabola or higher order polynomials may be a better fit for this relation, which constitutes a reason to study these cohorts separately. Patients vaccinated only in school (cohorts C and D) showed relatively stable protection (weighted R = -0.31) from 6 to 17 years after the vaccine received at ages 9 or 18.

Variable anti-AgHBs titres in vaccinated birth cohorts - a cross-sectional population-based study

Titruri variabile de anticorpi anti-AgHBs la cohorte vaccinate - studiu populațional transversal