Studies were eligible for inclusion if they fulfilled the following criteria: 1) reported effectiveness of HPV vaccination against infection, anogenital warts, or cervical abnormalities (based on cytological or histological outcomes); 2) assessed effectiveness of HPV vaccination by number of doses received (one, two, or three). We excluded studies if vaccination was administered as part of a clinical trial (e.g., post hoc evaluations of clinical trials).

In the previously published review, we first searched Medline and Embase databases from January 1, 2007, through June 15, 2017. For this review, we updated the search through September 29, 2021. We used the same methodology in the original and updated searches, which was a combination of Medical Subject Headings (MeSH) terms, title or abstract words, without restriction on language of publications: (“papillomavirus vaccines”, “HPV vaccine”, “HPV vaccination”, “papillomavirus vaccine”, or “papillomavirus vaccination”) and (“program evaluation”, “immunization programs”, “population surveillance”, “sentinel surveillance”, “incidence”, “prevalence”, “rate”, “rates”, “effectiveness”, “doses”) and (“papillomavirus infections”, “HPV”, “uterine cervical neoplasms”, “cervical intraepithelial neoplasia”, “HPV related diseases”, “condylomata acuminata”, “genital warts”). The selection of eligible articles was performed first by NP on title and abstract, and second by NP and MD on the full-text article.

Three authors (NP, LM, RL) independently extracted main study characteristics and outcomes using standardized forms. Discrepancies were resolved by a fourth author (MD). The main study characteristics included country, study design, age at vaccination and at outcome assessment, sample size according to the number of doses received, case definition, and statistical analyses (procedure used to assign the number of doses and adjustment for potential confounders).

Two teams that were part of this review (Université Laval and Centers for Disease Control and Prevention (CDC)) independently assessed quality; discrepancies were resolved by consensus. We assessed included studies for selection bias, information bias, and confounding based on the Risk Of Bias In Non-Randomized Studies - of Interventions (ROBINS-I). We used the same main sources of biases and ratings as the ROBINS-I, but adapted questions to consider the particularities of reduced dose observational studies. For selection bias, we examined whether selection of participants could be influenced by participants’ characteristics or outcome. For information bias, we examined potential biases in measurement of intervention (e.g., validity of data sources to determine dose groups, sufficient interval between first and second dose among two-dose recipients) and in measurement of outcome (e.g., validity of the algorithm used to identify outcomes, use of lag time or buffer period between time of vaccination and counting of outcome to exclude outcomes originating from prevalent infections at a given dose). For confounding, we examined the likelihood of differences between dose groups in: 1) prevalence of HPV infection at first dose, 2) risk of HPV acquisition during study follow-up, and 3) immunogenicity (for studies with formal comparisons between three, two and one doses). We also examined methods used to control for these potential confounders.

For each domain, possible ratings were “Low”: confident that there is little chance for bias; “Moderate”: unlikely that there is a substantial bias, but a slight bias is possible; “Serious”: significant possibility of a substantial bias; “Critical”: confident that a substantial bias exists. The risk of bias in each category was assigned based on the highest (worst) domain rating within that category. If more than one age group was used in the analyses, we rated the age group with the lowest risk of bias. Because one aim of this systematic review was to discuss study limitations, no studies were excluded on the basis of methodological quality. Quality assessment findings were compiled in a descriptive synthesis.

A narrative synthesis was conducted using data reported in the publications. The main outcome was effectiveness of HPV vaccination comparing the incidence or prevalence of HPV-related endpoints between individuals vaccinated with different numbers of doses (three vs none, two vs none, one vs none, three vs two, three vs one, two vs one) of quadrivalent HPV vaccine (Gardasil, 4vHPV) or bivalent HPV vaccine (Cervarix, 2vHPV). Results are presented as crude or adjusted risk ratios (RR), hazard ratios (HR), incidence rate ratios (IRR), prevalence ratios (PR), or odds ratios (OR). Because eligible studies used different buffer periods or age groups at vaccination and at outcome assessment, heterogeneity between studies was significant. Therefore, it was not possible to pool results from the different studies.

In the updated search from June 16, 2017, through September 29, 2021, 3784 additional articles were identified, 116 full text articles were assessed and 21 included. Overall, there were 35 articles in the review, 14 from the initial review[11–24] and 21 from the update.[25–45] (Figure 1). The 35 articles were from 12 countries: United States (eleven), Scotland (six), Australia (four), Denmark (three), Sweden (two), Canada (two), and one each from Belgium, Denmark/Sweden, Italy, Mongolia, Netherlands, Spain, and New Zealand (Table 1, Table 2, Table S1).

All studies except one were conducted within the context of a recommended three-dose schedule. Evaluations were in countries that used 2vHPV (seven), 4vHPV (27) or both (one). Nine studies evaluated effectiveness for prevention of prevalent (detected at a single time point) or incident vaccine-type (4vHPV or 2vHPV) HPV infection,[11, 12, 25–31] ten anogenital warts,[13–18, 32–35] and 16 cervical cytological or histological abnormalities[19–24, 36–45].

All 35 studies were determined to be at moderate or serious risk of bias (Table 1, Tables S2-S5, Figures S1–S6). No study had a domain rated as critical. Comparisons involving one and two doses were more likely to be affected by serious biases than three doses. Of seven potential sources of biases, three were more likely to be rated at serious risk of bias. These serious biases fell in two broad categories, i.e., information bias and confounding; each are likely to underestimate the effectiveness of one and two doses.

First, the majority of studies were deemed at moderate or serious risk of information bias for measurement of outcome (Figures S1, S4–S6). This is because prevalent infections at a given dose could cause an attribution of outcome to the wrong dose if outcomes are detected after a given dose but originate from an infection acquired before that dose. Studies using buffer periods to exclude outcomes originating from prevalent infections (≥12 months for infection, ≥six months for anogenital warts, ≥24 months for cervical abnormalities)[46–48] or restricting analyses to girls vaccinated at an age when they were less likely to have prevalent infections were deemed at lower risk of information bias. Of note, while buffer periods decrease the likelihood that a prevalent infection at the time of a given vaccination was responsible for the outcome detected, they do not guarantee this.

Second, studies examining two-dose effectiveness were deemed at serious risk of information bias in measurement of intervention if the interval between the two doses was less than five months (Figures S2, S4–S6).[2–4] Because the majority of studies were conducted when a three-dose schedule was recommended, the interval between the first and second dose was often only one or two months; longer intervals were found when individuals were late for the second dose in a recommended three-dose series.

Third, the majority of studies were at moderate or serious risk of confounding as a result of differences between dose groups in prevalence of HPV infection at first dose or start of follow-up and/or in risk of HPV acquisition during follow-up (Figures S3–S6). The use of buffer periods or restriction of analysis to younger age groups could control for differences in prevalent infection at first dose between dose groups. More recent studies tend to include individuals vaccinated at a younger age and were less likely to be affected by biases related to prevalent infections at vaccination. Few studies were able to control for the potential difference in risk of HPV acquisition between dose groups by adjusting for sexual activity during follow-up (Tables S3–S5).

In the original review, two studies reported vaccine effectiveness for prevention of prevalent vaccine-type infection, both from Scotland where 2vHPV was introduced.[11, 12] In the updated review, seven additional studies were identified, two from countries were 2vHPV was used (Scotland and Netherlands[30, 31]), and five from countries where 4vHPV was used (four from the United States[25–28] and one from Mongolia[29]) (Table 1). In five of the nine studies, three-, two- and one-dose vaccine recipients were compared with those unvaccinated;[11, 12, 27, 28, 30] four included a formal comparison between those with different numbers of doses (Table 2).[25–28] One study was conducted after a 2vHPV two-dose schedule was implemented in the Netherlands.[31] None of the studies assessed effectiveness using different buffer periods or intervals between doses in two-dose vaccine recipients.

Figure 2 shows estimates from eight studies that provided vaccine effectiveness estimates against HPV infection, either overall or limited to persons vaccinated at ≤18 years. Of five studies that evaluated effectiveness with three doses (compared to no vaccination), all found significant effectiveness.[11, 12, 27, 28, 30] Five of six found some significant effectiveness with two doses,[12, 27, 28, 30, 31] and four of six with one dose.[12, 27–29] Six studies conducted analyses limited to persons vaccinated at a younger age or only included such persons in the study.[11, 12, 28–31] In general, effectiveness estimates were higher with younger age at vaccination. In the one study conducted in the setting of a routine 2vHPV two-dose schedule, girls who were vaccinated at age 12–13 years were followed prospectively with self-collected vaginal swabs for HPV DNA.[31] The adjusted VE of two doses against incident HPV16/18 infection was 84% (95% CI 27.0, 96.5).

The three reports from Scotland used the same data sources, specifically data from assessment of vaccine-type HPV prevalence in women attending their first cervical screen generally at age 20–21 years,[28] immunization records from the Scottish Immunization Recall System, and other national registries. The first Scottish study reported statistically significant effectiveness in adjusted analyses for three doses but not for two or one doses.[11] The two subsequent studies found significant effectiveness for three and two doses.[12, 30] Effectiveness of one dose was observed when one dose vaccinees were oversampled and in some adjusted analyses of a cohort that included girls vaccinated at routine ages as well as those vaccinated during catch-up.[12] The most recent report stratified by age at vaccination for three doses only; higher effectiveness was found with younger age at vaccination.[30] There was no formal comparison of three vs. fewer doses effectiveness in any report from Scotland. For 4vHPV, two studies from the United States both found similar effectiveness for three, two and one doses. One used data from a national survey that obtained vaccination history from self-report.[27] The other used data from women aged 20–29 years continuously enrolled in an integrated health care system with vaccination histories from medical records.[28] The later study found high and similar effectiveness with three, two and one doses when analyses were limited to women who received the first dose at age ≤18 years. A study from Mongolia included women who were part of a pilot 4vHPV vaccination campaign[29]; the analysis included 118 women who received only one dose at age 11–15 years compared with 357 unvaccinated women, frequency-matched on age. The adjusted PR was 0.08 (95% CI 0.01, 0.56).

Two articles reported effectiveness in men aged 14–26 years, both from the same U.S. study at different time points; there was no statistically significant effectiveness for prevention of genital or anal vaccine-type HPV among those who received ≥ one dose (compared to no vaccination) and no difference in effectiveness by number of doses.[25, 26] The number of vaccinated men was small in both, and almost half had initiated sexual activity at the same age as or before vaccination.

In summary, among nine studies of effectiveness against HPV infection, three, all from Scotland, reported highest point estimates with three doses.[11, 12, 30] Two studies among women in the United States reported similar effectiveness regardless of number of doses.[27, 28] A study from Mongolia only reported single-dose effectiveness (92%),[29] and a study from the Netherlands only reported two-dose effectiveness (84%).[31] Two studies among men included a small number of vaccinated participants; no significant effectiveness was reported with ≥ one dose.[25, 26]

The original review included six studies of anogenital wart outcomes.[13–18] In the updated review, four additional studies were identified; one included men and women.[32–35] The 10 studies were from six countries that had introduced 4vHPV. In nine of ten studies, analyses were adjusted or stratified for age at vaccination; all 10 were able to adjust for markers of socioeconomic status, and several attempted to adjust for differences in sexual behavior using composite measures (Table 1).[32–34] Nine studies analyzed at least one of the vaccine dose groups compared with no vaccination; seven conducted a formal comparison of effectiveness between number of doses (Table 2).[13, 14, 16, 18, 32, 33, 35] Three included assessment of effectiveness using different buffer periods,[13, 16, 32] and five evaluated different intervals between doses in two-dose vaccine recipients.[14, 16, 18, 32, 33]

Figure 3 shows estimates from nine studies that provided vaccine effectiveness estimates against anogenital warts and buffer periods used. Of eight studies that formally evaluated effectiveness with three doses (compared to no vaccination), all found significant effectiveness. Six of seven found some significant effectiveness with two doses,[13, 15, 17, 32, 33, 35] and six of eight with one dose.[13, 14, 17, 32, 33, 35] Effectiveness estimates of one and two doses were generally higher when using buffer periods of longer duration. In the three studies that examined different buffer periods,[13, 16, 32] two found that a longer buffer period decreased differences by number of doses. Herweijer et al, used a three-month buffer period in the primary analysis.[13] In analyses adjusted for attained age and parental education, there was statistically significant effectiveness for three, adjusted IRR (aIRR)=0.20 (95% CI 0.17, 0.23); two, aIRR=0.32 (95% CI 0.26, 0.40); and one doses, aIRR=0.54 (95% CI 0.43, 0.68). In analyses stratified by age at vaccination (10–16 years and 17–19 years), with a buffer period > four months, there was no statistically significant difference between three and two doses, regardless of age at first vaccine dose. With a 12-month buffer period, there was no difference in effectiveness for three (aIRR=0.19 [95% CI 0.14, 0.24]), two (aIRR=0.19 [95% CI 0.13, 0.29]) or one (aIRR=0.24 [95% CI 0.15, 0.39]) doses among those who initiated vaccination at age 10–16 years. In a U.S. study, using data from electronic medical records and chart review, the adjusted HR of a single dose with a 6-month buffer from the last received dose was 0.81 (95% CI 0.60, 1.08) and with a 12-month buffer from the first dose was 0.32 (95% CI 0.20, 0.52).[32]

Seven studies that stratified by age at vaccination found higher vaccine effectiveness with younger compared to older age at vaccination, although differences were not formally tested in all.[13–15, 18, 33–35] Four studies evaluated vaccine effectiveness for three, two and one doses stratified by age at vaccination.[13, 33–35] Herweijer et al conducted the most detailed analysis, stratifying by age at vaccination (10–16 years and 17–19 years) and using different buffer periods.[13] They found higher effectiveness within the younger age at vaccination group, particularly for one-dose recipients when no or short buffer periods were used. Willows et al found higher effectiveness for all dose groups in those vaccinated at younger ages (9–18 years vs ≥19 years),[34] but differences by number of doses remained in all age at vaccination groups. In contrast, Zeybek et al found similar effectiveness by number of doses with younger age at vaccination.[33] A study by Navarro-Illana et al was limited mainly to girls vaccinated at age 14 years due to the national vaccination program in Spain; that study reported similar point estimates of effectiveness regardless of number of doses.[17]

In the five studies that explored the interval between doses in two-dose vaccine recipients,[14, 16, 18, 32, 33] three found that a longer interval changed effectiveness estimates or resulted in no difference between three and two doses.[14, 18, 32] For example, in a large study in Denmark where 70% of girls had an interval of two months between doses one and two, effectiveness was significantly higher with three doses than two doses in overall analyses.[14] However, there was no statistically significant difference between three and two doses with an interval > four months between doses and the IRR was close to one with an interval of six months.

In summary, among the ten studies evaluating 4vHPV effectiveness against anogenital warts, a range of age groups and different sensitivity analyses were included. Most studies reported highest effectiveness with three doses in the primary analyses. Sensitivity analyses suggested biases that could result from differences in age at vaccination across dose groups. Some found that differences between three and fewer doses decreased in analyses limited to persons vaccinated at younger ages, with longer buffer periods, or with longer intervals between two doses.

The original review included six studies that evaluated vaccine effectiveness for prevention of cervical cytological or histological abnormalities. In the updated review, ten additional studies were included (Tables 1 and 2).[36–45] Of the 16 studies, 12 were from countries that had introduced mainly 4vHPV, three 2vHPV, and one both. By number of doses, eight studies evaluated histological outcomes only, three cytological outcomes only,[22, 23, 45] and five both histological and cytological outcomes.[19–21, 41, 44] Histological abnormalities included cervical intraepithelial neoplasia (CIN) grade 1, 2, 3, CIN2+ (CIN grade 2 or higher or adenocarcinoma in situ [AIS]), and CIN3/AIS. Characteristics of women, including age at first vaccine dose, differed across dose groups in most studies. All studies except one evaluated cytological or histological outcomes irrespective of HPV type.[40]

Among the 16 studies, 15 found significant effectiveness for three doses. Figure 4 shows estimates from 13 studies that provide estimates of vaccine effectiveness against CIN2+. Either in the main analysis, in analyses restricted to certain age groups, or those using longer buffer periods, all 13 found significant effectiveness for prevention of high grade lesions (CIN2+ or CIN3+) with three doses; five of these 13 studies found significant effectiveness with two doses;[20, 38–41] and five with one dose.[20, 38–41] In general, analyses using longer buffer periods found higher effectiveness estimates and smaller differences by number of doses.

Eight studies used no buffer or buffer periods <12 months when evaluating prevention of high grade histological lesions, and five used an explicit buffer ≥12 months.[20, 21, 39–41] Four explored different buffer periods;[20, 21, 39, 40] of these, two found that longer buffer periods increased estimates of effectiveness for all dose groups.[21, 40] Brotherton et al found that point estimates for effectiveness of three and one doses were similar with a 12-month buffer but not with shorter buffers.[21] Another study using a 12-month buffer found that among those vaccinated at age 15–19 years, effectiveness was similar for three, two and one doses. Although Johnson et al found that effectiveness estimates increased with longer buffer periods, that study also formally compared three and one doses; a significant difference remained even with a 24-month buffer.[40]

Overall, 12 of 16 studies presented data stratified by age at vaccination or birth year group.[20–22, 36–42, 44, 45] Most found higher point estimates of vaccine effectiveness against cytological or histological outcomes with younger age at vaccination or later birth year, although the differences were not all formally tested. In eight studies that evaluated high grade histological lesions by number of doses stratified by age at vaccination, or in studies limited to women vaccinated at younger ages,[20, 21, 37–42] three found similar and significant effectiveness regardless of number of doses.[38, 39, 41] These three studies evaluated women vaccinated at age ≤16 years in Denmark, age <16 years in Australia and age 15–19 years in the United States. A study from Sweden and Demark also reported similar point estimates by number of doses among those vaccinated at age ≤16 years and 17–19 years, but only the the effectiveness estimate for three doses was significant among those vaccinated at ≤16 years.[37]

Most two-dose vaccine recipients received doses at a one- or two-month interval because they received vaccination under recommendations for a three-dose schedule. Four studies using histological outcomes examined intervals between two doses; three found no impact of interval on the estimate of effect,[21, 39, 41] and one found similar effectiveness with three and two doses with an interval ≥ five months but not < five months between doses in those vaccinated at age ≤16 years.[37]

In summary, we identified 16 studies evaluating vaccine effectiveness for prevention of cervical abnormalities. In many studies, baseline characteristics of women who received fewer than three doses were different than three-dose vaccine recipients, and investigators conducted stratified and/or adjusted analyses to control for these differences. All studies except one found effectiveness with three doses compared to no vaccination. While many of the earliest published studies reported highest efficacy with three doses and no significant effectiveness with fewer doses, more recently published studies have reported effectiveness with three, two and one doses, and three of five reported similar effectiveness by number of doses.[38, 39, 41] Limiting analyses to persons vaccinated at younger ages and using buffer periods, factors that impact several different domains in the quality assessments, resulted in higher effectiveness estimates.