Postexposure Effects of Vaccines on Infectious Diseases

Abstract

We searched the PubMed database for clinical trials and observational human studies about postexposure vaccination effects, targeting infections with approved vaccines and vaccines licensed outside the United States against dengue, hepatitis E, malaria, and tick-borne encephalitis. Studies of animal models, serologic testing, and pipeline vaccines were excluded. Eligible studies were evaluated by definition of exposure; attempts to distinguish pre- and postexposure effects were rated on a scale of 1 to 4. We screened 4,518 articles and ultimately identified for this review 14 clinical trials and 31 observational studies spanning 7 of the 28 vaccine-preventable diseases. For secondary attack rate, the following medians were found for postexposure vaccination effectiveness: hepatitis A, 85% (interquartile range (IQR), 28; n = 5 sources); hepatitis B, 85% (IQR, 22; n = 5 sources); measles, 83% (IQR, 21; n = 8 sources); varicella, 67% (IQR: 48; n = 9 sources); smallpox, 45% (IQR, 39; n = 4 sources); and mumps, 38% (IQR, 7; n = 2 sources). For case fatality proportions resulting from rabies and smallpox, the median vaccine postexposure efficacies were 100% (IQR, 0; n = 6 sources) and 63% (IQR, 50; n = 8 sources), respectively. Many available vaccines can modify or preclude disease if administered after exposure. This postexposure effectiveness could be important to consider during vaccine trials and while developing new vaccines.

immunization, immunotherapy, infectious disease, postexposure prophylaxis, vaccines

Abbreviations

 
• FDA

  Food and Drug Administration

 
• IQR

  interquartile range

INTRODUCTION

Since the advent of variolation in the early second millennium, vaccination has been considered a way to prevent healthy individuals from acquiring disease (1). However, to implement informed trials and curb future outbreaks and epidemics, postexposure effectiveness must be better understood. Substantial efforts have been made to develop therapeutic vaccines for treating chronic conditions such as cancer, diabetes, human immunodeficiency viruses, and obesity (2), but pre-exposure vaccination remains the focus for communicable disease. One notable exception is the rabies vaccine, which has achieved near total efficacy in exposed individuals for the last century. The smallpox vaccine also provided well-documented postexposure prophylaxis until the disease was eradicated 1980, with recommendations dating back to the mid-19th century (3). Nevertheless, uncertainty surrounding exposure status and length of incubation make postexposure properties difficult to estimate for any vaccine, and relatively few studies have done so for those that are currently available.

The effectiveness of postexposure vaccination varies widely depending on disease course, in terms of individual immune response and population-level spread. Although pre-exposure vaccination protects uninfected individuals from infection, postexposure vaccination modifies or prevents clinical disease among those who are already infected. As a result, postexposure trials must operate within a constrained time: participants must be identified and treated between exposure and symptom onset. Measurable benefits may occur if the vaccine stimulates an immune response faster or greater than that provoked by the natural infection alone. For smallpox, the vaccine induces antibody response 4–8 days before the variola virus does, probably because the response bypasses the initial respiratory-tract stages of a natural infection (4). These response kinetics explain historical evidence of a postexposure window and provide a basis for comparing surrogate models to humans in the case of reemergence (5). For rabies, though, this explanation has been unable to fully account for the vaccine’s postexposure mechanisms. Protection has generally been attributed to neutralizing antibodies, but rabies-exposed patients with human immunodeficiency viruses have been known to remain well despite poor or undetectable antibody levels after vaccination (6, 7).

At the community level, postexposure measures could be instrumental in reducing disease burden, especially when mass pre-exposure immunization is not feasible. Mathematical models for tuberculosis have been used to estimate that a postexposure vaccine would initially be more effective at reducing disease incidence compared with a preventive vaccine, although over time, a pre-exposure vaccine would result in increasing impact as more uninfected persons were vaccinated and protected against infection (8). One promising postexposure anti-tuberculosis candidate, the M72/AS01E vaccine, recently exhibited 54.0% efficacy against disease in a latently infected population (9). Although most diseases do not have such a large pool of hosts harboring latent infections—nearly a third of the global population carries tuberculosis—postexposure vaccination has already curtailed smaller outbreaks of varicella (10, 11), hepatitis A (12, 13), and measles (14, 15). These diseases tend to spread widely during outbreaks, but reactive vaccination studies usually operate under accelerated time frames and researchers have struggled to distinguish pre- and postexposure effects. One approach involves considering only the impact of vaccination on symptoms occurring before 1 incubation period after vaccination, which accounts for some of the uncertainty surrounding exposure status and timing. Variable incubation periods further complicate this method and make it especially difficult to draw conclusions about vaccination delays. As a result, the most robust information results from studies involving a known exposure during a definite interval before vaccination, as is the case for most percutaneous exposures or in settings that practice quarantine.

To address postexposure effectiveness across multiple diseases, we reviewed all infections that currently have a US Food and Drug Administration (FDA)-licensed vaccine, plus dengue, hepatitis E, malaria, and tick-borne encephalitis, for which vaccines are available in areas outside the United States (Table 1). The evidence could guide treatment protocols and also concerns the design and interpretation of more informative vaccine trials. We evaluated vaccines administered between exposure and symptom onset, but we also discuss the current state of research on therapeutic vaccines. These are a subset of postexposure vaccines designed to intervene after the onset of clinical disease, about which we include a brief summary illustrating the broader territory of nonpreventive, postexposure vaccination.

Table 1

 All Infections With a US Food and Drug Administration–Licensed Vaccine, Plus Dengue, Hepatitis E, Malaria, and Tick-Borne Encephalitis

Eligible Infection	Incubation Period, Days (Range) (83–85)a	Interval for Successful Postexposure Vaccination (Maximum)  b	Current ACIP Recommendations for Postexposure Prophylactic Vaccine Use
Cholera	0.08–5.00	No data	None
Pneumococcal disease	1–3	No data	None
Anthrax		No data	0, 2, 4 weeks
Cutaneous	1–7
Inhalation	7–60
Rotavirus	2	No data	None
Diphtheria	2–5 (1–10)	No data	Toxoid recommended
Haemophilus influenzae type b	2–7	No data	None
Meningococcal meningitis	3–4 (2–10)	No data	None
Yellow fever	3–6	No data	None
Denguec	5–7 (3–10)	No data	None
Influenza	5–7	No data	None
Japanese encephalitis	5–15	No data	None
Typhoid	6–30	No data	None
Pertussis	7–10 (4–21)	No data	None
Malariac	7–30	No data	None
Poliomyelitis	7–21	No data	None
Tick-borne encephalitisc	8 (4–28)	No data	None
Tetanus	10 (3–21)	No data	Toxoid recommended
Measles	11–12 (7–21)	≤3 days	≤3 days
Smallpox	10–14 (7–19)		N/A (case-by-case decisions)
For secondary attack		Insufficient data
For fatality		≤3 days
Chickenpox/varicella	14–16 (10–21)	≤3 days	≤5 days
Human papillomavirus	14–240	No data	None
Mumps	16–18 (12–25)	Insufficient data	Vaccine recommended, no timing provided
Rubella	17 (12–23)	No data	None
Hepatitis A	28 (15–50)	≤14 days	≤14 days
Rabies	30–90	≤5 days (+4 boosters)	0, 3, 7, 14 days
Hepatitis Ec	42 (14–63)	No data	None
Hepatitis B	120 (45–160)
For infants		≤1 month (+2 boosters)	Routine vaccine series
For adults		Insufficient data	Vaccine series
Tuberculosis	Weeks to years	No data	None

Abbreviations: ACIP, Advisory Committee on Immunization Practices; N/A, not applicable.

^a Infections are organized by minimum length of incubation according to the US Centers for Disease Control and Prevention.

^b The maximum interval is defined as the longest delay, with at least 3 studies suggesting an effectiveness of >75%. Studies with no evidence are indicated with ”no data.”

^c These infections have vaccines available in areas outside the United States.

METHODS

Strategy

We searched PubMed for clinical trials or observational studies pertaining to all FDA-licensed vaccines and their effectiveness after exposure. Search terms for each disease included but were not limited to “(disease) postexposure vaccine,” “(disease) postexposure vaccination,” “(disease) post-exposure vaccine,” “(disease) post-exposure vaccination,” “(disease) vaccine after exposure,” and “(disease) vaccination after exposure.”

Studies of animal models, serologic testing, or pipeline vaccines were excluded from this analysis. To ensure demonstration of a post- rather than pre-exposure effect, we limited consideration to secondary cases either confined to the first incubation period after vaccination or otherwise attributed to exposure before vaccination. Satisfactory evidence of such limitation included recording the time of vaccination and symptom onset within the vaccinated group or proving that exposure to the disease ended definitively before vaccination. In most studies, secondary attack rates were compared between vaccinated and unvaccinated groups, but for diseases with high rates of infection after exposure, such as rabies and smallpox, researchers sometimes looked to fatality proportions among secondary cases as a measure of vaccine effectiveness. We included both types of studies, along with reactive vaccination trials that met our susceptibility and exposure criteria. To determine the exposure ratings (defined in the following section), both authors conducted independent methodological assessments. Any discrepancies were resolved through discussion.

Definitions

Evidence of postexposure effectiveness was considered in terms of secondary attack and fatality outcomes. The effectiveness itself was defined as the relative reduction in outcome risk after having been exposed to a pathogen and subsequently vaccinated against it versus no vaccine or placebo. Because definitions of susceptibility and exposure vary, the exact descriptions have been compiled for each report in Web Appendix 1 and Web Tables 1–8 (available at https://academic.oup.com/aje). In most studies, susceptible individuals were characterized as those with negative history of vaccination, but this was confirmed with serologic testing in only a few studies. Especially in the case of smallpox, prior immunity was likely underestimated (4). An abridged summary of exposure certainty relative to vaccination is given in Table 2. Each study in our analysis was put in 1 of the following 4 exposure-rating classifications: 1) Exposure is uncertain or ongoing without a precise record of timing or timing is followed but without a control group; 2) study indicates a known point of exposure but authors offer no explanation for their approach; 3) point of exposure can be inferred with some confidence because timing has been followed closely and/or it is likely but not explicitly stated that index patients were isolated; and 4) there is a clear point where exposure occurs before vaccination.

Table 2

Summary of Reports Meeting Inclusion Criteria

First Author, Year (Reference No.)	Population	Study Type	Dates	Exposure Rating  a	Vaccine Type	Outcome for Vaccinated Group	Outcome for Unvaccinated Group	Time After Exposure	Effectiveness	95% CI
Chickenpox/Varicella
Asano, 1977 (16)	Children; household contacts	Cohort study	1977 (?)	1	Oka/Biken	0/18 SAR	19/19 SAR	≤3 days	100	57, 100
Asano, 1982 (17)	Children; household contacts	Prospective observational	1982 (?)	1	Oka/Biken (800–15,000 PFU)	0/30 SAR	None	≤3 days	100 ESTb
Arbeter, 1986 (18)	Children 18 months–16 years; household contacts	Double-blind RCT with placebo	1979–1982	1	Oka/Merck (4,350 PFU)	4/13 SAR	12/13 SAR	≤5 days	67	24, 85
Salzman, 1998 (19)	Children 14 months–12 years; Household contacts	Prospective observational	1995–1997	1	Oka/Merck	5/10 SAR	None	≤3 days	38 EST
Watson, 2000 (10)	Children <13; shelter contacts	Prospective, observational	1998	3	Oka/Merck (Varivax)	2/42 SAR	1/10 SAR	≤36 hours	52	−375, 95
Mor, 2004 (20)	Children 12 months–13 years; household contact	Double-blind RCT with placebo	1999–2000	1	Oka/RIT (Varilrix; GlaxoSmithKline, London, United Kingdom)	9/22 SAR	9/20 SAR	≤3 days	9	−83, 55
Brotons, 2010 (21)	Individuals >1 year old; household contacts	Prospective observational	2002–2007	1	Varilrix or Varivax	22/67 SAR	None	≤5 days	59 EST
Gétaz, 2010 (22)	Inmates in an overcrowded Swiss prison	Prospective observational	2009	4	Not specified	0/14 SAR	None	2–5 days	100 EST
Wu, 2018 (11)	Grade 8 students; school contact	Prospective observational	2013	1	VarV	0/10 SAR	4/27 SAR	Unclear; potentially ≤19 days	100	−483, 100
Hepatitis A
Werzberger, 1992 (12)	Children 2–16 years; community contacts	Double-blind RCT with placebo	1991	3	Formalin inactivated	7/519 SAR	12/518 SAR	Unclear; potentially ≤8 days	42	−47, 77
Sagliocca, 1999 (13)	Individuals 1–40 years; household contacts	RCT	1997	3	Havrix (GlaxoSmithKline, London, United Kingdom)	2/110 SAR	12/102 SAR	≤8 days	84	33, 96
Victor, 2007 (23)	Individuals 2–40 years; household and day-care contacts	Double-blind RCT with IG as control	2002–2005	4	Vaqta (Merck & Co.)	25/568 SAR	None	≤14 days	78 EST
Whelan, 2013 (24)	Any contact; household, sexual partner, cared for an infected child	Prospective observational	2004–2012	2	Not specified	8/167 SAR	None	≤14 days	76 EST
Parrón, 2017 (25)	Any community, household, daycare/school contact <40 years old	Retrospective cohort	2006–2012	1	Not specified	17/2,381 SAR	184/611 SAR	≤14 days	98	96, 99
Hepatitis B
Szmuness, 1980 (26)	Homosexual men known to be at high risk for HBV	Double-blind RCT with placebo	1978–1980	1	Plasma-derived (Merck & Co.)	1/549 SAR	6/534 SAR	Unclear	84	−34, 98
Beasley, 1983 (27)	Infants of e-antigen– positive/HbsAg carrier mothers	Double-blind RCT with placebo	1981–1982	4	Merck & Co.	1/51 SAR	74/84 SAR	3, 4, 9 months	98	84, 100
						3/50 SAR		4–7 days, 1, 6 months	93	80, 98
						5/58 SAR		1, 2, 7 months	90	77, 96
Roumeliotou-Karayannis, 1985 (28)	Spouses of patients with acute hepatitis B	Double-blind, RCT with placebo	1981–1982	3	H-B-Vax (Merck & Co.)	11/75 SAR	12/71 SAR	≤20 days	13	−84, 59
Ip, 1989 (29)	Infants born to mothers positive for HbsAg and HbeAg	RCT with placebo	1983–1986	4	Heat-inactivated (CLB)	11/64 SAR	33/47 SAR	0, 1, 2, 6 months	76	57, 86
Xu, 1995 (30)	Infants born to mothers positive for HbsAg and HbeAg	RCT with placebo	1982–1984	4	Plasma-derived (NIAID)	3/27 SAR	22/29 SAR	0, 1, 6 months	85	57, 95
					(BIVS)	11/28 SAR			48	14, 69
Measles
Berkovich, 1963 (14)	Children with tuberculosis; hospital contacts	Prospective observational	1961	3	Edmonston strain	3/6 SAR	None	2–5 days	44 EST
Watson, 1963 (31)	Household contacts	Prospective observational	1962	2	Not specified	0/3 SAR	2/2 SAR	≤3 days	100	−109, 100
Ruuskanen, 1978 (32)	Children aged 1–5 years; daycare contacts	Prospective observational	1975	2	Rimevax (GlaxoSmithKline, London, United Kingdom)	5/74 SAR	None	≤14 days	92 EST
Sheppeard, 2009 (33)	Household, social, hospital, school contacts	Retrospective cohort	2006	2	MMR	0/82 SAR	13/288 SAR	≤3 days	100	−125, 100
Barrabeig, 2011 (34)	Children aged 6–47 months; daycare contacts	Retrospective cohort	2006–2007	2	MMR	12/54 SAR	13/21 SAR	≤12 days	64	35, 80
Arciuolo, 2017 (15)	Household, community, and health care contacts	Prospective observational	2013	2	MMR	2/44 SAR	45/164 SAR	≤3 days	83	34, 96
Mumps
Wharton, 1988 (35)	School contacts during countywide outbreak	Prospective observational	1986	3	MMR	15/53 SAR	51/125 SAR	Unclear	31	−12, 57
Fiebelkorn, 2013 (36)	Household contacts	Prospective observational	2009–10	3	Not specified	1/44 SAR	8/195 SAR	3 days	45	−332, 93
Rabies
Bahmanyar, 1976 (37)	Individuals with potential rabies exposure	Prospective observational	1975–1976	4	PVRV	0/45 CFP	None	≤8 daysc (0, 3, 7, 14, 30 days)	100 EST
Anderson, 1980 (38)	Individuals with confirmed rabies exposure	Prospective observational	1978–1979	4	HDCV	0/21 CFP	None	≤10 daysc (0, 3, 7, 14, 28 days)	100 EST
Helmick, 1983 (39)	Individuals with confirmed rabies exposure	Prospective observational	1980–1981	4	HDCV	0/374 CFP	None	≤5 daysc (0, 3, 7, 14, 28 days)	100 EST
Wilde, 1995 (40)	Individuals with confirmed rabies exposure	Prospective observational	Not specified	4	HDCV	0/100 CFP	None	≤3 daysc (0, 3, 7, 14, 28 days)	100 EST
Wilde, 1996 (41)	Children who received unsuccessful postexposure prophylaxis after severe rabies exposure	Retrospective case study	1998–1993	4	PVRV or PCECV	5/5 CFP	None	≤66 hours (normal vaccine schedule; other deviations from WHO guidelines)	0 EST
Quiambao, 2000 (42)	Individuals with confirmed rabies exposure	Prospective trial	1996–1999	4	CPRV	0/57 CFP	None	≤5 daysc (0, 3, 7, 14, 28 days)	100 EST
Smallpox
Henderson, 2003 (3)	Secondary cases	Prospective observational (assumed)	1882	3	Not specified	5/26 CFP	None	During incubation period	N/Ad
McVail, 1902 (43)	Secondary cases after postexposure vaccination	Prospective observational	1900–1902	3	Not specified	7/101 SARe	None	≤3 days	N/A
						47/101 SARe		4–7 days	N/A
						41/101 SARe		8–11 days	N/A
Hanna, 2002 (44)	Patients who received postexposure vaccination	Prospective observational	1902–1911	3	Not specified	0/19 CFP	60/220 CFP	During incubation period	100	−50, 100
Sutherland, 1943 (45)	Secondary cases after postexposure vaccination	Retrospective cohort	1920–1942	3	Not specified	1/12 CFP	7/16 CFP	≤3 days	81	−35, 97
						5/12 CFP		4–9 days	5	−127, 60
Dixon, 1948 (46)	Household contacts	Retrospective cohort	1946	3	Not specified	0/21 CFP	34/132 CFP	≤5 days	100	−45, 100
						6/31 CFP		6–10 days	25	−63, 65
						1/4 CFP		≥10 days	3	−443, 83
Rao, 1972 (47)	Patients admitted to the Infectious Diseases Hospital, Madras, India	Retrospective cohort	1961–1972	1	Not specified	118/502 CFP	620/1,453 CFP	Unclear	45	35, 52
Douglas, 1962 (48)	Hospital contacts of a single smallpox case	Case studies	1962	4	Not specified	6/14 CFP	None	≤6 days	N/Ad
Heiner, 1971 (49)	Household and compound contacts	Prospective observational	1968–1970	1	Not specified	0/2 SAR	73/92 SAR	≤7 days	100	−248, 100
Mack, 1972 (50)	Household contacts	Retrospective cohort	1967	1	Not specified	12/16 SAR	26/27 SAR	≤10 days	59	−205, 94
Mack, Smallpox in Europe, 1972 (51)	Secondary cases among hospital staff, hospital clientele, and general public	Retrospective cohort	1950–1971	3	Not specified	20/70 CFP	41/79 CFP	Unclear	45	16, 64
Sommer, 1974 (52)	Secondary cases among household contacts	Prospective observational	1972	3	Lyophilized; WHO potency standards	14/1772 SAR	4/277 SAR	≤9 days	45	−65, 82
Mazumder, 1975 (53)	Patients admitted to the Infectious Diseases Hospital, Calcutta, India	Prospective observational	1973	3	Not specified	14/34 CFP	482/901 CFP	During incubation period	23	−16, 49

Abbreviations: BIVS, Beijing Institute of Vaccine and Serum; CI, confidence interval; CFP, case fatality proportion among cases; CLB, heat-inactivated hepatitis B vaccine; CPRV, chromatographically purified Vero-cell rabies vaccine; EST, estimated efficacy; HbeAg, hepatitis B e antigen; HbsAg, hepatitis B surface antigen; HBV, hepatitis B virus; HDCV, human diploid cell culture (rabies vaccine); MMR, measles-mumps-rubella; N/A, not applicable; NIAID, National Institute of Allergy and Infectious Diseases; PCECV, purified chick embryo cell vaccine; PVRV, purified Vero cell rabies vaccine; RCT, randomized controlled trial; SAR, secondary attack rate; VarV, varicella vaccine; WHO, World Health Organization.

^a The exposure ratings are classified as follows: 1) exposure is uncertain or ongoing without a close record of timing or timing is followed but without a control group; 2) study indicates a known point of exposure no explanation for the approach is offered; 3) point of exposure can be extrapolated because timing has been followed closely and/or it is likely but not explicitly stated that index patients were isolated; and 4) there is a clear point when exposure occurs before vaccination.

^b EST using accepted information about control group SAR/CFP (80% SAR for varicella, 20% for hepatitis A virus, 90% for measles, 100% for rabies, unknown for smallpox).

^c Point of initiation (subsequent schedule in parentheses).

^d Unable to determine a valid control to use for an efficacy estimation for evidence from Henderson et al. (3) and Douglas and Edgar (48).

^e Of individuals vaccinated within incubation period, this is the fraction of secondary attacks arising within given time frames. These data show that significantly fewer cases arise among individuals vaccinated within 0–3 days of exposure compared with ≥3 days.

The vaccines included target infections with an FDA-licensed vaccine, plus dengue, hepatitis E, malaria, and tick-borne encephalitis, for which vaccines are available in areas outside of the United States. These vaccines are referred to as “eligible vaccines” in this report (Table 1). To illustrate the findings from this review, a maximum delay was calculated for each eligible vaccine with evidence, defined as the maximum timespan between exposure and vaccination with at least a 75% effectiveness according to 3 or more studies. This delay was then compared with the incubation distribution for each disease according to the US Centers for Disease Control and Prevention (Figure 1). For the smallpox vaccine, which has been studied using both secondary attack and fatality proportions as indicators, the included data concerned fatality.

Figure 1

 Interval for postexposure vaccination versus incubation period for all eligible infections with evidence. The horizontal axis displays the incubation distribution according to the US Centers for Disease Control and Prevention (83, 85); the vertical axis shows the maximum delay between exposure and vaccination with at least a 75% effectiveness according to 3 or more studies in this review (at that delay or longer). This effectiveness concerns secondary attack rate for all vaccines except rabies and smallpox, for which fatality is used as the measured outcome. The 2 variables have a nonsignificant correlation (ρ = 0.76; P = 0.08). Hep, hepatitis.

Where possible, primary data were re-extracted and recorded in Table 2. We elected not to conduct a meta-analysis of results, because the studies, and, in particular, the inclusion criteria and timing of vaccination, were highly heterogeneous for most vaccines, and an average effect across such studies would be difficult to interpret. Because all point estimates for all vaccines and all studies considered were positive, but not all 95% confidence intervals excluded zero, we simply classified vaccines as those for which statistical evidence of postexposure protection had been observed in at least 1 study and those for which it had not.

RESULTS

Of the 4,518 sources identified, 45 ultimately met inclusion criteria (Figure 2). We reviewed the abstracts of 434 studies after preliminary exclusion by title; most were excluded because they measured a preventive vaccine effect (n = 330 of 360 clinical trials; n = 51 of 73 observational studies). After discarding postexposure studies focused on immunoglobulin or surrogate models and adding 15 studies found through bibliographies, a total of 14 clinical trials and 31 observational studies reported data for chickenpox/varicella (10, 11, 16–22), hepatitis A (12, 13, 23–25), hepatitis B (26–30), measles (14, 15, 31–34), mumps (35, 36), rabies (37–42), and smallpox (3, 43–53). No clinical or observational studies that fit our criteria were located for 21 of the 28 eligible vaccines, although tetanus and diphtheria toxoids are effective forms of postexposure prophylaxis (1).

Figure 2

Study selection process.

All vaccines but the mumps vaccine demonstrated statistical evidence of a positive postexposure effect in at least 1 study. Although the effects measured for smallpox were variable and the median was lower than for most other vaccines, reports in which effectiveness was stratified by delay after exposure showed a clear relation between prompt immunization and modified disease course. Vaccine effectiveness against secondary attack varied in strength across diseases, with medians of 85% (interquartile range (IQR), 28) for hepatitis A, 85% (IQR, 22) for hepatitis B, 83% (IQR, 21) for measles, 67% (IQR, 48) for varicella, 45% (IQR, 39) for smallpox, and 38% (IQR, 7) for mumps. These medians exclude the efficacies estimated using a historical control (Table 2) and incorporate only studies with an internal control population. For studies in which stratification was by vaccination delay (e.g., the Sutherland (45) and Dixon (46) smallpox studies), the included estimate refers to the shortest postexposure interval. Of the 46 studies, vaccine effectiveness against fatality was considered in 15—7 for rabies and 8 for smallpox—and median vaccine efficacies of 100% (IQR, 0) and 63% (IQR, 50) were determined for the 2 diseases, respectively.

Within these studies, vaccination timing varied widely and must be a factor when considering these results. According to data in this review, there was a nonsignificant correlation between the length of incubation and the length of delay between exposure and vaccination that allowed for at least 75% effectiveness (ρ = 0.76; P = 0.08), but it appears that a longer incubation period allows for a longer delay (Figure 1). Sample populations may also affect vaccine effectiveness: for instance, evidence for varicella and measles tends to be focused on children (7 of 9 and 4 of 6 studies, respectively), and that for hepatitis A tends to be from individuals younger than 40 years (n = 3 of 5 studies). Most data for these diseases, as well as for mumps and smallpox, derive from reactive vaccination campaigns, which often had large enough sample sizes to study. Still, their favorable characteristics sometimes came at the cost of exposure certainty, such as studies conducted in health care (15, 21, 47) and school (11, 25) settings in which researchers struggled to determine when exposure occurred. Exposure already is difficult to trace and isolate in outbreak settings, and many reactive attempts focused on disrupting an outbreak rather than investigating vaccination as postexposure prophylaxis to begin with. Conversely, percutaneous or mucosal exposure to viruses like hepatitis B and rabies is often much simpler to identify and link to infection.

DISCUSSION

Of the eligible vaccines with relevant postexposure evidence, all but mumps had compelling evidence of some form of postexposure protection. In previous reviews, authors have already investigated hepatitis A and B, smallpox, and varicella individually, and although this study is the first, to our knowledge, to incorporate findings on multiple diseases, our conclusions align with those of the targeted reports (54–58). Using historical smallpox data, Keckler et al. (54) and Henderson et al. (3). concluded that a 3–4 day interval suffices for significant postexposure protection, whereas a vaccine administered any time before symptom onset could be advantageous. Expert opinions culled using the Delphi technique corroborated this, estimating that vaccination is 93%, 90%, and 80% effective at preventing smallpox within 0–6 hours, 6–24 hours, and 1–3 days of exposure, respectively; and 80%, 80%, and 75% effective, respectively, at modifying disease among those in whom illness develops (4). In a review of varicella vaccines, researchers also determined a 3-day window but could not locate enough evidence to draw conclusions about vaccine effectiveness beyond 3 days (58). Perinatal hepatitis B vaccination was compared with placebo or no intervention in another review; authors found a relative risk of 0.28 (95% confidence interval: 0.20, 0.40), even with varying immunization schedules (56).

The findings here agree with those reported in previous reviews and indicate that even common vaccines have properties that have not been fully explored. Of course, postexposure trials are especially complicated because they involve a pool of infected individuals. The few studies that do exist were conducted in 2 main ways, largely depending on how easy it was to identify such a sample. The first approach was to record the time elapsed between vaccination and disease, then infer exposure timing and the window for effective vaccination on the basis of an incubation period estimate. This approach allows for less precise knowledge about exposure timing and thus lends itself to outbreak situations for diseases like hepatitis A, measles, mumps, varicella, and smallpox. However, authors of each study reconcile these uncertainties differently, so the study designs must be considered before comparing their results. For instance, reported events can vary in definition; for example, smallpox onset could refer to either fever (53) or rash (45, 49), which typically develops 2–3 days after the fever.

The second approach was more straightforward and involved restricting consideration to exposure that occurred in a distinct period before vaccine receipt. Examples include outbreak settings that enforced quarantine (22, 48), or diseases like hepatitis B and rabies for which transmission could be linked to specific events. Rabies is unique in that vaccines are typically administered after exposure; however, because current guidelines are successful, deviations longer than the typical 5-day delay have not been studied rigorously (59, 60). Most documented, postexposure prophylaxis failures tend not to have resulted from vaccine schedule changes but rather from insufficient wound infiltration or rabies immunoglobulin dose (41). Tetanus also spreads through cuts and punctures, although its established treatment protocol has been investigated even less systematically than that of rabies. The success of postexposure vaccination by tetanus toxoid has ruled out the option of randomized trials, and the success of preventive vaccination has reduced incidence of tetanus dramatically since the mid-1930s, eliminating most opportunity for case studies (1).

For both approaches, experimental designs create potential sources of variation across studies. Studies isolated exposure with varying levels of certainty, and although no significant pattern emerged between exposure rating and vaccine effectiveness, this rating scale indicates which of them are most likely capturing the desired effects. Other variables include vaccine types and doses; sample sizes, which range from 10 or fewer individuals (14, 19, 31, 41) to more than 2,000 (25, 52); prior immunity among participants, which was monitored in all but a few studies (25, 43, 48); and settings, whether in schools, hospitals, prisons, or households. More detailed information specific to each study and infection is reported in Web Tables 1–8 and Web Appendix 2.

In addition to these discrepancies, postexposure vaccination itself was assumed in most postexposure studies to not cause symptoms that could be mistaken for mild versions of the illness. If erroneous, this assumption could lead to an underestimation of vaccine effectiveness by mistaking vaccine-associated rash, for example, for breakthrough infection. In a study of healthy (presumably unexposed) children, 5.9% of measles, mumps, rubella, and varicella vaccine recipients and 1.9% of MMRII/VARIVAX (Merck & Co., Kenilworth, NJ) recipients experienced a very mild rash after vaccination, although both groups demonstrated greater than 90% response rates to the vaccine (61); therefore, in studies of postexposure vaccination for varicella and measles, minor adverse events may be conflated with secondary attacks and how often the vaccine works may be underestimated. For ethical and logistical reasons, some reports also lack information about important controls: individuals exposed to and untreated for rabies, for example, or individuals who received a postexposure smallpox vaccination and were protected from disease. This is the case for all but 3 historical smallpox reports (49, 50, 52); the rest include information only about people presenting with smallpox, such as the severity and outcome of their disease.

As for the vaccines without evidence of a postexposure effect, a likely explanation is that they do not have time to induce an adequate response before clinical disease manifests. The shortest incubation period among diseases with postexposure protection is 11–12 days for measles, whereas all but 4 of the 21 diseases without evidence have incubation periods of 10 days or less (Table 1). For diseases from which patients benefit from postexposure vaccination, a longer incubation period may permit a longer window for postexposure vaccination to be effective (Figure 1). Similar biological considerations already inform treatment guidelines when no trials are available. For instance, tetanus toxoid is recommended promptly after exposure, because too long a delay would allow additional tetanus neurotoxin to bind to neurons in the peripheral and central nervous systems (62). However, postexposure vaccination against tetanus produces an adequate amount of antitoxin in just 4–7 days, leaving a small window for a postexposure vaccine to outpace the natural, 10-day incubation (1). As discussed previously, the smallpox vaccine has a similar advantage over the natural infection and begins its course a few days ahead, spreading directly to regional lymph nodes and lymphoid organs (4). These results indicate that the success of postexposure prophylaxis depends on the time frames of the vaccine as it relates to disease mechanisms. It is important to note, though, that a postexposure vaccine could still supplement other treatments even if it cannot prevent or modify disease alone. Anthrax infections are 1 example and require antibiotics, because of a short incubation period and rapid onset. However, because anthrax spores can survive antibiotic prophylaxis, a vaccine should also be administered to counter long-lasting threats (1).

An exception to this generalization is the bacille Calmette-Guérin vaccine, which is the only licensed vaccine for tuberculosis and offers no known benefit to individuals with latent infection, despite that latent (asymptomatic) infection can last for years or decades (63). Better defense against tuberculosis ultimately will require a new vaccine and ideally 1 that functions before and after infection. More than a dozen candidates for priming and boosting have entered clinical trials to date, 6 of which are designed for individuals with latent or active infections—termed postexposure or therapeutic candidates, respectively (63). The M72/AS01E vaccine (GlaxoSmithKline, Brentford, United Kingdom), 1 of the most promising postexposure vaccines, recently exhibited 54.0% efficacy during a 2-year follow-up of a pool of healthy individuals infected with Mycobacterium tuberculosis (9). This finding, along with a recent bacille Calmette-Guérin revaccination trial in which 45% efficacy was reported within an uninfected population (64), has encouraged efforts toward new vaccine strategies for tuberculosis. Future solutions will have to reimagine how vaccines are implemented along the disease course, whether that involves a new way of using the bacille Calmette-Guérin vaccine or a new vaccine altogether.

In addition to these investigative tuberculosis vaccines, several others in the pipeline may have unexplored postexposure effects. Filoviruses incubate for a week or longer and thus might be promising candidates according to the findings of this review. In case reports, the recombinant vesicular stomatitis virus–Zaire Ebola virus vaccine has appeared to alter the course of Ebolavirus after several needlestick exposures in humans (65, 66), and in controlled studies, researchers have found a 50% postexposure efficacy in nonhuman primates after receiving a lethal challenge (6). Because the disease course is both faster than in humans and uniformly lethal in this nonhuman primate model, the protective effectiveness of the Ebolavirus vaccine could plausibly surpass 50% in humans (67). Preliminary research for Marburg virus indicated the recombinant vesicular stomatitis virus–Marburg virus vaccine had similar properties, demonstrating a significant effect if administered within 1 day of exposure in rhesus macaques (68). In light of the 2013–2016 Ebolavirus epidemic, filovirus countermeasures should remain a high priority, and systematic plans for gathering evidence should be set in place for the next outbreak.

An intermediate case not considered explicitly in this review is vaccination against herpes zoster disease. This disease, also called shingles, results from the reactivation of latent varicella-zoster virus infection primary, typically symptomatic, varicella disease (chickenpox). Two vaccines are licensed in the United States and have been proven effective in preventing herpes zoster (69, 70). Although technically a postexposure effect, we have not included herpes zoster vaccine in the main review because it is not an effect in preventing primary symptomatic disease but rather reactivation.

Classic vaccine efficacy trials are designed for study of the preventive effectiveness of vaccination against infection and disease. In most cases, if there were a postexposure effect of vaccination, it would have little effect on the outcome of such trials, because a small proportion of trial participants would be infected at the time of vaccination. Two classes of exceptions are worthy of note. First is vaccines against bacteria that colonize the respiratory or digestive tract asymptomatically and for which disease is a comparatively rare complication of colonization. For vaccines against such bacteria, the limited evidence available is that vaccination does not terminate the carriage state but rather reduces the acquisition of carriage and also the probability of disease, given carriage (71). If this is the case, then the postexposure effect should be modest for these vaccines and not complicate estimates of effectiveness. On the other hand, a setting where postexposure effectiveness could have greater consequences for the interpretation of vaccine trials is in ring-vaccination trials of vaccines against acute viral diseases, such as the Ebola Ça Suffit! (in English, “Ebola that’s enough!”) trial in which the recombinant vesicular stomatitis virus–Zaire Ebola virus vaccine was evaluated against Ebola virus disease in Guinea (72). In such a trial, by design, vaccination occurs close to the time of likely exposure to infection, and cases occurring within 1 incubation period of vaccination (plus 1 week for vaccine immunity to ramp up) were excluded from analysis as unpreventable by vaccination. On one hand, it is possible that postexposure effects, such as those observed in nonhuman primates with this vaccine, mean that some cases before that window could be preventable by vaccination, and it would be interesting to analyze the data from before the window in the main analysis to see if there is evidence of such an effect. On the other hand, because incubation periods are variable, it is possible that a window designed to exclude any individuals infected before vaccination would do so imperfectly, leading to some inclusion of a postexposure effect in the main effect estimate. If this design is used again, it would be valuable to quantify the likely impact of postexposure effectiveness on estimates obtained from the trial.

With the exception of tuberculosis, most postexposure benefits discussed so far arise as adverse effects of successful preventive vaccines. However, active therapeutic immunization has recently become a focus for chronic, mainly noncommunicable diseases. Many of these experimental vaccines induce antibody production, but the goal of their use is altering a disease that has already begun. Some vaccines treat drug abuse by binding to addictive substances like nicotine and cocaine (73), and others block tumor necrosis factor-α, an inflammatory cytokine linked to Crohn disease, rheumatoid arthritis, and psoriasis (74). Several active therapy candidates for Alzheimer disease have entered human clinical trials since 2000, mostly to target amyloid-β plaques that are thought to be causative agents (75). However, despite promising preclinical results, no significant cognitive benefit has been observed to date (76), to our knowledge. A similar narrative characterizes the efforts toward therapeutic HIV-1 vaccines: early optimism because of a slow and relatively well-understood disease progression, followed by decades of research and few positive clinical outcomes (77).

Other therapeutic vaccine candidates induce T cells rather than antibodies, an approach well suited to treating cancer. In theory, the paradigm mirrors that of classical viral vaccines in that tumor-associated antigens are used to activate T cells, which differentiate and proliferate to target the unwanted tumor cells (1). However, despite several promising phase III trials (78, 79), in an objective review of cancer vaccine trials, the authors found a response rate of less than 4% (80). The FDA has only approved 1 cancer vaccine, Sipuleucel-T (Dendreon Pharmaceuticals, Seal Beach, CA), which targets metastatic prostate cancer and prolongs median survival by 4.1 months (81). Still, the same phase III trial that led to its licensure presented no significant effect on other important disease factors, namely, time to disease progression, tumor regression, and long-term immunity. Interestingly, a different licensed vaccine may have the ability to eradicate other malignant tumors, specifically squamous cell carcinomas. In a recent case report, the 9-valent human papillomavirus vaccine (Gardisil-9; Merck & Co., Kenilworth, New Jersey) resolved all of a woman’s tumors within 1 year of her first injection (82). Although the human papillomavirus vaccine has already been linked to preventing cervical cancer and other diseases, its therapeutic properties remain unclear (1).

Therapeutic vaccination represents a promising frontier for disease treatment but also has implications for how we consider vaccines as a tool for prevention. As evidenced by this report, postexposure effectiveness has not been fully explored even for common vaccines, and there are several scenarios in which they are important, including responding to unpredictable health emergencies, designing new treatments, and interpreting vaccine trials. Especially in outbreak settings, attributing all outcomes to preventive effects could lead to an overestimation of a vaccine’s preventive effectiveness and underappreciation of its postexposure effectiveness. Therefore, postexposure effects not only concern innovative treatments for exposed or infected individuals but could also improve how we anticipate and understand the impact of any vaccine. This review represents the current body of evidence for vaccines that are already available, and our findings indicate there is still much to learn about postexposure vaccination.

ACKNOWLEDGMENTS

Author affiliations: Dartmouth College Department of Physics and Astronomy, Hanover, New Hampshire (Tara Gallagher); and Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts (Marc Lipsitch).

This project was supported by the National Institute of General Medical Sciences (grant U54GM088558).

We thank Dr. Mark Slifka for helpful discussions before this project began.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health.

Conflict of interest: M.L. discloses consulting or honoraria from Merck, Pfizer, Antigen Discovery, and Affinivax, and grant funding through his institution from Pfizer. T.E.G. reports no conflicts.