Examination of the Safety of Pediatric Vaccine Schedules in a Non-Human Primate Model: Assessments of Neurodevelopment, Learning, and Social Behavior

Background In the 1990s, the mercury-based preservative thimerosal was used in most pediatric vaccines. Although there are currently only two thimerosal-containing vaccines (TCVs) recommended for pediatric use, parental perceptions that vaccines pose safety concerns are affecting vaccination rates, particularly in light of the much expanded and more complex schedule in place today. Objectives The objective of this study was to examine the safety of pediatric vaccine schedules in a non-human primate model. Methods We administered vaccines to six groups of infant male rhesus macaques (n = 12–16/group) using a standardized thimerosal dose where appropriate. Study groups included the recommended 1990s Pediatric vaccine schedule, an accelerated 1990s Primate schedule with or without the measles–mumps–rubella (MMR) vaccine, the MMR vaccine only, and the expanded 2008 schedule. We administered saline injections to age-matched control animals (n = 16). Infant development was assessed from birth to 12 months of age by examining the acquisition of neonatal reflexes, the development of object concept permanence (OCP), computerized tests of discrimination learning, and infant social behavior. Data were analyzed using analysis of variance, multilevel modeling, and survival analyses, where appropriate. Results We observed no group differences in the acquisition of OCP. During discrimination learning, animals receiving TCVs had improved performance on reversal testing, although some of these same animals showed poorer performance in subsequent learning-set testing. Analysis of social and nonsocial behaviors identified few instances of negative behaviors across the entire infancy period. Although some group differences in specific behaviors were reported at 2 months of age, by 12 months all infants, irrespective of vaccination status, had developed the typical repertoire of macaque behaviors. Conclusions This comprehensive 5-year case–control study, which closely examined the effects of pediatric vaccines on early primate development, provided no consistent evidence of neurodevelopmental deficits or aberrant behavior in vaccinated animals. Citation Curtis B, Liberato N, Rulien M, Morrisroe K, Kenney C, Yutuc V, Ferrier C, Marti CN, Mandell D, Burbacher TM, Sackett GP, Hewitson L. 2015. Examination of the safety of pediatric vaccine schedules in a non-human primate model: assessments of neurodevelopment, learning, and social behavior. Environ Health Perspect 123:579–589; http://dx.doi.org/10.1289/ehp.1408257


Background
During the 1990s, thimerosal, an ethyl mercury (EtHg)-based preservative, was included in several vaccines given to U.S. infants (Clements et al. 2000). Many infants received up to 187.5 μg EtHg by 6 months of age by following the recommended pediatric vaccination schedule (Pichichero et al. 2008). This cumulative exposure exceeded the U.S. Environmental Protection Agency's safe intake level, estimated in 1997 to be ≤ 0.1 μg of mercury/kg body weight (BW)/day (U.S. Environmental Protection Agency 1997). However, these safety recommendations are based on data from exposure to oral methylmercury (MeHg), not intra muscular (IM) EtHg. Some parent and advocacy groups raised concerns over a possible link between the use of EtHg in vaccines and the increasing rates of develop mental dis orders, which has in turn negatively impacted immunization rates (Biroscak et al. 2003). In 1999, the Centers for Disease Control and Prevention (CDC) and the American Academy of Pediatrics recommended that thimerosal be removed from pediatric vaccines (CDC 1999).
Since that time, the Advisory Committee on Immunization Practices has markedly expanded pediatric vaccination recommenda tions (Fiore et al. 2008). By 2008, multiple doses of rotavirus, hepatitis A, pneumo coccal, varicella, and meningococcal vaccines, as well as a yearly influenza vaccine for all children 6 months to 18 years of age, had been added to the vaccine schedule. Despite the recommended removal of thimerosal from pediatric vaccines in the United States, multi dose influenza and meningo coccal vaccines still include thimero sal as a preservative (Food and Drug Administration 2012) and are administered to many infants and/or pregnant women (Dórea et al. 2013). Additional thimerosalcontaining vaccines (TCVs), such as that for hepatitis B, are also adminis tered to millions of children globally volume 123 | number 6 | June 2015 • Environmental Health Perspectives (Dórea et al. 2013). As the U.S. vaccine schedule has expanded, parental perceptions that vaccines pose safety concerns have grown (Gust et al. 2009;Kempe et al. 2011), espe cially since there have been no pre clinical studies examining the safety of new pediatric vaccine schedules in their entirety before universal recommendation.
Much of the research examining the safety of pediatric vaccines is based on rodent data. Specifically, these studies have investigated potential neuro behavioral effects of pre natal and/or post natal thimerosal exposure (Berman et al. 2008;Hornig et al. 2004;Laurente et al. 2007;Olczak et al. 2011;Sulkowski et al. 2012). At thimerosal doses equiva lent to those previously present in pediatric vaccines, few, if any, neuro behavioral effects were identified (Berman et al. 2008). When an adverse effect was reported, it was typi cally when very high doses of thimerosal (as much as 250 times that found in vaccines) were used (Li et al. 2014;Olczak et al. 2011) and/or the route of exposure differed (Li et al. 2014;Sulkowski et al. 2012). Several studies have already established that oral treatment and IM injections with thimerosal in mice result in different toxico kinetics (Harry et al. 2004;Rodrigues et al. 2010), indicating that the route of adminis tration is crucial in these studies. Furthermore, small improvements to experimental methodology, such as a reduc tion in injection volume (thereby avoiding possible hindlimb damage), resulted in a previously reported adverse neuro behavioral effect (Hornig et al. 2004) no longer being significant (Berman et al. 2008). Clearly, one must take into account the dose of thimero sal used, the route of adminis tration, and the injection volume when reviewing the litera ture to avoid misinterpretation of the findings. Ultimately, although the rodent literature has helped inform us about experimental design for thimerosal studies, the small size of mouse pups represents signifi cant challenges particu larly when adminis tering IM thimerosal (Harry et al. 2004).
With these limitations in mind, we developed a nonhuman primate model to examine the effects of different vaccine schedules on neuro behavioral development. Nonhuman primates (hereafter referred to as primates) share a great deal of evolutionary history with humans, and as such, are particu larly relevant for neuro behavioral and neuro cognitive evaluations. Questions addressing more complex cognitive processes and intricate social interactions may therefore be better suited for nonhuman primate studies (Nelson and Winslow 2009;Patten et al. 2014). Furthermore, primates are especially useful for studies of develop mental exposures because they, like humans, have relatively prolonged periods of gestation, infancy, and adolescence (Rice 1987). This long period of vulnerability allows investigation of critical variables during sensitive periods of exposure. Moreover, the nervous system of primates is quite comparable to that of humans (Nelson and Winslow 2009) and often responds simi larly to toxic insult (Burbacher and Grant 2000;Golub 1990;Rice 1987;Schneider et al. 2011). Because infant development in primates shares many parallels with that of humans, a wide range of neuro behavioral tests, adapted from assessments used with human infants, are routinely implemented for monitoring developmental trajectories in infant primates following exposure to envi ronmental neuro toxicants (Burbacher and Grant 2000;Gunderson et al. 1988;Rice 1999;Rice and Hayward 1997).
In summary, primates provide a relevant animal model for exploring potential neuro behavioral consequences of environ mental neuro toxicant exposures, such as thimerosal. In a controlled, blinded primate study, we examined the safety of pediatric vaccines, including TCVs, on a number of neuro behavioral tests: acquisition of neonatal reflexes, development of object perma nence, formation of discrimination learning strategies, and assessments of social behavior.

Materials and Methods
Animal assurances. Animal procedures followed the guidelines of the Animal Welfare Act and the Guide for Care and Use of Laboratory Animals (National Research Council 2011). The Washington National Primate Research Center (WaNPRC) and the University of Washington are fully accred ited by the Association for Assessment and Accreditation of Laboratory Animal Care. The experimental design and research proto cols were approved by the University of Washington Institutional Animal Care and Use Committee, and all animals were treated humanely and with regard for alleviation of suffering.
Animal husbandry. Rhesus macaque (Macaca mulatta) pregnancies were produced by natural mating at the California National Primate Research Center (CNPRC). We selected pregnant dams based on their overall health and confirmation of a male fetus of suitable gestational age by ultrasound. Prior pregnancy records were also reviewed to avoid nulli parous dams or dams with a history of miscarriage. Pregnant dams were transported from the CNPRC to the WaNPRC Infant Primate Research Laboratory (IPRL) by a specialized animal trucking company and monitored 24 hr/day using infra red cameras until delivery.
Study design. A total of 79 male infant macaques were studied in six groups (Table 1): a) control (animals received saline injections in place of vaccines); b) MMR (animals received only the MMR vaccine); c) TCV (animals received all TCVs but no MMR vaccines); d) 1990s Pediatric (animals received TCV and MMR vaccines following the pediatric schedule recommended in the 1990s); e) 1990s Primate (animals received all vaccines recommended in the 1990s but with the timing accelerated approximately 4:1); and f ) 2008 (animals received the expanded pediatric vaccine schedule that was in place in 2008, which remains very similar to the current recommended vaccine schedule).
We preassigned infants to a study group prior to delivery to distribute them across multiple study groups within a single breeding season (see Supplemental Material, Table S1). Within each study group, infants were further assigned to a peer group such that their birth dates were within 30 days of  Table S2, and added thimerosal. To calculate the thimerosal content for each vaccine, we first determined the amount of EtHg (micrograms) admin istered to a male human infant in the 10th percentile for weight at the recommended times of vaccination (Table 2). Using the weights of male infant macaques on the 95th percentile (Ruppenthal 1985), we calculated the weight ratio for male human infants:male primate infants at each scheduled vaccination. This maximized possible infant exposure to thimerosal while still maintaining an appro priate clinical exposure. An average weight ratio of 6.3:1 for human:primate infants across the entire study period was used to calculate the final dosing of the TCVs. Standardization of thimerosal content for each vaccine across the study also allowed for valid comparison of outcomes and minimized errors in vaccine dosing.
The preparation of TCVs and all quality assurance/quality control were performed at the University of Kentucky Environmental Research and Training Laboratory. Briefly, purchased vaccines were pooled prior to thimerosal addition. Stock thimerosal (T5125; SigmaAldrich) solutions were prepared such that a 50μL dose added to the pooled vaccines would yield the desired EtHg concen trations. Triplicate stock thimerosal solutions and spiked vaccine solutions were digested in 5% nitric acid at 100 o C for 2 hr and analyzed for EtHg concentration using a Varian Vista Pro CCD ICPOES (simultaneous induc tively coupled plasma optical emission spec trometer) to verify that target concentrations were achieved. Matrix effects were evaluated and corrected for using an yttrium internal standard. Furthermore, secondsource curve verifiers and spike recoveries were > 95%. Laboratory control samples consisting of three different dilutions of the stock solu tions bracketing the expected concentrations of the dosed vaccines were also prepared and analyzed alongside the dosed vaccines on a Nippon MA2000 mercury analyzer (Nippon Instruments Corporation). Recoveries on the laboratory control samples were again > 95%. The TCVs contained either 1.98 μg EtHg per 0.5 mL dose (Hep B) or 3.96 μg EtHg per 0.5 mL dose (DTaP and Hib). We peri odically verified the concentration of EtHg in vaccine aliquots throughout the study using an independent testing laboratory (Quicksilver Scientific).
For the 2008 schedule, additional vaccines were purchased from the manufacturers listed in Supplemental Material, Table S2. These included rotavirus, pneumo coccal, inactivated polio virus, varicella, hepatitis A, meningo coccal, and influenza vaccines, which were administered according to the schedule listed in Supplemental Material, Table S3. Because the multi dose vials of meningococcal and influenza vaccines currently available for pedi atric use contain 25 μg EtHg per 0.5 mL dose (Fiore et al. 2008), we purchased multiple singledose thimerosalfree vaccines and added thimerosal so that the influenza and meningococcal vaccines doses used contained 3.96 μg EtHg per 0.5 mL, as described above. In 2002, the CDC recommended that pregnant women be vaccinated against influ enza (Bridges et al. 2002). To replicate this, a single pre natal influenza vaccine containing 25 μg EtHg was adminis tered approximately 4 weeks before estimated delivery to all pregnant dams giving birth to infants assigned to the 2008 study group. All other dams received a single saline injection.
Vaccine administration. According to study group assignment, all animals received either a vaccine or saline injection, adminis tered IM, subcutaneously, or by oral gavage, depending on the manufacturer's recom mendations (see Supplemental Material, Table S2). For each IM injection, the needle was inserted at a 90degree angle and a 0.5 mL dose injected into the left or right biceps femoris of the hamstring. For subcu taneous injections, the skin of the thigh was pinched, the needle inserted at a 45degree angle, and a 0.5 mL dose administered. When multiple vaccines were to be administered at the same time, different sites within the same area were selected and/or the left and right side alternated.
To adjust the timing of vaccination to human age equivalents, we used a truncated schedule of vaccination. The development of the human and macaque infant visual system is very similar, with the post natal develop mental ratio between the two groups being about 4:1 (Atkinson 1979;Boothe et al. 1980;Teller et al. 1974). This 4:1 ratio is further demonstrated in the develop ment of pattern recognition (Gunderson and Sackett 1984) and the acquisition of object  (Ruppenthal 1985). c EtHg content of primate vaccines was determined by first averaging the weight ratios for human infant boys:male infant primates across the six time points of vaccine administration; this yielded an average weight ratio of 6.3:1. The EtHg content in each pediatric vaccine was then divided by 6.3 to determine the dosing of EtHg for each primate vaccine. This provided a similar dosing of μg EtHg/kg BW for infant boys and primates.
volume 123 | number 6 | June 2015 • Environmental Health Perspectives concept permanence (Williams 1979). Thus, the vaccinedosing schedule was adjusted to accommodate this projected 4:1 develop mental trajectory of infant primates. Implementation of neuro behavioral assessments. Assessments of infant development were based on protocols developed at the IPRL and have been extensively published (Burbacher et al. 2013;Chamove and Molinaro 1978;Harlow 1959;Piaget 1954;Sackett et al. 2006a;Schneider and Suomi 1992). All assessments were conducted by three trained testers (see Supplemental Material, Table S4) who were reliability tested to a minimum 85% agreement every 6-9 months, and who were blinded to the assignment of animals to study groups. Infants underwent developmentally appro priate assessments from birth to 12 months of age. Brief descriptions are given below (for detailed information, see Center on Human Development and Disability 2009). The timing of neuro behavioral assessments in relation to vaccine administration is shown in Figure 1.
Acquisition of neonatal reflexes. Infants were assessed for the presence of 19 neonatal reflexes based on the Neonatal Behavioral Assessment Scale (Brazelton 1978). Tests, performed daily from birth to 20 days of age, measured days to criterion for survival reflexes, basic motor reflexes, visual and auditory orienting, muscle tone, and behavioral state (Chamove and Molinaro 1978;Sackett et al. 2006a;Schneider and Suomi 1992).
Object concept permanence testing. The object concept permanence (OCP) physical search test consisted of four tasks: plain reach, screen, well, and AnotB (Sackett et al. 2006b). The object used as the reward consisted of a small toy covering a grape. The screen and well tasks had three conditions: no hide with the reward in plain view, partial hide with the object half covered, and full hide with the object fully hidden behind the screen or fully covered by a lid over the well. OCP was tested for each infant for 4 days/week from 14 days of age until the infant reached performance criteria on all tasks. Fifteen trials were presented in each session, and data were recorded as the number of sessions to criterion.
Discrimination/reversal learning and learning set. Discrimination/reversal testing was initiated at 75 days of age and imple mented using a touch screen computer program modeled after the Wisconsin General Testing Apparatus (Harlow and Bromer 1938). Computer testing proce dures followed those previously reported Sackett 2008, 2009). Infants were placed in a wire mesh cage with a touchscreen computer monitor mounted vertically to an opening of the cage. An initial adaptation procedure trained the infants to use the touch screen. Training was accomplished through successive approxi mation by rewarding the infants for approaching, touching, and finally activating the touch screen where a colored stimulus appeared. A stimulus appeared randomly in one of nine possible screen locations. The infant was considered trained when it correctly touched the screen only where the stimulus appeared on 23 of the 25 trials on a single day.
Discrimination and reversal testing imme diately followed the adaptation phase and consisted of 25 trials/day. Test trials were a maximum of 60 sec, and the inter trial interval was 10 sec (Mandell and Sackett 2009). Throughout testing, no correction proce dure was used. Two stimuli differing only in color were presented in random locations on the screen. A balk was recorded if there was no response within 60 sec after stimulus presentation, which is the accepted method for calculating non responsive trials. If the animal balked on 5 trials in a row, the session was terminated.
In the initial discrimination phase, the color of the rewarded stimulus was randomly chosen for each infant. The initial discrimi nation was run until the infant reached the criterion of 80% correct on a single day. After attaining criterion, the color of the rewarded stimulus image was reversed to the non rewarded color and 25 trials/day were run again to the same criterion. This was repeated for a total of four reversals. Six animals were removed from the analysis due to experi menter error (1990s Primate, n = 4; MMR, n = 1; and TCV, n = 1). These animals were moved to the first reversal on discrimination learning without reaching criterion. All of these animals were performing above 70% correct when this was done, but they had not yet met the required 80% correct to reach criterion.
Learning set presented the animal with a series of discrimination problems. Each problem had two unique stimulus images, with one randomly selected as the reward image. Each unique problem was presented to the infant for six trials, and then the images were changed to a new problem. Each infant was presented with 6 problems/day and received 240 problems over a minimum of 40 test days. If an infant balked for five trials in a row, that session was terminated. During the study, there was a modification to the software that affected the way the learning set was presented. The spatial distribution of the stimuli changed from three screen locations to nine, potentially increasing the difficulty of this test. Because the majority of animals (n = 54) started learningset testing after this software change, only these animals were included in the analyses (control, n = 8; TCV, n = 8; MMR, n = 12; 1990s Primate, n = 8; 1990s Pediatric, n = 12; and 2008, n = 8).
Although the software change did not affect the discrimination/reversal task, the same 54 animals were analyzed for both tasks so that the groups of animals were consistent.
Social behavior. Social behavior was evaluated in 40min daily playroom sessions for each peer group of four animals from approximately 30 days to 12 months of age. The playroom was approximately 2 m wide × 2 m deep × 1.5 m high and contained wire mesh shelves, climbing platforms, and toys. Scoring was conducted by a blinded observer in 5min focal periods using a coding system of mutually exclusive and exhaustive behav iors (Burbacher et al. 1990;Sackett et al. 1973). The order of testing was randomized for each session. Scored behaviors included passive, explore, withdraw, fear disturbance, rockhuddleselfclasp, stereotypy, play, sex, and aggression, and could be scored as either a social interaction or a non social behavior (see Supplemental Material, Table S5). Statistical analyses. Neonatal reflexes. The acquisition of neonatal reflexes was coded as the number of days from birth to reaching criterion for a putative reflex. Days to criterion was modeled using Cox regres sion for reflexes that had a single outcome (snout, suck, righting, or startle) and multi level Cox regression for all reflexes that were highly correlated (e.g., right and lefthand grasping). Cox proportional hazards regres sion models were fitted using the R survival package (Therneau and Grambsch 2000) with Breslow's method for tied time to events. Mplus software (Muthén and Muthén 2012) was used to fit multi level Cox regres sion models with a random intercept for animal, which accounts for the correlation in responses between observations from the same animal. In the event that criterion was not met, the days to criterion was truncated at 21 days and right censored. Condition was dummy coded so that the control group was the reference condition, and vaccine groups were each coded one if an animal participated in a putative condition or zero otherwise. The proportional hazards assumption was assessed for each reflex. The joint null hypothesis that all conditions had identical hazard functions was tested using a likelihood ratio test (LRT) that compared a null model with a model fitted with the experimental conditions, where a significant LRT indicates group differences; the null model for the multi level Cox model included a random intercept. In the event of a significant LRT, we examined individual parameters to assess whether differ ences represented differences between the control and a vaccine group. False discovery rate (FDR) corrections to pvalues were applied across LRTs and within each unique control versus vaccine group (e.g., control vs. TCV) to determine a significance cutoff (Benjamini and Hochberg 1995).
Object permanence. To analyze the development of object permanence, we used a Cox proportional hazards regression, fit in a manner identical to the method described above. In the event that the criterion was not met, days were truncated at 75 and right censored. Condition was dummy coded as described above for the reflex models. LRTs of the joint null hypothesis of identical hazard functions across conditions for object perma nence were performed as described above. FDR corrections to pvalues were applied in the manner as described above.
Discrimination learning. Data were initially summarized as the number of trials to attain 80% criterion on a single test day. We analyzed trials to criterion using survival analysis with Cox regression, and identified median trials to criterion for the control group. This median point was the 25trial interval, at which the probability of passing was 0.5 for the control group. The probability of passing at this trial interval was calculated for all the other groups, allowing for compar ison of the vaccine groups to the mid point of the survival curve for the control group. Groups with a higher probability of passing than the control group at this trial interval were quicker to attain criterion, whereas groups with a lower probability of passing were slower to attain criterion.
Learning set. Data were cleaned following published procedures . Briefly, trials on which the animal balked were removed. If the animal completed fewer than three trials in the problem, the entire problem was excluded from the analysis. All remaining trials and problems were rese quenced so that trial 1 in the analysis repre sents the first attempt at the problem and problem 1 represents the first problem where three or more trials were completed. The resequenced data were then aggregated across 40 problem blocks of the 240 total problems, creating a percentage of correct responses per trial on the problem block. Multilevel modeling was used to analyze the learning set data, which were fit using an auto regressive covariance structure to reflect the incremental increase in performance that is expected between trials and between problem blocks. Trial, problem block, and group were included as fixed factors, and the intercept was modeled as a random effect. Vaccine groups were compared with the performance control group using the coding procedure described above.
Social behavior. Prior to model building, we examined descriptive statis tics for duration and frequency of social and non social behaviors (see Supplemental Material, Table S6). Because duration and frequency were highly correlated, we used only duration as an outcome in the analytic models. Durations of the negative behaviors (withdrawal, fear/disturbance, rockhuddle selfclasp, and stereotypy) were summed for each animal, as were durations of the positive behaviors (play, sex, and aggression). Thus, for both social (involving one or more animals) and non social (involving no other animal) behaviors, four behavior outcomes were used in the analysis: passive, explore, negative, and positive. A 30day average was computed for the duration of each of the four non social and social behaviors for each animal for each 30day period from 30 days to 360 days of age. Duration values were natural logtransformed to reduce the possi bility of disproportionate impact of extreme values. Models were fit following longi tudinal modelbuilding strategies in which the unconditional growth model (i.e., the average rate of change in a putative outcome) was established by comparing longitudinal models using the Akaike information crite rion. Nochange, linear, and quadratic models were fit for each outcome. Time was centered at month 2, the first month of the data. The assessment of unconditional growth models indicated that a quadratic model (i.e., change was non linear) was the best model for all outcomes, except for a linear trend for social positive behavior. After establishing the growth model for each outcome, we added the intervention condition and an inter action between time parameter and the inter vention condition to the models to test for differences in experimental conditions and for differences in develop mental trajectory of a putative behavior as a function of experimental condition, respectively. An FDR correction was applied to each parameter across the eight models. In the event of either a significant effect for group or a group × time inter action, we estimated simple slope comparisons (Bauer and Curran 2005) between the control group and each of the vaccine groups. The differences were computed at 2 months and 12 months of age to assess any differences between the experimental groups and the control group at the beginning and at the end of the study period, using an FDR within each timepoint.
Object concept permanence. Sessions to criterion for the four stages of object permanence testing are shown in Table 4. No significant differences between groups were observed.
An error analysis was conducted to assess differences in perseverative behavior between groups. Perseveration was defined as any day of testing that an animal performed < 34% correct or balked on the session, if the balk day was preceded by a persevera tive day. All other test days were classified as nonperseveration. Classifying test days in this way has been shown to be sensitive to prefrontal lesions (Jones and Mishkin 1972), as well as to development in humans (Overman et al. 1996) and primates (Mandell and Ward 2011) of a compa rable age. A oneway analysis of variance revealed no significant differences between groups for perseverative behavior and balks for any discrimination or reversal phase (see Supplemental Material, Table S7).
Learning set. The key outcome in a successful learningset analysis is a significant 2way inter action between Block and Trial that shows better performance on trials 2-6 as the animal progresses through testing. Overall, there was not a significant Block × Trial interaction [ Table 6; F(35, 1606.7) = 0.8, p = 0.79], nor was there a significant main effect for Group [F(5, 543.1) = 2.03, p = 0.07]. Percent correct for the Block × Trial interaction for each group revealed a similar pattern to the overall Block × Trial interac tion with no evidence for learningset formation and only modest withinproblem learning by trials 5 and 6 in the later blocks (see Supplemental Material, Figures S2,S3). Although there was a significant threeway interaction (Table 6), the lack of evidence for learningset formation with any of the groups, as well as the lack of a clear pattern of differ ences in contrast testing, suggests that this result does not reflect an interpretable learning difference between the groups. Finally, overall latency for the Block × Trial inter action was highest on trial 1 and remained high on subsequent blocks (see Supplemental Material, Figure S4). When we examined the Block × Trial interaction for each study group, we found that all groups had the same general pattern of high reaction times on trial 1(see Supplemental Material, Figure S5). The TCV group had the slowest overall reaction times and was significantly slower than the control group (mean difference, 1.83; 95% CI: 0.96, 2.69). The 2008 group also had reaction times significantly slower than the control group (mean difference, 0.91; 95% CI: 0.12, 1.70).
Social behavior. Overall means and SDs for duration and frequency of social and non social behaviors scored for all infants is shown in Supplemental Material, Table S6. The duration and frequency of negative   behaviors by animals in all groups was very low; in fact, there were no instances of stereotypies recorded across all sessions (see Supplemental Material, Table 6). Analyses of social interaction data identified a significant Group × Quadratic interaction [F(5, 752) = 2.92, p = 0.030] for negative behaviors, indicating that longitudinal change in negative behaviors differed across groups. Followup contrasts indicated that at 2 months of age, relative to the controls, animals in the 1990s Primate and 2008 groups exhib ited significantly fewer negative behaviors [t(752) = -2.47, p = 0.034 and t(752) = -2.85, p = 0.023], respectively (Figure 2; see also Supplemental Material, Table S8). At 12 months of age, there were no significant differences in behaviors in the vaccine groups compared with the control group. Analyses of non social interaction data revealed a significant Group main effect [F(5, 211) = 3.62, p = 0.011] for passive behaviors. However, animals in the control group exhibited no significant differences in passive behaviors from the experimental groups at both 2 months and 12 months. There was a significant Group × Quadratic interaction [F(5, 751) = 3.32, p = 0.021] for explore behaviors. Followup contrasts indicated that at 12 months of age, relative to the controls, the 1990s Pediatric group exhibited significantly fewer explore behaviors [t(751) = -4.62, p < 0.001] (Figure 3; see also Supplemental Material, Table S9). There was also a significant Group × Quadratic interaction [F(5, 751) = 3.68, p = 0.021] for negative behaviors. Followup contrasts indicated that at 2 months of age, relative to the control group, the 1990s Primate and MMR groups exhibited significantly fewer negative behaviors [t(751) = -4.12, p < 0.001] and [t(751) = 2.35, p = 0.048], respectively. We observed no significant differences in negative behaviors in the vaccine groups relative to the control group at 12 months. There was a significant Group × Linear time interaction [F(5, 751) = 13.97, p < 0.001] for positive behaviors. Followup contrasts indi cated that, relative to the control group, the 1990s Pediatric group exhibited significantly fewer positive behaviors [t(751) = -2.95, p < 0.016] at 2 months, and significantly greater positive behaviors at 12 months [t(751) = 4.75, p < 0.001] (see Supplemental Material, Table S9).

Discussion
In this primate study of vaccine safety, we examined a number of neuro behavioral tests-the acquisition of neonatal reflexes, the development of object permanence, the forma tion of discrimination learning strategies, and assess ments of social behavior-in a primate model of vaccine safety. Using a modified version of the Neonatal Behavioral Assessment Scale, we found that days to criterion for the acquisition of neonatal reflexes was similar for animals irrespective of vaccination status, suggesting that auditory and motor function at this age were normal. The only exception was for the acquisition of the hand top of counter reflex for the 1990s Pediatric group, which took longer than the control group. These data are in contrast to our previous pilot study in which a delay in the acquisition of the root, suck, and snout survival reflexes were reported for primate infants following exposure to the birth dose of the thimerosal containing Hep B vaccine (Hewitson et al. 2010a). This discrepancy is most likely due to the larger number of animals in the present study providing more accurate estimates. Furthermore, in the present study, reflexes were examined from birth to 21 days of age, during which some animals received multiple TCVs (not just a single Hep B vaccine, as was used in the previous study), and yet no detri mental effects on the acquisition of survival reflexes were reported for these animals.
Several rodent studies have examined the effect of thimerosal on auditory and motor function (Berman et al. 2008;Hornig et al. 2004;Olczak et al. 2011;Sulkowski et al. 2012). For example, lowdose thimerosal exposure was found to decrease motor function and increase anxiety in SJL mice, which are susceptible to auto immunity, but not in C57BL/6J or Balb/c mice (Hornig et al. 2004), suggesting that an altered immune system might confer heightened susceptibility to thimerosal in mice. However, SJL mice are functionally blind as early as 4 weeks of age as a result of retinal degeneration (Chang et al. 2002), and they demonstrate poorer perfor mance in tasks that rely heavily on the visual system (Wong and Brown 2006); therefore,   Hornig et al. (2004), is questionable. The timing, dosing, and location of thimerosal injections in rodent studies can also have a significant effect on data outcome. The small size of mouse pups and the limited muscle development at times of IM dosing would have resulted in injections that were a combi nation of IM and subcutaneous routes (Harry et al. 2004), and any vascular involvement or damage to the hindlimb would have negative implications for tests of motor function. In a study similar to that of Hornig et al. (2004), Berman et al. (2008) examined a number of neuro behavioral outcomes in SJL mice following vaccination with low dose thimer osal. They specifi cally lowered the vaccine injection volumes and verified at 2-3 days post injection that there was no vascular damage at the site of injection. In that study, no deficits in tests of social interaction, sensory gating, or anxiety were reported. Although Berman et al. (2008) did report a significant locomotor effect, it was limited to female mice in the open field test only at 4 weeks of age, an age when visual acuity may be diminished (Wong and Brown 2006). Other studies have reported a delay in development of the startle reflex and motor learning (Sulkowski et al. 2012) or a decrease in social behavior (Olczak et al. 2011) in rat pups receiving either subcutaneous or IM injections of thimerosal, respectively. These effects were found only at doses of 200-3,000 μg EtHg/kg/BW, which is 15-500 times the level of EtHg found in pedi atric vaccines. Such high doses do not allow for sufficient clearing of EtHg, which has been shown to persist in the rat brain for > 30 days following a single acute IM injection of thimerosal (Olczak et al. 2009). Because much of the rodent data reflects different method ologies and timing and dosing of thimerosal, with adverse effects being found only at very high doses, it is difficult to directly correlate these findings with results of our study.
In the present study, we also examined OCP, discrimination/reversal, learning set, and social behavior. Attainment of object permanence requires some understanding that objects are permanent in space and time and continue to exist when removed from the visual field (Piaget 1954), and has been closely linked to early memory development (Diamond 1990). We found no statistically significant differences between vaccinated and control animals on performance in any phase of the OCP testing. Several primate studies have shown that OCP testing is sensi tive to various highrisk conditions, such as prenatal exposure to MeHg, prematurity, low birth weight, and birth asphyxia (Burbacher et al. 1986(Burbacher et al. , 2013. Twochoice color discrimination tests have been used to evaluate basic learning skills in infant primates for many years (Harlow 1959). Mastery of this task requires the animal to learn a simple discrimination between two identical objects that differ in color. In the present study, we found no significant differ ences in performance in the discrimination phase across all groups. However, there were two consistent group differences during the reversal phases: Animals in both the TCV and 1990s Primate groups achieved criterion in fewer trials than control animals in three of the four reversal phases, although not the same three reversals. Animals in both groups received similar dosing and timing of TCVs; thus, it appears that animals receiving TCVs on the accelerated schedule demonstrated improved performance during reversal testing. In agreement with this finding, previous studies in macaques have shown that both pre natal and post natal exposure to MeHg resulted in facilitated learning on this task, as well as a spatial alternation task (Gilbert et al. 1993;Rice 1992). Conversely, animals in the 2008 group, which had a higher cumu lative exposure to thimerosal at the time of testing due to both pre natal and post natal vaccinations, showed no evidence of facilitated learning in any phase of reversal testing.
Several clinical studies have examined the relationship between infant thimerosal exposure from TCVs and pediatric outcome. For example, in a British cohort study examining child development and behavior, Heron et al. (2004) reported that exposure to thimerosal at 3 months of age was inversely associated with hyperactivity and conduct problems, motor development, and require ment for speech therapy. More recently, several studies have reported on the effects of exposure to TCVs and subsequent tests of memory and learning, attention, execu tive function, language, and motor skills in children at 7-10 years of age (Barile et al. 2012;MrozekBudzyn et al. 2012;Thompson et al. 2007;Tozzi et al. 2009). In the original CDC study, Thompson et al. (2007) iden tified a few significant associations with exposure to thimerosal, but these were small and divided equally between both positive and negative effects. For example, among boys, there was a beneficial association between thimerosal exposure and performance IQ but a detrimental association with both behavioral regulation and motor tics. This analysis was  Positive Negative then expanded using measurement models to further assess any associations between thimerosal exposure and neuro psychological outcomes. In the subsequent analysis (Barile et al. 2012), the only consistent finding was an association between early thimerosal exposure and the presence of motor tics in boys. In an Italian cohort, Tozzi et al. (2009) found that greater thimerosal exposure was associated with lower scores in motor function (finger tapping test) and language (Boston Naming test) only in girls. On the basis of the overall study outcomes, Thompson et al. (2007) and Tozzi et al. (2009) concluded that the pattern of results was consistent with these associa tions occurring by chance and that exposure had no relation to outcome.
Learningset formation refers to the learning of visual and other types of discrimi nation problems progressively more quickly as a function of training on a series of problems (Schrier 1984). In the present study, animals in the TCV group demonstrated increased response latencies in learningset testing compared with the control group but this was not found in animals in the 1990s Primate group, which received the same EtHg exposure. Furthermore, the TCV group showed little evidence that they had performed at a level, or that their responses had organized into a strategy, that was different from that of controls. In fact, the only performance differ ence was in the overall mean averaged across all of the blocks and trials, not in their learning across trials or blocks, which is the outcome needed to indicate a learning or strategy differ ence. In fact, the reported difference was found only in the overall mean averaged across all of the blocks and trials, not in their learning across trials or blocks, which is the outcome needed to indicate a strategy difference.
It is well established that primates who are at high risk for poor developmental outcomes may not develop normal social behaviors charac teristic for that species. For example, Burbacher et al. (1990) reported that chronic pre natal exposure to 50 μg/kg/day oral MeHg altered the expression of social behavior in primates, such that exposed infants spent more time being passive and less time engaged in play behaviors with peers. Post natal exposure to lead (Bushnell and Bowman 1979;Levin et al. 1988) or pre natal exposure to TCDD (2,3,7,8tetrachlorodibenzop dioxin) (Bowman et al. 1989) have also been shown to negatively influence social behavior in macaques. Early differences such as these may translate into enduring social deficits that impact the animal's ability to interact effec tively with other animals into adulthood. In the present study, TCVs did not affect the develop ment of social behaviors charac teristic of infant macaques of this age. In all study groups, we observed that each of the four social and non social behaviors developed as expected for normal laboratoryreared macaque infants (Worlein and Sackett 1997). It is particu larly relevant that, under the hypothesis that TCVs may impact behavior, there were very few instances of negative behaviors, such as rocking, selfclasping, and stereotypy, reported across the entire infancy period for all groups. This is reassuring because infants would have received the full schedule of TCVs during behavioral testing, representing the period of development at highest risk for neuro toxicity.
Based on the observed toxico kinetics in infant primates receiving lowdose IM thimerosal injections (Burbacher et al. 2005), toxicity following TCV administration would appear unlikely. For example, the halflife of Hg in the blood is 7 days in primates (Burbacher et al. 2005), which is similar to data from comparable studies in mouse pups (Zareba et al. 2007) and human infants (Pichichero et al. 2002(Pichichero et al. , 2008. Furthermore, there is minimal accumulation of Hg in the blood after adminis tration of multiple TCVs (Burbacher et al. 2005;Pichichero et al. 2008), suggesting that Hg is rapidly metabo lized and either excreted or deposited in tissue. In primates, the halflife of Hg in the brain following thimerosal exposure is 24 days, more than three times that seen in blood (Burbacher et al. 2005). Accumulation of Hg in the brain of primate infants is therefore likely to occur over time with repeated admin istration of IM thimerosal (Burbacher et al. 2005), although there is no clear evidence in the literature that this accumulation would directly impact neuro behavioral outcome.
Our study has several limitations. First, studies of lowdose thimerosal exposure in primates have employed an accelerated schedule of exposure similar to that used in rodent studies (Burbacher et al. 2005;Hewitson et al. 2010b). This schedule is based on the theoretical developmental ratio of 4:1, that is, 4 weeks of human development is comparable to 1 week for a primate (Boothe et al. 1985). In the present study we examined neuro behavioral effects of TCVs using both an accelerated vaccine primate schedule and the recom mended pediatric schedule, neither of which appeared to affect neuro behavioral outcomes, thus suggesting that the toxico kinetics of EtHg in infant primates is not a limiting factor when using an accelerated schedule of dosing.
Second, we used only male animals in our study, and many clinical studies have reported genderspecific effects of organo mercurials (reviewed by Llop et al. 2013). For example, higher exposure to EtHg through vaccination in boys was associated with poorer behavioral regulation and a higher likelihood of motor tics, whereas girls performed significantly better in tests of visual-motor coordination when tested at 7-10 years of age (Thompson et al. 2007). Conversely, pre natal and post natal exposure to dietary MeHg has been reported to have a negative effect on visuo spatial testing at 9 years of age, but only in girls (Davidson et al. 2008).
Finally, because of the large sample size in our study, infants were added to the protocol over several breeding seasons spanning 5 years. There is always a possibility of changes in environmental conditions over time, which is a challenging variable to control for, and therefore a potential limitation to this study. Every care was taken to ensure that all testers remained blinded to study group assign ment and that they were reliability trained to the highest standard. Furthermore, neuro behavioral assessments followed very detailed protocols that have been used at this facility for more than three decades (Burbacher and Grant 2012;Burbacher et al. 2013).

Conclusions
We found no evidence of an adverse impact of vaccination status on early neuro developmental measures, including the acquisition of neonatal reflexes and the develop ment of object perma nence. This was true for animals receiving TCVs, as well as animals in the 2008 group, which received the expanded pediatric vaccine schedule that is very similar to the currently recommended schedule. Although some animals that received TCVs performed better than controls in the reversal phase of discrimi nation learning, this association was not consis tent across all study groups with thimerosal exposure. Furthermore, response latency on learningset testing was slowest for animals in the TCV group, but this observation was not mirrored in the 1990s Primate group, which received the same EtHg exposure. Finally, all infants, irrespective of vaccine status, devel oped the typical social behaviors for their age, with very few instances of negative behav iors reported. Although the data as a whole do not support a consistent adverse effect of TCVs on primate development, factors that may modulate the toxico kinetics and toxico dynamics of thimerosal-such as genetics, sex, birth weight, gestational age, maternal health, and chemical coexposures-should be thoroughly investigated.