INTRODUCTION
The use of combined vaccines or the simultaneous administration of multiple vaccines permits a number of desirable outcomes (1). Most importantly, the number of health care provider visits can be reduced while vaccine coverage can be increased. This permits better protection against more diseases earlier in life, and in the case of combined vaccines, reduced number of injections and the associated morbidity. These efficient and positive outcomes need to be carefully weighed, however, against the possibility that without careful evaluation, the sum may be less than the parts with respect to immunogenicity or efficacy. In terms of adverse reactions or safety, there is the opposite concern, that the sum may be greater than the parts, the combining of vaccines may increase the frequency of existing serious reactions or lead to new reactions. Sorting out which reaction is due to which vaccine may be quite difficult, however. Finally, with respect to economics, combining vaccines may increase the complexity of vaccine production for manufacturers, and may lead to the creation of a monopoly (2).

Before adopting a policy of simultaneous administration of multiple vaccines or licensing a combined vaccine, randomized clinical trials with careful follow-up of the patients are usually conducted (3). These trials are usually limited in sample size and duration of follow-up, and may be conducted in populations not fully representative of the general population, however.

The recommendation for simultaneous administration of measles-mumps-rubella (MMR) vaccine with diphtheria-tetanus-pertussis (DTP) and oral polio vaccines (OPV) at 15 months of age, for example, was based on a comparison of approximately 250 to 400 children in each group (4). The recent recommendation for the use of HbOC (haemophilus influenzae type b, Hib) conjugate vaccine during infancy was based on a larger study. Approximately 500 to 4000 infants in each arm of the trial (HbOC alone, HbOC + diphtheria-tetanus-pertussis [DTP], and DTP alone) were interviewed by telephone within 72 hours of vaccination for local and systemic reactions. About 30,000 vaccinees were also followed for emergency room visits, hospitalization, and sudden death (5). Even with this larger sample size, detection and evaluation of rare or delayed adverse events require surveillance postlicensure or postadoption of new recommendations. While such surveillance is important, methodologically, it is quite difficult to do well.

Classically, such postmarketing surveillance has been passive in nature, relying on spontaneous reporting system (SRS) of adverse events. These reports are frequently difficult to assess for attribution of causality to vaccine because such attribution requires either a specific laboratory finding, for example, the isolation of vaccine virus from a normally sterile site, as in isolation of mumps vaccine virus from the cerebrospinal fluid of patients with aseptic meningitis (6); a unique clinical syndrome, for example, acute flaccid paralysis in an OPV recipient in the absence of circulation of wild poliovirus; or an epidemiologic study that shows that the risk of the adverse event is greater among vaccinated than among the unvaccinated persons. SRS data are usually inadequate for epidemiologic assessment, however, for several reasons.

Any epidemiologic assessment requires data necessary to complete a two-by-two table of exposures and outcomes (table 1). In this case exposures are vaccinations, and outcomes are adverse events. However, SRS reports represent at best only one cell, cell "a", of such a table -- those who are vaccinated and have the outcome of interest. Usually, because of selective or biased reporting, SRS reports are only an unrepresentative sample of cell a. The remaining three cells of the two-by-two table, b, c, and d are usually not reported to the SRS. Finally, in a postlicensure setting allocation to vaccinated and nonvaccinated is no longer random; the data necessary to control for potential confounding is also usually not reported to SRS. Because most SRS reports do not contain laboratory or clinical findings that can be specifically attributed to vaccination, it is clear then that SRS are quite limited in assessing causality of reported adverse events.

Recognizing these weaknesses of the SRS, the Centers for Disease Control and Prevention (CDC) began in 1991 to organize a large-linked database study of about 500,000 children zero to 6 years of age in four HMOs (7). Information on all four cells of a two-by-two table plus potential confounders are being collected to permit controlled epidemiologic assessment of vaccine safety concerns, including those raised by this workshop.

Despite the shortcomings described, SRS can still provide a modicum of useful postlicensure data on the safety of simultaneous administration of vaccines. If prior to the introduction of the new procedure or new combined vaccine baseline SRS experience with at least one of the antigens is available and the reporting behavior to the SRS remains otherwise unchanged, a crude ecologic analysis can be done to see if there is any change in the profile of reported adverse events from the baseline. If no change is observed, this can be interpreted to be generally reassuring that there is no augmentation of adverse reactions with simultaneous or combined vaccinations. On the other hand, if a change is observed, a controlled study is still needed to validate this ecologic observation. A more refined ecologic analysis is possible if denominator data is available on how often and in which combination vaccines are administered. In which case, estimated rates of adverse events can be compared rather than numerator analysis alone.

To illustrate the utility of SRS for this purpose using the two types of ecologic analyses, we focused on the following case studies of simultaneous administration: 1) the recommendation to permit the simultaneous administration of MMR, DTP, and OPV vaccines at 15 month of age; (8) and 2) the recommendation for administration of various types of Hib vaccines, initially at 24 months, (9) then 18 months, (10) and more recently for infants; (11) these vaccinations are usually administered with DTP and OPV.

METHODS
The data sources used for this ecologic evaluation are the Monitoring System for Adverse Events Following Immunization (MSAEFI), operated by the CDC from 1979 to October 31, 1990 (12). Persons who received a publicly purchased vaccine also received an Important Information Statement, which encouraged vaccinees to report any adverse event which resulted in a health care visit within 30 days (4 weeks) of vaccination. The MSAEFI form was a basically "closed" report form with check-off boxes for specific adverse events. A number of improvements were made to MSAEFI in 1985, so the following analyses are based on MSAEFI data between 1985 and 1990.

On November 1, 1990, the Vaccine Adverse Event Reporting System (VAERS) became fully operational as a single unified system for the United States (13) . VAERS is jointly supervised by the CDC and the FDA. Reports are submitted using an "open-ended" report form. The reports are then assigned standard codes by using Coding Symbols for a Thesaurus of Adverse Reaction Terms (COSTART) (similar in principle to assignment of International Classification of Diseases or ICD codes). The VAERS analyses are based on data collected between November 1, 1990 and July 15, 1993.

Estimates of doses of vaccine administered were derived from two sources. The 1991 National Health Interview Survey, a cross-sectional interview of approximately 50,000 households and 135,000 persons annually based on a multistage area probability sampling design (14), and routine administrative data on doses administered submitted quarterly to the National Immunization Program from recipients of public sector immunization grants.

RESULTS
The age of vaccination on reports to MSAEFI in 1985 for children 12-23 months of age are shown on figure 1. As might be expected, reports of adverse events for which MMR was the only vaccine received peaked shortly after 15 months of age, and reports for which only DTP or DTP and OPV were administered peaked shortly after 18 months of age. There were few reports in which DTP, OPV, and MMR were administered simultaneously. In contrast by 1990, most of the reports in this age group were for DTP, OPV, and MMR vaccines administered simultaneously, and many fewer reports for the vaccines administered separately (figure 2). This change in schedule then permits us to examine for major changes in profiles of adverse event reports.

Examining a fairly specific adverse event first, febrile convulsions, table 2 shows the total number of reports to MSAEFI for children 15-23 months of age after different DTP, OPV, and MMR vaccine combinations, and the proportion of reports with febrile convulsions. We see that this proportion did not change significantly with the successive addition of OPV and MMR to DTP vaccine. While we do not know the exact denominator for each of these categories, we know that they are only the order of millions of persons. The results of this very crude ecologic analysis of postadoption surveillance data are in general agreement with preadoption clinical trials.

Serious neurologic adverse events are by themselves difficult to evaluate due to their rarity and the crudeness of SRS. But grouped together and performing a similar analysis, the data is generally reassuring again (table 2). The successive addition of OPV and MMR to DTP vaccines does not change the baseline proportion of reports which contain >1 neurologic event. In this instance, it actually suggests that simultaneous administration may be safer; 23% of reports with DTP-OPV-MMR vaccines reported at least one neurologic event compared to 30% of reports with DTP alone (p = 0.02). Due to the many potential biases and confounding in SRS data, however, one would need to validate this finding in a controlled study before reaching a definite conclusion.

Table 3 examines the overall safety profile more broadly in terms of DTP and MMR, administered either alone or in combination with other vaccines at 15-23 months of age in MSAEFI. Overall, we again see a general constancy in profile irrespective of simultaneous vaccination. The only exceptions are in settings like local reaction after MMR, which was more frequent with simultaneous administration (23%) than when given alone (8%) (p<0.01), but no higher than DTP given alone (48%).

Hib or Hib conjugate vaccination at age 18 to 36 months of age will now be considered. Because of the small numbers and the similarity in adverse event profile of Hib vaccine alone and Hib conjugate vaccine (HbCV) alone, these vaccines are grouped in these analyses. With accurate documentation of vaccinations and lot numbers that will be possible under the proposed National Vaccine Registry, product-specific evaluation using SRS data will be even easier. We see again that there is no significant increase in febrile convulsions as a percentage of total reports when these are given with successive combinations with DTP and OPV (table 4). Similarly for neurologic events reported after various DTP, OPV and Hib combinations, there are no significant changes in the proportion of neurologic events with the administration of additional antigens. Again with the exception of reports of more frequent local reactions with simultaneous administration, there were no major differences in the broader safety profile of Hib given alone or administered simultaneously with other vaccines (table 5).

When denominator data are available, a more refined type of ecologic analyses of SRS data is possible. In late 1990, Hib conjugate vaccine began to be administered to infants at 2, 4, and 6 months of age. From vaccine coverage data collected by the National Health Interview Survey, the percent of DTP vaccine simultaneously administered with Hib increased rapidly, from less than 10% in 1990 to 60% in 1991.

The switch from MSAEFI to VAERS in the public sector occurred on November 1, 1990. Because there were major differences in reporting forms between the two systems, comparing ecologic trends over time for most adverse events is difficult. However, hospitalization and deaths are two very specific outcomes that were captured by both MSAEFI and VAERS; therefore, each of these outcomes can be analyzed for changes over a time frame that spans both SRSs. For consistency, we examined only reports after vaccines received in the public sector. Because doses-administered data by specific vaccine are reported in the public sector also, we could examine the estimated rates of hospitalization and death reported per million doses of DTP vaccine administered. The relative stability of these rates in 1991 and 1992 despite the large increase in simultaneous administration of Hib vaccine suggests that this change in schedule did not result in significant increase in hospitalization or deaths(figure 3).

Any death reported after vaccination understandably raises concern. One might even ask why should there be any baseline rate of death reported after DTP or any other vaccine. Because vaccinations are recommended during the first year of life and sudden infant death syndrome (SIDS) also occurs during the first year of life, one would expect some SIDS to occur shortly after vaccination by chance alone. If the SRS is working, one would ,in fact, expect these deaths to be reported. The difficult question is whether one can determine which of these deaths are coincidental and which might be causally related to vaccination. Our FDA colleagues examine reports of deaths to VAERS for consistent patterns in clinical presentation; we have done some analyses for epidemiologic patterns.

Figure 4 shows the age distribution of all deaths among infants within 30 days after DTP vaccination reported to VAERS compared to the age distribution of 1) SIDS cases in the 1983 to 1987 birth cohorts from the national mortality statistics, and 2) DTP vaccinations from the National Health Interview Survey 1986-90. Drawing a line across the peak of the VAERS deaths matches the shape of the SIDS curve extremely well, but not the distribution of DTP vaccinations. The small hills correspond to when vaccinations are administered and when temporal clustering of reports are expected. Figure 5 shows the similar data for MSAEFI 1985 to 1990. With more years of data these hills smooth out, so that the two curves now coincide even more closely.

Figure 6 plots the seasonality of all infant deaths reported to VAERS within 30 days after DTP vaccination. This line appears similar to the shape of the seasonality of SIDS, and dissimilar from the seasonality of DTP administration as reported by the public sector. From the age distribution and seasonality figures, we conclude that most of the infant deaths after DTP vaccination reported to MSAEFI or VAERS are probably due to coincidence. A 1991 Institute of Medicine review of this issue arrived at a similar conclusion (15).

CONCLUSION
In summary, the benefits of simultaneous administration have generally been shown to outweigh the risks in prelicensure/adoption trials (1). Ecologic analyses of data from two postlicensure surveillance systems for adverse events were used to confirm that simultaneous administration of DTP, OPV, and MMR at 15 months of age and DTP, OPV, and Hib vaccines at various ages probably have not resulted in an increase of serious adverse events. If baseline SRS experience is available, similar ecologic analyses of SRS surveillance data can be used in the future to crudely assess the safety of new combinations of vaccine antigens. Due to the intrinsic methodologic deficiencies of SRS, however, controlled studies will be needed to confirm or refute signals generated by these relatively crude ecologic analyses of SRS data. CDC has embarked on a large linked database study to provide a ready setting with sufficient power for such validation studies (7).