Immunization Issues

Vaccine Safety Datalink

Updated: February 12, 2009

Vaccine safety monitoring is conducted both before and after a vaccine is licensed. Prelicensure trials typically involve thousands to tens of thousands of participants. Common adverse events can be identified by comparing the incidence of those events in vaccine recipients with the incidence rate in placebo recipients. A rate significantly higher in the vaccine recipients is supportive of vaccine playing a causal role in the illness.

However, adverse events that are rare (for example, 1 in 10,000 or lower) may not be detected because the numbers of participants in the prelicensure trials are not large enough to find these rare events.1 For example, to identify a rare adverse event that occurs 1 time per every 10,000 vaccinees may require 30,000 vaccinees plus 30,000 controls or more.2

Postlicensure or postmarketing evaluation of vaccine safety is done through passive reporting systems such as the Vaccine Adverse Event Reporting System (VAERS) or non passive systems such as the Clinical Immunization Safety Assessment (CISA), as well as large linked databases, such as the Vaccine Safety Datalink (VSD) project.

The VSD project was created in 1990 by the Centers for Disease Control and Prevention (CDC) and currently involves partnerships with 8 large health maintenance organizations (HMOs) to continually monitor vaccine safety.3

The VSD is a large-linked database that is used to analyze vaccine related information pertaining to more than seven million people affiliated with the partner HMOs. The HMOs include Group Health Cooperative of Puget Sound in Washington, Northwest Kaiser Permanente in Oregon, Northern California Kaiser, Southern California Kaiser Permanente health programs, Marshfield Clinic, Kaiser Permanente Colorado, Health Partners Healthcare Minneapolis, and Harvard Vanguard.4

The patients’ complete medical records at these HMOs are used to analyze vaccine related information pertaining to more than 5.5 million people.

Although originally the VSD surveyed only children between 0 and 6 years of age, it has been expanded to include information on adolescents and adults.

The VSD database allows researchers to access already gathered data on:

  • vaccine type
  • date of vaccination
  • concurrent vaccinations (those given during the same visit)
  • the manufacturer
  • lot number
  • injection site 
  • potential adverse events3

The VSD has several advantages over other methods of post-licensure vaccine safety monitoring. First, it is a comprehensive medical database. All or almost all significant medical encounters should be in the available information. Thus, the VSD does not rely on a physician or parent recognizing that a given clinical illness might be related to vaccine and making a report. All medical illnesses whether vaccine related or not are already in the data base.

Second, all or almost all information on vaccination is also available. Thus, investigators can determine the most valid incidence rate of an illness following vaccination. Most importantly, the VSD allows the comparison of the incidence rate of a given clinical illness after vaccination with the incidence of the same event among unvaccinated individuals. Significantly higher rates after vaccination are supportive of a potential causal role for the vaccine in the illness.

On the other hand, if the rates are similar between vaccinees and nonvaccinees, the vaccine is unlikely to be playing a causal role in the event. Neither VAERS nor CISA has the capacity to make these types of comparisons.

Further, when there is concern that a vaccine may be causing an illness in a certain time frame after vaccination, the incidence during this period can be compared with the incidence at other times to see if it is significantly higher in the period of concern. Finally, the electronic records of diagnoses can be validated through actual review of the medical charts to assure the information is as accurate as possible.

Thus, the VSD allows researchers to plan vaccine safety studies and examine hypotheses generated from the medical literature, study possible signals from the passive VAERS program, and evaluate changes in the immunization schedule or from the introduction of new vaccines.3

For example, VSD data were used to study the rate of Sudden Infant Death Syndrome (SIDS) after anecdotal evidence suggested a possible link to the DTP vaccine. The risk of SIDS was found to be the same for vaccinated children as for unvaccinated children enrolled in the participating health maintenance organizations.5

The VSD has been used to examine a number of vaccine safety issues, including the safety of inactivated influenza vaccine among children, and the relation between immunizations and conditions such as asthma and neonatal death.67

Rapid cycle analysis is a relatively recent approach to using the VSD. (A) This is a method that allows looking for specific adverse event “signals” in an active, hypothesis-driven process rather than in a passive “data mining” process as is used with the VAERS database. That is, this system asks if a particular vaccine safety concern is more frequent or not in those who got the vaccine. Then if a signal is detected, the analysis can be refined into a more specific study. Here’s how it works for a newly introduced vaccine:

  • A specific vaccine safety concern is selected. 
  • Each week, the administrative database (lists of diagnoses) is examined for that specific adverse event in vaccinated persons.
  • The rate of that adverse event in a comparison group is determined (this becomes the “expected rate”).
  • The observed frequency of the adverse event after vaccine is then compared weekly to the expected frequency from the comparison group.
  • Statistical tools allow the researchers to set a “threshold” for when the number of cases in the vaccinated group should be recognized as greater than expected (it is a “signal”).
  • If a signal is detected, specific epidemiologic studies can be performed using data from the medical records.

This rapid cycle analysis was used to look for an association between the new meningococcal conjugate vaccination and Guillain-Barré Syndrome when a signal had seemed to have been detected. No association between MCV4 and GBS was found.8 Studies using the VSD database have some limitations:

  • The population in the participating HMOs is not wholly representative of the United States in terms of geography and socioeconomic status. • Vaccine coverage rates for most vaccines are very high in the participating HMOs, and thus few nonvaccinated controls are available to do a comprehensive comparison.
  • The VSD cannot easily assess mild adverse events that may escape medical attention such as fever.
  • The project is not large enough to examine the risk of extremely rare events (such as 1 in a million vaccinees) such as Guillain-Barré Syndrome after each influenza season.3
  • Because the database contains clinical information, it can only be accessed under circumstances which maintain patient confidentiality.9

Despite the above caveats, the VSD project is a powerful and cost-effective tool for the constant monitoring of vaccine safety. Studies performed using these data sets—along with many other studies—were used by the Institute of Medicine’s Vaccine Safety Review of thimerosal and autism to favor the rejection of a cause-effect relationship.