Vaccinations and the Risk of Relapse in Multiple Sclerosis

Study subjects were chosen from among patients with multiple sclerosis attending neurology centers affiliated with the European Database for Multiple Sclerosis (EDMUS) network.⁴⁰ The EDMUS software, introduced in 1990 and currently used in more than 180 clinics worldwide,⁴¹ enables users to collect in a standardized fashion retrospective and prospective data on patients with multiple sclerosis. Among hospital-based medical centers with more than 500 patients enrolled in their data bases, six centers in France, Spain, and Switzerland successfully underwent a pilot phase and were selected for the study.

Eligible patients were initially identified from the EDMUS data base of each participating center by means of a computer algorithm. All patients with a definite or probable diagnosis of multiple sclerosis according to the diagnostic criteria of Poser et al.⁴² and with at least one index relapse of the disease were considered for inclusion. A relapse was defined as an occurrence, reappearance, or worsening of symptoms of neurologic dysfunction that lasted more than 24 hours. Fatigue alone or a transient, fever-related worsening of symptoms was not considered to be a relapse. Symptoms that occurred within one month of each other were considered to be part of the same relapse.⁴⁰ The index relapse was defined as the first relapse that occurred between January 1, 1993, and December 31, 1997, was confirmed at an outpatient visit or during a hospitalization in the center within 2 months, and was preceded by a relapse-free period of at least 12 months. According to the design of the study, the initial manifestations of the disease could not be selected as the index relapse.

The fact that patients met the criteria for inclusion and the features of relapses were confirmed by a neurologist on the basis of medical files. The level of certainty of the index relapse was assessed as follows: definite, if the neurologist determined that highly suggestive symptoms were present and that there was objective confirmation on examination; probable, if highly suggestive symptoms were present but there was no objective confirmation on examination; and possible, if only atypical or minimal symptoms were present. Additional information regarding the history of the disease was also collected, including the course of disease,^43,44 the course of disability,⁴⁵ the occurrence of any relapses before the index relapse, and prescriptions for corticosteroids or immunosuppressive or immunomodulating treatments during the 18 months before the index relapse.

Eligible patients were invited by letter to seek information about their vaccinations between January 1, 1992, and December 31, 1997 — i.e., for a standard six-year period starting one year before the period of interest for index relapses. They were provided with a list of existing vaccines and conditions that might be related to vaccination.

Interested patients were then interviewed by the study administrator of their center, who used a standardized telephone questionnaire. For all patients, data were collected regarding vaccinations received during the entire six-year period (1992 through 1997). Patients were unaware of the specific hypothesis being tested and of the date of the relapse of interest. The interviewer was unaware of which relapse was of interest. The interview included systematic questions related to all vaccinations the patient might have received. Whenever a vaccination was reported, detailed information on the nature, name, and date of the vaccine was collected. A vaccination was defined as the administration of a commercial preparation of vaccine, regardless of whether it was monovalent or multivalent (i.e., specific for several infectious diseases), whether it was a single vaccination or part of a series, and whether it was for primary immunization or was a booster. Information about reported vaccinations was confirmed on the basis of written medical documentation, usually a copy of the vaccination record. When the copy could not be obtained from the patient, written confirmation was sought from the physician who had performed the vaccination; a high priority was placed on obtaining such confirmation for vaccinations reported to have been received during the 15 months before the index relapse. At the end of the study, an additional confirmation procedure was performed with the data from a random sample of 97 patients drawn from all centers who reported having received no vaccinations between 1992 and 1997: a copy of the patients' vaccination records was requested, and their general practitioners were contacted to verify the absence of vaccinations.

The case–crossover design of this study is equivalent to a case–control approach in which patients serve as their own controls, with data used from different points in time.⁴⁶ The subjects were patients who had had a relapse of multiple sclerosis. We assessed the relative risk of relapse associated with vaccination by comparing the exposure to vaccination in the period immediately preceding the index relapse (the risk period) with that in control periods that preceded the risk period and were not followed by a relapse (Figure 1).

On the basis of the literature,^1–39 analyses were performed under the assumption that a relapse due to vaccination would occur within two months after vaccination; we also performed analyses using one- and three-month periods. Patients who had received at least one vaccination during the two months before the index relapse were considered to have been exposed during the risk period. For each patient, the four two-month periods preceding the risk period were used as control periods. In the analyses based on a one-month risk period, we used the previous nine one-month periods as control periods, and in the analyses based on a three-month risk period, we used two three-month control periods.

Each exposure to vaccine during the 10 months before the index relapse was allocated either to the risk period or to a control period, depending on when it occurred with respect to the date of the index relapse and on the duration of the risk period chosen for the analysis. We measured the association between exposure to vaccine and the occurrence of a relapse within two months after exposure by estimating the odds ratio for exposure during the risk period as compared with exposure during the control periods. Conditional logistic regression was used for these estimations, because there was a 1:4 ratio between the risk period and the control periods for each patient. The main analysis was performed with the inclusion of the data for all patients, regardless of the level of certainty about their index relapse, and all confirmed vaccinations. On the basis of the results of a pilot phase, the probability of exposure to vaccination during any two-month control period was estimated to be 2.8 percent. With four control periods per patient and a one-tailed α level of 0.05, the inclusion of 600 patients allowed the study to detect an odds ratio of 2.0 with a statistical power of more than 90 percent. Therefore, the protocol specified that enrollment would stop when this target number of patients was reached. To assess the robustness of the findings, sensitivity analyses were performed for one- and three-month periods of exposure, for all reported vaccinations even if they were not confirmed, and for only the patients with a definite index relapse. Additional analyses were performed according to the type of vaccine. All analyses were performed with the use of SAS software.⁴⁷

To ensure the strict application of the protocol, a user's manual was provided to the study administrator at each center and the study was monitored by research assistants from a clinical research organization. All patients gave informed consent for their inclusion in the study; consent was oral in all cases and written in the majority of cases. The use of computerized files at the hospitals in France was authorized by the French Commission Nationale de l'Informatique et des Libertés. The study protocol was approved by the ethics committees of the hospitals in Basel, Switzerland, and Barcelona, Spain. The organization of the study was approved by the Société Française de Neurologie. The study was overseen by a scientific advisory board composed of independent experts who approved the protocol, conduct, analysis, and publication of the study.

At the six participating centers, a total of 1176 patients were found by the computer algorithm to be eligible; 1037 of these patients were confirmed as eligible on the basis of medical records; 1009 were sent an informational letter; and 960 consented to be interviewed. Recruitment was stopped when the target number of patients was reached. At that point, 643 patients had been interviewed (Figure 2). The base-line characteristics of the 216 patients who did not consent to be interviewed or who were excluded for the reasons listed in Figure 2 did not differ substantially from those of the patients who were included.

The index relapse was considered by the neurologists at each center to be definite in 492 patients (77 percent), probable in 112 patients (17 percent), and possible in 39 patients (6 percent). By design, the interval between the index relapse and the previous relapse was at least 12 months. It was greater than 18 months in 468 patients (73 percent).

The patients reported having received 960 vaccines during the period between 1992 and 1997; of these, 748 (78 percent) could be confirmed. Eighty percent of the confirmed vaccinations (596 of the 748) were confirmed by means of copies of vaccine certificates. During the confirmation procedure used for the 97 randomly sampled patients with no reported vaccinations, 3 patients could not be contacted, nonvaccinated status could be formally confirmed for 89 patients, and confirmation was not possible for 5. No vaccinations were discovered in any patients during this procedure.

Of the 643 interviewed patients, 260 (40 percent) had received at least one confirmed vaccination between 1992 and 1997, 57 (9 percent) reported vaccinations that could not be confirmed, and 326 (51 percent) reported receiving no vaccination at all. The patients with confirmed vaccinations were younger, had had fewer relapses before the index relapse, and had a lower Kurtzke disability score⁴⁵ (range, 0 to 10, with higher scores indicating more severe disability) than the patients with no validated vaccinations (Table 1).

During the 12 months preceding the index relapse, 96 patients (15 percent) reported 144 vaccinations, 135 of which (94 percent) could be confirmed. Of these 135 vaccinations, 49 percent were against tetanus, either alone or in combination with poliovirus or diphtheria or both, and only 2 were live attenuated vaccines (Table 2). A total of 89 patients (14 percent) received these 135 confirmed vaccinations during that 12-month period, 60 receiving 1 vaccination only, 18 receiving 2, 8 receiving 3, and 3 receiving 4 or more. Their base-line characteristics are provided in Table 1. In 71 of these patients (80 percent) the index relapse was assessed as definite, in 15 patients (17 percent) as probable, and in 3 patients (3 percent) as possible.

The proportion of patients vaccinated during any one of the four two-month control periods ranged from 2.8 to 4.0 percent; the proportion vaccinated during the two-month risk period was 2.3 percent (Table 2). The relative risk of relapse associated with any vaccine exposure during a two-month period was 0.71 (95 percent confidence interval, 0.40 to 1.26), with specific relative risks for tetanus, hepatitis B, and influenza vaccines ranging from 0.22 to 1.08 (Table 3).

The overall relative risk of relapse associated with exposure to vaccination remained stable between 0.68 and 0.79 when one to five two-month control periods were used in the analysis. It did not change substantially when we performed sensitivity analyses using periods of varying lengths. The relative risk of relapse associated with any vaccine exposure during a three-month period was 0.58 (95 percent confidence interval, 0.35 to 0.99). With a one-month risk period and three to seven control periods, the relative risk varied from 0.97 to 1.18, with confidence intervals wider than 0.6 to 1.8. When all reported vaccinations were considered even if they were not confirmed, the relative risk was 0.76 (95 percent confidence interval, 0.44 to 1.31). The relative risk remained stable when the analysis was restricted to patients with a definite relapse (0.58; 95 percent confidence interval, 0.29 to 1.15). The relative risk was not substantially modified by consideration of a series of other characteristics of the index relapse (calendar year, month of onset, or time elapsed since the previous relapse).