What Can Pharmacovigilance Databases Tell Us About SIRVA?
Pharmacovigilance databases have played an important role in the development of the SIRVA literature.
They have also been frequently misunderstood.
When hundreds or thousands of shoulder injuries are identified in a vaccine adverse-event database, it can be tempting to interpret the number as evidence that vaccination caused those injuries.
That is not what these databases are designed to establish. Systems such as VAERS are primarily early-warning and signal-detection systems. CDC explicitly states that VAERS accepts adverse events occurring after vaccination even when the reporter is unsure whether the vaccine caused the event, and that a VAERS report by itself does not establish causation.
For SIRVA, these databases are valuable, but only when their limitations are understood.
What Is Pharmacovigilance?
Pharmacovigilance is the ongoing surveillance of possible adverse events after a drug or vaccine enters widespread use.
In the United States, VAERS is jointly managed by CDC and FDA. Patients, family members, clinicians, manufacturers, and others can submit reports of health events occurring after vaccination.
The strength of this approach is scale. A rare event that might never appear in a clinical trial can begin to emerge when millions of doses are administered.
The purpose is therefore often to ask:
“Is something unusual being reported often enough that we should investigate it further?”
rather than:
“Did this vaccine cause this patient's condition?”
What Did VAERS Contribute to the SIRVA Literature?
One of the most important SIRVA pharmacovigilance studies was performed by Hibbs and colleagues, who reviewed VAERS reports following inactivated influenza vaccination from 2010 through 2017. Their analysis identified 1,220 reports of atypical shoulder pain and dysfunction. The study was useful for identifying recurring patterns across a much larger number of reports than had been available in the original SIRVA case series.
Pharmacovigilance Can Help Describe Patterns
Large reporting databases can help identify recurring features such as:
timing of symptom onset,
age and sex distributions,
vaccine types,
reported injection technique,
common symptoms,
imaging use,
treatment patterns,
and whether patients reported recovery at the time the report was submitted.
For example, the Hibbs VAERS analysis found that 70.7% of reports described symptom onset on the day of vaccination, a finding that contributes useful evidence to the observation that classic SIRVA tends to have rapid onset.
That does not prove causation in those individuals.
But when a similar temporal pattern repeatedly appears across independent case reports, clinical cohorts, medicolegal data, and pharmacovigilance reports, it becomes relevant to defining the clinical phenotype.
Bass and Poland subsequently examined VAERS reports following COVID-19 vaccination. Rather than simply counting every shoulder-related report, they attempted to improve specificity by manually reviewing cases, requiring physician evaluation, excluding prior shoulder dysfunction or trauma, and separately analyzing cases with diagnostic imaging. Their initial search identified 621 potentially relevant VAERS entries, of which 476 met their study definition; 306 had undergone physician evaluation. The authors also cross-referenced published case reports to reduce duplicate counting.
Among the more stringently selected imaging-confirmed subgroup, adhesive capsulitis and bursitis were prominent diagnoses.
This demonstrates an important principle:
Pharmacovigilance data become more clinically informative when reports undergo additional case validation.
They still do not become equivalent to a prospective controlled clinical study.
What Pharmacovigilance Databases Cannot Tell Us
The most important limitation is causation.
A report generally means:
vaccination occurred, and
a health event subsequently occurred.
It does not necessarily mean:
vaccination caused the health event.
CDC and FDA specifically caution that VAERS report counts cannot by themselves establish a causal association, determine the frequency of an adverse event, or establish its rate. Reports may be incomplete, inaccurate, coincidental, biased, or unverifiable.
That distinction is especially important for SIRVA because shoulder pain, rotator cuff disease, bursitis, and adhesive capsulitis also occur in people who have not recently been vaccinated.
Reporting Bias Matters
VAERS is a passive surveillance system. Someone has to decide to submit a report.
That creates several forms of bias.
Some SIRVA cases may never be reported. Conversely, awareness of SIRVA may increase reporting after media coverage, litigation, publication of new research, or changes in public awareness.
CDC specifically identifies both underreporting and reporting bias as limitations of VAERS.
This means the number of reports should not be treated as a direct measure of how many SIRVA cases actually occurred.
The Denominator Problem
To calculate incidence, we need both:
number of cases, and
number of people or vaccine doses at risk.
Traditional VAERS analyses frequently lacked a reliable denominator of administered doses corresponding precisely to the reports being studied.
Mackenzie and colleagues specifically noted that earlier VAERS SIRVA analyses could not reliably estimate incidence because the number of vaccines administered was unavailable. This is why statements such as:
“There were 1,220 VAERS reports, therefore SIRVA occurs at X rate”
are generally not justified without additional denominator and case-validation data.
EudraVigilance Provides Another Perspective
Mackenzie and colleagues later used EudraVigilance, the European Medicines Agency pharmacovigilance database, together with COVID-19 vaccine administration data.
EudraVigilance contains a mixture of patient, healthcare-professional, and pharmaceutical-company reports. The investigators applied explicit SIRVA criteria and attempted to compare identified cases with the number of administered COVID-19 vaccine doses.
This approach begins to address the denominator problem. But even with denominator data, spontaneous reports remain subject to incomplete reporting, diagnostic uncertainty, and variable case ascertainment.
Thus, an incidence estimate derived from pharmacovigilance data should still be interpreted differently from incidence measured through systematic active surveillance or a well-defined population cohort.
Pharmacovigilance Is Not the Same as a Clinical Cohort
This distinction is important in interpreting the SIRVA literature.
A case series begins with clinically evaluated patients.
A population cohort begins with a defined population and systematically evaluates outcomes.
A pharmacovigilance study begins with reports submitted to a surveillance system.
A medicolegal claims study begins with individuals who filed claims.
These populations are not interchangeable.
For example, Hesse and colleagues analyzed 476 conceded SIRVA claims from the National Vaccine Injury Compensation Program. That dataset contains unusually rich medical-record information, including examination findings, imaging, treatments, and reported administration errors.
But it is a compensation-selected population, not a pharmacovigilance database or population-based cohort. Our evidence synthesis therefore treats it separately.
This matters when comparing percentages across studies.
Even Better-Documented Cases Require Caution
More clinical detail does not automatically solve causation. Hesse and colleagues specifically warned that MRI abnormalities in SIRVA claims should not automatically be interpreted as evidence of vaccine-caused structural injury, because rotator cuff and other shoulder abnormalities are common in middle-aged and older adults independent of vaccination.
The same principle applies even more strongly to unverified pharmacovigilance reports.
A report containing the term “rotator cuff tear” tells us that the diagnosis was reported.
It does not tell us that vaccination created the tear.
So What Are These Databases Good For?
Their greatest value is signal generation and pattern recognition. For SIRVA, pharmacovigilance data can help investigators notice that:
persistent shoulder symptoms are repeatedly being reported after vaccination;
many reports describe rapid onset;
certain diagnoses such as bursitis and adhesive capsulitis recur;
some patients report injections placed unusually high; and
particular demographic or clinical patterns may deserve further investigation.
Those observations can then generate better questions for studies using medical records, imaging, active surveillance, or population-based methods.
CDC describes exactly this surveillance pathway: VAERS identifies potential safety signals, which may then be evaluated through systems such as the Vaccine Safety Datalink, CISA, or FDA's BEST system that are better suited to assessing whether an actual association exists.
Why This Matters for Causation
Pharmacovigilance evidence can contribute to general causation by helping demonstrate that a particular clinical phenomenon has repeatedly been observed after an exposure and deserves scientific investigation.
It is much less capable of establishing specific causation in an individual patient.
A VAERS report cannot tell us whether a particular patient's shoulder condition was caused by vaccination.
That analysis still requires consideration of:
prior shoulder status,
symptom timing,
injection technique,
anatomical plausibility,
imaging,
examination findings,
competing diagnoses,
and clinical course.
The existence of similar reports can provide context. It cannot substitute for analysis of the individual case.
The Bottom Line
Pharmacovigilance databases have contributed meaningfully to our understanding of SIRVA.
They allow researchers to examine far more reported events than would be available through conventional case series, identify recurring clinical patterns, generate safety signals, and develop hypotheses for subsequent study.
But they are fundamentally surveillance tools.
The most appropriate interpretation is:
Pharmacovigilance databases can tell us what is being reported after vaccination, how often certain patterns appear among those reports, and what questions deserve further investigation. They generally cannot, by themselves, tell us how often SIRVA truly occurs or whether vaccination caused the shoulder condition in a particular patient.
That distinction is particularly important in medical-legal analysis, where a reported association should not be presented as equivalent to established causation.
References
Hibbs BF, Ng CS, Museru O, Moro PL, Marquez P, Woo EJ, Cano MV, Shimabukuro TT. Reports of atypical shoulder pain and dysfunction following inactivated influenza vaccine, Vaccine Adverse Event Reporting System (VAERS), 2010-2017. Vaccine. 2020;38(5):1137-1143.
Hesse EM, Atanasoff S, Hibbs BF, et al. Shoulder Injury Related to Vaccine Administration (SIRVA): petitioner claims to the National Vaccine Injury Compensation Program, 2010-2016. Vaccine. 2020;38(5):1076-1083.
Bass JR, Poland GA. Shoulder injury related to vaccine administration (SIRVA) after COVID-19 vaccination. Vaccine. 2022;40(34):4964-4971.
Mackenzie LJ, Bushell M-J, Newman P, Cunningham J, Woodward AP, Silk-Jones J, Nguyen C, Bushell MA. What three years of COVID-19 vaccine administration reveals about the incidence of shoulder injury related to vaccine administration (SIRVA). Vaccine. 2025 Apr 2;51:126892.
Martín Arias LH, Sanz Fadrique R, Sáinz Gil M, Salgueiro-Vazquez ME. Risk of bursitis and other injuries and dysfunctions of the shoulder following vaccinations. Vaccine. 2017;35(37):4870-4876.
MacMahon A, Nayar SK, Srikumaran U. What do we know about shoulder injury related to vaccine administration? An updated systematic review. Clin Orthop Relat Res. 2022;480(7):1241-1250.
Centers for Disease Control and Prevention; U.S. Food and Drug Administration. Guide to Interpreting VAERS Data. Vaccine Adverse Event Reporting System (VAERS). U.S. Department of Health and Human Services. Accessed August 11, 2026.
Comments