Pharmacovigilance Interview Questions

"I studied pharmacy, and in my final year we had a module on adverse drug reactions where we looked at real case reports. What struck me was that a medicine can pass every trial and still show a new risk once thousands of people use it, and someone has to catch that. After graduating I did a short course in pharmacovigilance and practised reading case reports and coding events. I like work where detail matters and there's a clear reason behind every rule. I want to start in case processing because that's where you really learn how a report turns into safety data, and later I'd like to move towards medical review or signal work."

"I expect a day built around a case queue. I'd pick up new reports, check whether each one is valid, look for duplicates, enter the data, code the events and drugs, assess seriousness, write the narrative and send it on for quality review, all while keeping an eye on due dates. Some cases will need follow-up questions to the reporter. I know it's repetitive in places, but I like that each case is a real patient and each field ends up in safety data someone relies on. My pharmacy background helps me read medical history and drug names quickly, and I'm careful by nature, which matters more here than speed on day one."

"I looked at all three. Clinical research is mostly about running the trial well, and regulatory affairs is about getting and keeping the approval. Pharmacovigilance runs through the whole life of the medicine, from the first trial patient to years after launch, and it's the part that answers one question: is this still safe for the people taking it? I like that a single well-processed case can be part of the evidence that changes a label or adds a warning. Medical writing appeals to me too, and narrative writing in safety gives me some of that, but I'd rather be close to the actual patient data."

"Pharmacovigilance is the work of detecting, assessing, understanding and preventing adverse effects and other problems with medicines. We need it after approval because clinical trials, however well run, have limits. They include a limited number of people, they run for a limited time, and they often leave out the elderly, pregnant women, children and people on many other medicines. So a reaction that happens in one patient in ten thousand, or only after years of use, or only with a certain other drug, may never appear in a trial. Once the medicine is used widely, reports from doctors, patients and literature are how those risks come to light, and that's what lets companies and regulators update labels and protect patients."

"An adverse event is any unwanted medical occurrence in someone who has taken the medicine, whether or not the drug caused it. If a patient on a blood pressure tablet breaks their arm in a fall, that's still an adverse event. An adverse drug reaction is narrower: it's a harmful, unintended response to the drug where a causal relationship is at least a reasonable possibility. A serious adverse event is any event that results in death, is life-threatening, needs hospitalisation or prolongs one, causes persistent or significant disability, causes a congenital anomaly, or is judged medically important. So seriousness and causality are two separate questions. An event can be serious and unrelated, or related and not serious."

"On what's described, no. Severity and seriousness are different things. Severity is intensity, so a headache can be mild, moderate or severe. Seriousness follows fixed criteria: death, life-threatening, inpatient hospitalisation or prolonged hospitalisation, persistent or significant disability, congenital anomaly, or another medically important event. A day in bed isn't any of those. Life-threatening also has a strict meaning: the patient was actually at risk of death when the event happened, not that it might have become fatal if it got worse. I'd still read the whole report, because if the headache turned out to be a bleed in the brain, or they were admitted, the answer changes. And if I'm unsure about medical importance, I'd flag it for the medical reviewer rather than guess."

"I look at a few things. Does the timing make sense between starting the drug and the event? Did it improve when the drug was stopped, which is dechallenge, and come back if it was restarted, which is rechallenge? Is there another explanation, like the disease itself, another medicine or a known condition? And is the reaction known for this drug or its class? The WHO-UMC system puts the answer into categories from certain and probable through possible and unlikely, down to conditional or unassessable when information is missing. The Naranjo scale does something similar with a set of scored questions. I've mostly used WHO-UMC. Whatever the method, I record my reasons, because a category without reasons is hard to review or defend."

"Expectedness is always judged against a reference document, not against what I've read in general. For a marketed product that's the company's core safety information or the local label, and in a trial it's the reference safety information, usually in the investigator's brochure. An event is unexpected if its nature, severity, specificity or outcome isn't consistent with that reference. So if the label mentions hepatitis and the case is liver failure, that's unexpected by severity. If the label says stroke and the case is a more specific event like cerebral thromboembolism, it's unexpected by specificity. Listed and labelled split the same idea: listed means it's in the core company safety information, labelled means it's in one country's approved label. Because labels differ, one event can be labelled in one country and unlabelled in another."

"Both opinions get recorded, separately. The doctor's view goes in as the reporter's causality, and our medical reviewer's view goes in as the company assessment, with the reasoning in the company comment. What the company can't do is downgrade the reporter's opinion to avoid reporting. In a trial that rule is spelled out: if the sponsor disagrees with the investigator, the case is still handled as related and both opinions go with the report. For a spontaneous report, the reporter's suspicion is enough, and spontaneous reports are generally treated as suspected reactions anyway, because someone chose to report them. So if this case is serious and unexpected, it still goes as an expedited report. The reviewer's reasoning isn't wasted, though. It tells anyone reading the case later, including in signal review, why the disease looks like the more likely cause."

"A colleague marked a case non-serious where a patient went to the emergency department with a severe allergic reaction and was sent home the same day. I agreed it wasn't a hospitalisation, but the report said the patient had swelling of the throat and needed adrenaline. I felt it met the medically important criterion. I didn't change it myself. I messaged her with the part of the source I meant and the definition from our SOP, and asked what she thought. She said she'd only looked at the outcome. We took it to the medical reviewer, who agreed it should be serious. What I took away is that it's always easier when you bring the exact words from the source and the rule, not just your opinion."

"It was a literature case about an elderly patient on eight medicines, two of them our products, who developed kidney injury, low sodium and confusion, and later died. The article mentioned the drugs in a table and the events in the discussion, so it was hard to see what happened when. I began by building a simple timeline from every date in the paper. That showed our first product was started months before, but the second was started days before the kidney injury. I coded each event separately rather than lumping them, because the causality was different for each. I flagged to the medical reviewer that the author suspected another drug, not ours, and I wrote the narrative in strict date order. The reviewer said the timeline saved a lot of time, and I now build one for every complex case."

"A valid case needs four things: an identifiable reporter, an identifiable patient, at least one suspect product and at least one adverse event or reaction. Identifiable doesn't mean a full name. For the patient, one qualifier is enough, such as initials, age or an age group, sex or date of birth, so that we know it's a specific person. For the reporter, we need enough to know a real person made the report, like a name, a qualification or contact details. If one element is missing, the case isn't valid yet, but I don't discard it. I log it, try to get the missing detail through follow-up, and the moment all four are present, that becomes the case's day zero."

"First the report arrives through a call, email, form or literature, and it's logged with the date we received it. At triage I check the four minimum criteria, run a duplicate search, and make a first call on seriousness so urgent cases go to the top. Then comes data entry: patient details, history, suspect and other drugs, events, dates, lab results and outcome. I code the events in MedDRA and the drugs against the product dictionary, and the case is assessed for seriousness, expectedness and causality. I write the narrative, then the case goes through a quality check and medical review. After that it's submitted electronically to the authorities or partners that need it, within their timelines. If details are missing, we send follow-up questions, and new information reopens the case as a follow-up version."

"Right now it's not a valid case, because there's no identifiable patient. 'Several customers' isn't a specific person. But I wouldn't just close it. I'd log the call with the date and her contact details, as our process says, and follow up while she's still on the line if I can. I'd ask if she can give anything about each person: age, sex, initials, when they started the product, and what the rash was like. If she can, each identifiable customer becomes its own case, and day zero is the date that minimum information came in. If she can't, it stays recorded as an invalid report, and it can still be useful as background information in safety review."

"A patient in the same age group with the same drug and event from two reporters could easily be one patient, like a doctor reporting and the patient reporting too. So I'd compare everything: initials, sex, date of birth, event onset dates, dose, country, hospital, lab values. If they match, it's the same case and I'd add the new report as follow-up to the existing one, keeping both sources linked, rather than making a second case. If key details conflict, it's likely two patients and I'd process them separately. If it's in between, I wouldn't guess. I'd raise it with my lead and maybe ask the reporters a question. Duplicates matter because they double-count events and can make a signal look stronger than it is."

"Even though she has no symptoms, pregnancy exposure is a special situation we collect and track. I'd thank her, stay calm and not give medical advice. I'd point her to her doctor about whether to continue the medicine. Then I'd gather what the process asks for: the dose, when she started and stopped, the date of her last period or how far along she is, the expected due date, other medicines, and her doctor's details if she agrees to share them. It gets recorded on the pregnancy follow-up path, and we follow up towards the expected due date to learn the outcome, whether that's a healthy baby, a birth defect or a loss. If an adverse event is reported at any point, then it becomes a case with an event and is assessed as normal."

"At my last company I had a case where a nurse reported a patient with a seizure on one of our products, but there was no dose, no start date and no history. Without those, causality was close to impossible. Instead of sending the long standard form, I wrote three short questions: the dose and start date, any history of seizures or other medicines, and the outcome. The nurse didn't reply, so I sent the reminder our process required and then tried the phone number on file. On the second attempt I reached her, and she told me the patient had epilepsy and had missed doses of their seizure medicine. That changed how the reviewer saw causality. If she'd never replied, I'd have documented every attempt and closed the follow-up as per the SOP."

"MedDRA has five levels. At the top is the System Organ Class, then High Level Group Term, High Level Term, Preferred Term, and at the bottom the Lowest Level Term. I code at the Lowest Level Term, choosing the one that matches the reporter's words most closely, and it rolls up to a Preferred Term, which is what's mostly used for analysis. Each Preferred Term has one primary System Organ Class. If the doctor reports a diagnosis with its usual symptoms, say pneumonia with fever and cough, I'd code pneumonia and not the symptoms that are typical of it. But if there's a symptom that isn't part of the diagnosis, I code that too. And I never create a diagnosis the reporter didn't make, even if the symptoms point to one."

"A good narrative lets a medical reviewer understand the case without opening the source. I start with who reported it, the patient's age and sex, and the suspect drug with its indication. Then I tell the story in date order: relevant history, other medicines, the dose and start date, when the event began, what tests showed, how it was treated, whether the drug was stopped or restarted, and the outcome. I end with the reporter's causality and any details that are still missing. The mistakes I see most are dates out of order, abbreviations nobody explains, and a writer's own opinion slipped in as if the reporter said it. Another one is leaving out important negatives, like a normal liver test, which can matter a lot for causality."

"A signal is information, from one or more sources, that suggests a new possible causal link between a drug and an event, or a new aspect of a known link, strong enough to be worth checking. It's a hypothesis, not a conclusion. We find them by reviewing individual cases, especially serious and unexpected ones, by watching the literature, and by statistical methods in large databases, like disproportionality scores such as the proportional reporting ratio, which flag when an event is reported more often with this drug than with others. Then comes signal management: validate it by checking the cases are real and relevant, prioritise by seriousness and impact, assess all the evidence in depth, and decide on action, such as a label update, more monitoring or closing it. A high score alone never proves anything, because publicity or stimulated reporting can inflate counts."

"An individual case report tells the authorities about one patient, often within days. A PSUR, or the newer format called a PBRER, steps back and looks at everything we know about the product over a period and cumulatively since approval. It covers how many patients were exposed, new safety findings from cases, trials and literature, signals opened or closed in the period, an evaluation of the known and potential risks, the evidence of benefit, and an integrated conclusion on whether the benefit-risk balance is still favourable, with any actions proposed. The data lock point is the cut-off date: data received up to that date goes in, and anything later goes in the next report. How often it's due depends on the product and the region. The PBRER's big shift was putting benefit and risk side by side, not just listing risks."

"Day zero is the date the company, or anyone acting for it such as a partner, a sales rep or a vendor, first has a report with all four minimum criteria. It's not when the safety team opens the email, and weekends count. From there, the logic is that the more urgent the risk, the shorter the clock. For marketed products, serious unexpected cases are commonly due within 15 calendar days, and some regions ask for all serious cases in that time and non-serious ones later. In clinical trials, fatal or life-threatening unexpected suspected reactions are due within 7 calendar days, with a complete report within about another week, and other serious unexpected ones within 15. The exact rules differ by region, so we follow each one's requirements. When significant follow-up arrives, like a new serious event or a change in causality, that receipt date becomes day zero for the follow-up report."

"First, day zero is the day that email reached the company, not the day I found it. I can't change that date to make it look on time. So I'd tell my lead straight away and put the case at the top of the queue, because it may already be close to or past its deadline. We'd process and submit it as fast as quality allows, and record why it was late. Then there's the bigger problem: if one case sat there, others might have too. I'd check the rest of that inbox for anything else unlogged. Finally, it needs a deviation or quality record and a root cause, for example nobody owning that inbox on certain days, with a corrective action like a named owner and a daily check."

"I'd treat it as urgent, because the company became aware when he heard it, not when safety hears it. A hospitalisation also makes it serious, so the clock may be well along already. Right there I'd ask him what he knows: the doctor's name and contact, anything about the patient like age or sex, the drug, what happened, and the date the doctor told him. I'd ask him to send it to the safety inbox today, or help him do it through the usual route. The case must show the real date he learned about it, even if that makes it late. I wouldn't lecture him over lunch, but afterwards I'd let my lead know, because it might mean field teams need a quick reminder on how and when to report."

"It means that if an inspector asks why a case was late, or whether we tried to follow up, the only answer that counts is the record. Saying 'I did call them' isn't enough. So in daily work I document every follow-up attempt with the date and method, record why I made a judgement call, like choosing one code over another, note the reason if a case goes late, and make sure training and SOP updates I've read are signed off. It also shapes how I think about mistakes. If I find an error, I'd rather it's raised and recorded with a fix than quietly corrected, because a documented deviation with a corrective action shows a working system. A hidden one is what really worries an inspector."

"That's a big inconsistency, because if the event really started first, the drug is unlikely to have caused it. First I'd re-check the source myself. Maybe I misread it, or the dates are in a different format, like day and month swapped, or the patient had an earlier course of the same drug that isn't clear. If it's still inconsistent, I wouldn't quietly fix it to what I think is right. I'd enter the data as reported, mention the inconsistency in the narrative so the reviewer sees it, and send a follow-up question to the reporter asking them to confirm both dates. When the answer comes back, the case gets updated and causality looked at again."

"In my first few months, a quality reviewer found I'd coded 'liver function test increased' when the report actually gave a diagnosis of drug-induced liver injury. I'd focused on the lab values and missed the doctor's conclusion further down the page. That mattered because it made the case look much milder in analysis than it was, and it could have affected expectedness. I corrected the coding and the narrative. After that I changed my habit: before coding anything, I read the whole source once from start to end, and I look specifically for any diagnosis the reporter gives. I also started keeping a short list of my own repeat errors and checking it before sending a case to review. I haven't had that finding again."

"At my last company, a large literature search returned many new articles in the same week two colleagues were on leave. My queue roughly doubled. The first thing I did was sort everything by due date and seriousness, so serious cases close to their deadline came first and non-serious ones with more time moved back. I told my lead the same day, with a count of cases due in the next few days, instead of waiting until something was late. She moved some non-serious cases to another team member. I kept my usual self-check before sending each case to review, because a rushed case that fails quality only costs more time. We submitted everything on time that week. What I learned is that flagging early, with specifics, gets help far more easily than a vague 'I'm busy'."

"When I moved onto a new client's products, they had their own conventions, like which fields were mandatory and how to write company comments, and a different database layout from what I knew. I read their case processing guideline end to end first, then worked through the training cases before touching live ones. I kept my own one-page note of everything that differed from what I was used to, and put it next to my screen. For my first couple of weeks, I asked a senior to look over my cases before quality review, and I noted every correction. Within about a month I was processing cases on my own with few findings. The note became something new joiners on that client used too."

"I'll be honest, the risk with repetitive work is that you start seeing what you expect instead of what's there. So I rely on habits, not willpower. I read the full source before entering anything, I use my own checklist before sending a case to review, and I take a short break between complex cases rather than rushing into the next one. I also try to remember that each case is a person who had something go wrong, and that the data might one day be the case that confirms a signal. That helps. And I like variety within the routine, like a literature case after a run of simple ones, so I'll ask to mix it up when I can."