L3 EBP Observational studies in clinical medicine

Overview

This lecture introduces the family of observational study designs (case study/series, ecological, cross-sectional, case-control, cohort) and the situations where they are most useful, then uses the history of AIDS investigation to show these designs building on one another in real practice. It then turns to case reports as a source of drug-safety knowledge, post-marketing surveillance in New Zealand, and pharmacoepidemiology as an applied field that draws on cohort and case-control methods to answer real-world drug-safety questions.

Types of observational studies

The lecture lists six observational study designs:

• Case study
• Case series
• Ecological study
• Cross-sectional study / survey
• Case-control study
• Cohort study (prospective or retrospective)

Ecological studies are noted as being able to be either descriptive or analytical in design; by the structure of the list, this may extend to the designs listed below it too (cross-sectional, case-control, cohort). The transcript does not otherwise define each design.

Situations where observational studies are used

• New diseases
• Very rare conditions
• Post-marketing surveillance (detecting rare adverse effects)
• Drug safety in pregnancy
• Identifying risk factors

The AIDS/HIV story: observational designs building on each other

Case series (1981). An MMWR report (June 1981) described 5 previously healthy, young, homosexual men treated for biopsy-confirmed Pneumocystis carinii pneumonia at three Los Angeles hospitals between October 1980 and May 1981; two died. All had laboratory-confirmed current or previous cytomegalovirus (CMV) infection and candidal mucosal infection. Case details for four of the five patients [flag: a fifth patient is referenced by the summary but not shown/legible on the slide]:

• Patient 1: 33-year-old man; P. carinii pneumonia and oral candidiasis after 2 months of fever, elevated liver enzymes, leukopenia, CMV viruria; falling CMV complement-fixation titre (256 to 32); deteriorated despite TMP/SMX, pentamidine and acyclovir; died.
• Patient 2: 30-year-old man; P. carinii pneumonia after 5 months of daily fever, elevated liver-function tests, CMV viruria and documented CMV seroconversion; leukopenia and mucosal candidiasis; pneumonia responded to intravenous TMP/SMX but daily fever persisted.
• Patient 3: 30-year-old man; oesophageal/oral candidiasis responded to amphotericin B; later hospitalised for P. carinii pneumonia that responded to oral TMP/SMX; oesophageal candidiasis recurred and was treated again with amphotericin B; oesophageal biopsy positive for CMV.
• Patient 4: 29-year-old man with Hodgkin’s disease successfully treated 3 years earlier with radiation alone; developed P. carinii pneumonia, did not improve on intravenous TMP/SMX and corticosteroids, and died; postmortem showed no Hodgkin’s disease but P. carinii and CMV in lung tissue.

Case-control study (1983). Jaffe et al. (Annals of Internal Medicine) conducted a case-control study in New York City, San Francisco, Los Angeles and Atlanta to identify risk factors for Kaposi’s sarcoma and PCP in homosexual men: 50 cases (39 Kaposi’s sarcoma, 8 PCP, 3 both) versus 120 matched homosexual male controls (from STD clinics and private practices). The variable most strongly associated with illness was the median number of male sexual partners per year (61 for cases versus 27 and 25 for clinic and private-practice controls respectively). Cases were also more likely than controls to have been exposed to faeces during sex, to have had syphilis and non-B hepatitis, to have been treated for enteric parasites, and to have used various illicit substances. Conclusion: aspects of a lifestyle shared by a subgroup of homosexual men were associated with increased risk of Kaposi’s sarcoma and PCP. [flag: the introduction text on this slide is cut off before completion]

Notably, the causative retrovirus (HIV) was not identified until May 1983, yet interim national recommendations for preventing AIDS had already been issued in March 1983, based on epidemiological findings such as these:

1. Avoid sexual contact with persons known or suspected to have AIDS; members of high-risk groups should know that multiple partners increase risk.
2. As a temporary measure, members of at-risk groups should refrain from donating plasma/blood.
3. Evaluate screening procedures (laboratory tests plus history and examination) for excluding high-risk plasma/blood.
4. Adhere strictly to medical indications for transfusion; encourage autologous transfusion.
5. Continue work toward safer blood products for haemophilia patients.

This sequence illustrates a key strength of observational studies: risk factors can be identified and acted on before an underlying cause is confirmed.

Cohort studies establishing natural history.

• The Multicentre AIDS Cohort Study (MACS), initiated in 1983 at Johns Hopkins, the University of Pittsburgh and UCLA, is a long-running cohort of around 6,972 men who have sex with men, still publishing on the natural history of untreated and treated HIV infection after more than 35 years, with a very large accumulated volume of person-years and biological specimens collected. [flag: several summary figures in this abstract were too small/blurred to transcribe reliably]
• The AGEhIV cohort study (Amsterdam) is a prospective, longitudinal cohort comparing ageing-associated comorbidities in 596 HIV-positive and 550 HIV-negative participants aged 45 years or older, assessed every two years; enrolled October 2010 to October 2012 and followed to October 2018 (median follow-up 5.9 years, IQR 5.7-6.9); 20.2% overall dropout. The number of comorbidities increased at a similar rate over time in both groups (rate ratio per year approximately 1.04 for HIV-positive and 1.05 for HIV-negative participants), but each additional comorbidity was associated with a substantially increased risk of death (hazard ratio 3.33), and HIV-positive participants had a greater increase in disability-adjusted life years (DALYs) over time than HIV-negative participants. The findings support strategies to optimise prevention, screening and early intervention in people ageing with HIV. [flag: several numeric values and one p-value in this abstract were partly illegible]

Case reports and case series in practice

Beyond AIDS, the lecture gives further examples of case reports/series characterising new or rare conditions:

• First case of 2019 novel coronavirus in the United States (NEJM, January 2020): described identification, diagnosis, clinical course and management of a single Washington State patient during the early Wuhan-associated outbreak, including progression to pneumonia by illness day 9; illustrated the value of coordinated clinical, laboratory and public-health responses to a newly emerging disease.
• Human cases of highly pathogenic avian influenza A(H5N1) (NEJM, February 2025): a case series of 46 US patients identified March to October 2024, following widespread H5N1 infection in dairy cattle and poultry. Of 45 patients with a known animal exposure, 20 had been exposed to infected poultry and 25 to dairy cows; median age 34 years; illness was mild in all cases, with no hospitalisation or deaths; conjunctivitis was the most common symptom (93%), followed by fever (49%) and cough (39%); 87% received a neuraminidase inhibitor; further cases were found among 97 household contacts; no human-to-human transmission was identified; personal protective equipment use among exposed workers was suboptimal. [flag: exact percentages/figures were read as legibly as possible from a small rendering]
• Duodenal gastrointestinal stromal tumour (GIST) in a 14-year-old boy (BMC Cancer, 2007): a rare paediatric presentation. The boy had severe anaemia from recurrent upper gastrointestinal haemorrhage; endoscopy, small-bowel series, scintigraphy and video capsule endoscopy had all been negative elsewhere. Emergency surgery for a bleeding recurrence found a 4 cm mass in the fourth part of the duodenum, resected with adequate margins; the tumour stained positive for CD117 (c-KIT), with fewer than 5 mitoses per 50 high-power fields and a wildtype c-KIT/PDGFRA genotype, confirming a low-aggressiveness GIST. The patient remained disease-free at two years. This illustrates how case reports can document extremely rare presentations, of a tumour usually seen in adults, that standard investigation had failed to identify.

Drug safety in pregnancy: case reports underpinning classification

Case reports are also a major source of information on drug safety in pregnancy, feeding into prescribing classification systems such as the Australian categories for prescribing medicines in pregnancy:

• Category A: taken by many pregnant women/women of childbearing age without any proven increase in malformations or other harmful fetal effects.
• Category B (subcategorised because human data are lacking or inadequate, based instead on animal data): taken by only a limited number of pregnant women without an observed increase in malformation or harm.
  • B1: animal studies show no evidence of increased fetal damage.
  • B2: animal data are inadequate or lacking, but available data show no evidence of increased damage.
  • B3: animal studies show evidence of increased fetal damage, of uncertain significance in humans.
  • Category B does not imply greater safety than Category C.
• Category C: due to their pharmacological mechanism of action, may cause harmful (possibly reversible) effects on the fetus or neonate, without causing malformations.
• Category D: has caused, is suspected to cause, or may be expected to cause an increased incidence of fetal malformation or irreversible damage (e.g. many anticonvulsants); not absolutely contraindicated in pregnancy.

Post-marketing surveillance

Post-marketing surveillance is used to identify rare adverse drug effects [the lecture’s “why” slide carries only this title, with no further body content transcribed]. In New Zealand this is coordinated by CARM (the Centre for Adverse Reactions Monitoring), part of the NZ Pharmacovigilance Centre (NZPhvC) based at the University of Otago. Suspected adverse reactions are reported on a standard form capturing:

• patient details (name/initials, date of birth, sex, pregnancy status, NHI number, weight/height, ethnicity)
• all medicines/vaccines in use (name, dose, route, start/stop dates, indication)
• a free-text description of the adverse reaction
• outcome (recovered, recovering, not recovered, recovered with sequelae, fatal, unknown)
• seriousness (death, life-threatening, hospitalisation/prolonged hospitalisation, disabling/incapacitating, congenital abnormality/birth defect, other medically important condition)
• other relevant information
• reporter details (health professional type, contact details, DHB/PHO, date)

Pharmacoepidemiology

Pharmacoepidemiology connects to five main goals, all of which feed forward into clinical and health-policy decision-making:

• drug effectiveness (by condition, comorbidity or population)
• patterns of drug use (including off-label use, by condition or population)
• drug-drug interactions (identifying unknown interactions and their clinical relevance)
• pharmacovigilance (identifying unknown, rare, serious and long-term adverse drug reactions)
• pharmacoeconomy (cost-effectiveness, cost-benefit and cost-utility)

Within pharmacoepidemiology, cohort and case-control studies (the two analytical designs highlighted) have complementary strengths and limitations:

• Cohort studies: suited to long-term drug effects; can assess multiple exposures and outcomes at once; require a large sample size and an extended study period; not useful for rare outcomes or diseases; can draw on a wide range of data sources.
• Case-control studies: suited to rare outcomes/diseases and those with long latency periods; accurate selection of controls is a challenge and cases can be hard to find; draw on a narrower range of data sources than cohort studies. [flag: the specific numbered list of data sources referenced by this table was not legible on the slide]

Two New Zealand examples of applied pharmacoepidemiology:

• Antidepressant dispensing in pregnancy (Donald et al., 2021): using the New Zealand Pregnancy Cohort (805,990 pregnancies), dispensing records from 270 days before pregnancy to 360 days after pregnancy end were linked to pregnancy dates. Antidepressant dispensing was lower in the first trimester than before or after pregnancy, and fell further in later trimesters. The proportion of pregnancies with at least one antidepressant dispensing rose from 3.1% to 4.9% over the study years, and around 80% of those dispensed received a selective serotonin reuptake inhibitor (SSRI).
• Dual long-acting bronchodilator therapy and acute coronary syndrome (ACS) risk in COPD (Parkin et al.): a nationwide retrospective cohort with a nested case-control (risk-set sampling) design, using routinely collected national health and dispensing data. A cohort of 83,417 patients aged over 45 who started long-acting bronchodilator therapy (LAMA and/or LABA) for COPD between February 2006 and December 2013 yielded 5,399 ACS cases; for each case, up to 10 controls were matched by date of birth, sex, cohort-entry date and COPD severity. Current use of combined LAMA and LABA therapy was associated with a higher risk of ACS than LAMA alone (odds ratio 1.28, 95% CI 1.13-1.44), an association that was larger restricted to initial users of therapy (rate ratio approximately 1.46, 95% CI 1.08-1.91) [flag: some follow-up and confidence-interval figures were partly illegible]. In real-world practice, using two long-acting bronchodilators rather than one was associated with about a 30% higher risk of ACS in this high cardiovascular-risk population.

Other important observational studies

Named as examples in three categories:

• Risk factors: Framingham Heart Study, British Doctors Study, Health Professionals Follow-up Study.
• Clinical management: cardiovascular risk assessment and treatment algorithms.
• Cross-sectional surveys: NZ Census, NZ Health Survey.

Self-test

1. List the six observational study designs presented in the lecture, and explain which design(s) may be either descriptive or analytical.
2. List the five clinical situations described as ones where observational studies are particularly useful.
3. Describe the 1981 MMWR case series of Pneumocystis carinii pneumonia in Los Angeles, and explain what it illustrates about the role of case series in medicine.
4. Describe the design of the 1983 case-control study of Kaposi’s sarcoma and PCP in homosexual men, and state the variable most strongly associated with illness.
5. Besides number of sexual partners, what other exposures were found to be associated with illness in this case-control study?
6. Why is it notable that interim AIDS prevention recommendations were already in place by March 1983?
7. Describe the Multicentre AIDS Cohort Study (MACS), and explain what kind of question a decades-long cohort study can answer that a case-control study cannot.
8. Describe the design and key finding of the AGEhIV cohort study.
9. Give two examples from the lecture of case reports or case series used to characterise emerging infectious disease outbreaks, and state one specific finding from each.
10. What value did the case report of the duodenal GIST in a 14-year-old boy add, given that standard investigations had already been negative?
11. Distinguish the Australian pregnancy drug-safety Categories A, B (including its subcategories), C and D.
12. Describe the purpose of post-marketing surveillance, name New Zealand’s scheme for it, and list the main categories of information collected on its report form.
13. Describe the five goals that pharmacoepidemiology connects to, and explain how they ultimately feed into decision-making.
14. Distinguish cohort and case-control study designs as used in pharmacoepidemiology, including which situations favour each and the practical limitation of each.
15. Describe the design and headline result of (a) the New Zealand study of antidepressant dispensing in pregnancy, and (b) the New Zealand study of dual long-acting bronchodilator therapy and ACS risk in COPD.
16. Using the history of AIDS investigation as an example, describe how observational study designs typically progress as understanding of a new disease develops.

Answers

Reveal answers

1. Case study, case series, ecological study, cross-sectional study/survey, case-control study, and cohort study (prospective or retrospective). Ecological studies (and, by the list’s structure, arguably the designs below it) may be either descriptive or analytical.
2. New diseases, very rare conditions, post-marketing surveillance (rare adverse effects), drug safety in pregnancy, and identifying risk factors.
3. Five previously healthy young homosexual men in Los Angeles were treated for biopsy-confirmed P. carinii pneumonia between October 1980 and May 1981, all with CMV infection and candidal mucosal infection; two died. It illustrates how a small case series of an unusual clustering of illness can flag a new disease before its cause is known.
4. A case-control study across four US cities compared 50 cases (Kaposi’s sarcoma and/or PCP) with 120 matched homosexual male controls. The variable most strongly associated with illness was the median number of male sexual partners per year (61 for cases versus 27-25 for controls).
5. Exposure to faeces during sex, history of syphilis and non-B hepatitis, treatment for enteric parasites, and use of various illicit substances.
6. Because the causative retrovirus (HIV) was not identified until May 1983, after the recommendations were issued, showing that risk-factor evidence from observational studies can guide public health action before a disease’s cause is confirmed.
7. MACS is a cohort of about 6,972 men who have sex with men, initiated in 1983 and still running over 35 years later. A long cohort like this can describe the natural history of untreated and treated HIV infection over decades, which a one-off case-control study cannot.
8. A prospective cohort of 596 HIV-positive and 550 HIV-negative participants aged 45+ followed for a median of 5.9 years. Comorbidity burden increased at a similar rate in both groups, but each additional comorbidity substantially raised mortality risk, and HIV-positive participants had a greater increase in DALYs over time.
9. The first US case of 2019 novel coronavirus (NEJM 2020), which described diagnosis and clinical course including progression to pneumonia by day 9; and the case series of 46 US human A(H5N1) avian influenza cases (2024), mostly exposed to poultry or dairy cattle, in which illness was uniformly mild with conjunctivitis the commonest symptom.
10. It documented a genuine, rare paediatric case of a tumour (GIST) usually seen in adults, arising in a site (fourth part of duodenum) and presenting with recurrent bleeding that four separate standard investigations had failed to identify, adding to the very limited literature on this presentation.
11. Category A: no proven increase in fetal harm despite wide use. Category B: human data lacking, so subcategorised by animal data alone (B1 no harm in animal studies, B2 inadequate/lacking animal data but no harm observed, B3 animal studies show harm of uncertain human significance); Category B does not imply greater safety than C. Category C: may cause reversible fetal/neonatal harm via the drug’s pharmacological action, without malformation. Category D: causes or is expected to cause fetal malformation or irreversible damage, though not absolutely contraindicated.
12. Post-marketing surveillance exists to catch rare adverse drug effects not detectable in pre-marketing trials. In New Zealand this is done through CARM. Its report form collects patient details, all medicines in use, a description of the reaction, outcome, seriousness, other relevant information, and reporter details.
13. Drug effectiveness, patterns of use, drug-drug interactions, pharmacovigilance, and pharmacoeconomy; all five feed forward into clinical and health-policy decision-making.
14. Cohort studies suit long-term drug effects and can assess multiple exposures/outcomes, but need a large sample and long study period and are not useful for rare outcomes. Case-control studies suit rare outcomes or those with long latency, but accurate control selection is difficult and cases can be hard to find.
15. (a) A New Zealand cohort study linking dispensing records to over 800,000 pregnancies found antidepressant dispensing fell in the first trimester and rose over the study years from 3.1% to 4.9% of pregnancies, mostly SSRIs. (b) A nationwide retrospective cohort/case-control study of over 83,000 COPD patients found that dual LAMA and LABA therapy was associated with a higher risk of acute coronary syndrome than LAMA alone (OR 1.28), roughly a 30% increased risk in real-world practice.
16. It typically moves from case reports/case series flagging an unusual clustering of illness, to case-control studies identifying risk factors (sometimes before a cause is confirmed), to long-running cohort studies establishing the natural history and long-term outcomes of the disease.