Breast cancer concerns in a call centre: cluster or coincidence?
A call centre employs 420 staff, 310 of them women aged 25–64. Over 4 years, six women have been diagnosed with breast cancer; one has died. The union has written to the chief executive citing 'a cancer cluster caused by the building's air conditioning' and demanding an independent inquiry. The local newspaper has run a story. The local age-standardised breast cancer incidence for women aged 25–64 is 250 per 100,000 per year. The company asks you, as contracted occupational physician, to advise. There is no record of any occupational carcinogen exposure; the building has a standard HVAC system.
1. Is this a cluster? Show your reasoning.
Read the worked answer
- First calculate the expected number. With 310 women over 4 years and a local rate of 250 per 100,000 per year: expected = 310 × 4 × 0.0025 = 3.1 cases.
- Six observed against 3.1 expected gives a ratio of about 1.9. Using a Poisson model with mean 3.1, the probability of observing six or more is about 0.09 - not statistically significant at conventional levels.
- The numbers are small, so random variation is wide. Case ascertainment may also be incomplete or biased: women may have moved, left, or been diagnosed elsewhere.
- A true cluster requires more than a numerical excess: cases should be unusual in type, age or time, and there should be a plausible exposure pathway. Breast cancer has a long latency and multiple hormonal and lifestyle risk factors; HVAC systems are not a recognised cause.
- Conclusion: this does not meet the criteria for a confirmed occupational cluster, but it is a real concern that requires a proportionate, transparent response, not dismissal.
2. What investigation, if any, would you advise, and what are the main methodological pitfalls?
Read the worked answer
- Start with case ascertainment: confirm diagnoses and dates from medical records or the cancer registry, and define the denominator (current and former staff, ages, employment dates).
- Compare observed cases with expected cases from the regional cancer registry, age-standardised. Consider a retrospective cohort using payroll records, with person-years at risk.
- Examine exposure: review building records, HVAC maintenance, any known carcinogens, and job tasks. If there is no plausible exposure, extensive environmental sampling is unlikely to be informative and may raise anxiety.
- Pitfalls: small numbers, multiple comparisons, healthy worker effect, latency, incomplete follow-up, and confounding by reproductive history, alcohol, obesity and screening uptake. Media attention can increase ascertainment.
- Recommend a limited, protocol-driven review with external public health input, rather than an open-ended inquiry. Agree in advance what would count as a signal.
3. How would you communicate with staff, the union and the media?
Read the worked answer
- Acknowledge the concern openly and early. Saying 'there is no cluster' before explaining the data will be heard as a cover-up.
- Share the calculation in plain language: six cases were diagnosed where about three would be expected; this is not statistically significant, but it is being taken seriously.
- Explain the uncertainty: small numbers mean we cannot exclude a small excess, and breast cancer is common in this age group.
- Describe what will be done, by whom and when, and commit to reporting back. Involve the local director of public health and the cancer registry; independent oversight helps credibility.
- Brief a single spokesperson, avoid speculation about the air conditioning, and provide staff with a route to ask questions and access support. Keep a written record of the communication.