Skip to content
PaperFren

Research method

Cox Proportional Hazards Model

The Cox proportional-hazards model estimates how a covariate shifts the instantaneous rate of an event — death, cancer recurrence, incident depression, a cardiovascular composite — without having to specify the baseline hazard. The headline number is a hazard ratio: HR 0.80 means about 20% lower hazard in that group, holding the model's other covariates fixed, provided the proportional-hazards assumption is close enough to true. Time-to-event methods also let investigators keep people who are censored (lost to follow-up, end of study) in the risk set until they leave.

Researchers reach for Cox models when the outcome is not a yes/no at a fixed week but whether and when something happens in a cohort or a trial. It answers 'is this exposure associated with a different event rate over follow-up?' Its main limitation is that a hazard ratio from an observational cohort is not a randomised treatment effect, and when events are rare even a large trial can return an HR near 1 that is hard to read as proof of equivalence.

Evidence

What the evidence shows

Drawn from 16 studies in this library. Each finding starts with a plain-language takeaway, then the denser detail. Supports means evidence for a finding; Challenges means evidence against a stated position; Qualifies marks scope with a short note on each study’s contribution. Challenged positions are labeled — they are not findings.

  • In observational cancer and sleep cohorts, Cox models turn deprivation, education and disease severity into excess-hazard or event-rate contrasts. ONS Longitudinal Study analyses found less-deprived areas had lower excess hazard for prostate cancer (EHR 0.80) and that absolute net-survival gaps between extreme individual SES profiles were 29.6% in the most deprived areas versus 9.1% in the least; an OSA sleep-lab cohort of 10,149 patients had 11.5% composite events over a median 68 months, with higher AHI linked to higher hazard (e.g. univariate HR 1.49 for high versus low AHI).

    2 studies
    1. 1Individual SES and cancer survival
    2. 2OSA severity and cardiovascular risk

    Study comparison

    StudyRoleDesignNPopulationOutcome
    Individual SES and cancer survival2022SupportsCohortONS Longitudinal Study; excess mortality models by SES markersN=9276 · 1522 men + 1237 women colorectal; 3044 prostate; 3473 breast (ONS Longitudinal Study analysis cohort)People with colorectal, prostate, or breast cancer in the ONS Longitudinal StudyExcess mortality / net-survival gaps by area deprivation and individual SES
    OSA severity and cardiovascular risk2014SupportsCohortDiagnostic polysomnography cohort linked to administrative CV outcomesN=10149 · 10,149 of 11,596 first diagnostic sleep studies linked to admin data (88%)Adults undergoing diagnostic polysomnographyComposite cardiovascular events and mortality
  • Diet and alcohol papers in this set report HRs for incident depression or CVD that are threshold-like or beverage-specific rather than strictly linear. Low-to-moderate alcohol (>5–15 g/day) had HR 0.72 for incident depression among 5,505 older Mediterranean adults; each extra 10 g/day olive oil in PREDIMED participants had HR 0.87 for major CVD events; higher diet-quality quintiles had depression HRs around 0.74–0.78.

    3 studies
    1. 1Wine, alcohol, and depression
    2. 2Olive oil intake and CVD risk
    3. 3Diet quality and depression risk
  • A randomised trial can still use Cox (or a closely related survival contrast) as the primary analysis. Patients undergoing curative breast-cancer surgery randomised to propofol versus sevoflurane had five-year overall survival around 92% in both arms, with a hazard ratio near 1 and no statistically significant difference in intention-to-treat or per-protocol analyses.

    1 study
    1. 1Propofol vs sevoflurane and breast cancer survival
  • When depression is not the trial's original primary endpoint, a Cox HR below 1 is easy to over-read. PREDIMED's merged Mediterranean-diet HR for incident depression was 0.85 and not significant; a nuts-arm benefit appeared in a diabetes subgroup after 224 cases over a median 5.4 years.

    1 study
    1. 1Mediterranean diet and depression risk

Open questions

Tensions and limits

Some items are genuine disagreements on the same question. Others mark different assays, populations, or outcomes — limits on how far one study travels — not a forced fight between papers.

  • Scope / different questions

    The same modelling family is asked to do two different jobs: describe inequalities and exposures in cohorts, versus test a randomised intervention. The SES-cancer and OSA papers cannot name a single causal mechanism or a treatment that would close the HR gap, whereas the propofol–sevoflurane trial can say the assigned anaesthetic did not shift 5-year survival under its protocol. Reading every HR as if it were that trial overclaims the observational papers and underclaims the trial's design.

    3 studies
    1. 1Individual SES and cancer survival
    2. 2OSA severity and cardiovascular risk
    3. 3Propofol vs sevoflurane and breast cancer survival

    Study comparison

    StudyRoleDesignNPopulationOutcome
    Individual SES and cancer survival2022SupportsCohortONS Longitudinal Study; excess mortality models by SES markersN=9276 · 1522 men + 1237 women colorectal; 3044 prostate; 3473 breast (ONS Longitudinal Study analysis cohort)People with colorectal, prostate, or breast cancer in the ONS Longitudinal StudyExcess mortality / net-survival gaps by area deprivation and individual SES
    OSA severity and cardiovascular risk2014SupportsCohortDiagnostic polysomnography cohort linked to administrative CV outcomesN=10149 · 10,149 of 11,596 first diagnostic sleep studies linked to admin data (88%)Adults undergoing diagnostic polysomnographyComposite cardiovascular events and mortality
    Propofol vs sevoflurane and breast cancer survival2023SupportsRCT1:1 propofol vs sevoflurane for anaesthesia maintenance in curative primary breast cancer surgeryN=1670 · ITT: 841 propofol, 829 sevofluranePatients undergoing curative primary breast cancer surgery5-year overall survival
  • Scope / different questions

    Null and protective HRs are not interchangeable with 'safe' or 'preventive.' The anaesthetic trial's HR near 1 is limited by rare mortality and differing opioid co-analgesia; the alcohol paper's HR 0.72 cannot prove drinking prevents depression; the antidepressant-adverse-event cohort links several drugs to more falls versus non-use, which may still be confounding by indication.

    3 studies
    1. 1Propofol vs sevoflurane and breast cancer survival
    2. 2Wine, alcohol, and depression
    3. 3Antidepressant adverse events cohort

    Study comparison

    StudyRoleDesignNPopulationOutcome
    Propofol vs sevoflurane and breast cancer survival2023SupportsRCT1:1 propofol vs sevoflurane for anaesthesia maintenance in curative primary breast cancer surgeryN=1670 · ITT: 841 propofol, 829 sevofluranePatients undergoing curative primary breast cancer surgery5-year overall survival
    Wine, alcohol, and depression2013SupportsCohortObservational analysis within PREDIMED participantsN=5505 · Older Mediterranean adults; 443 incident depression casesOlder adults in the PREDIMED Mediterranean cohortIncident depression by baseline alcohol/wine intake (Cox HRs)
    Antidepressant adverse events cohort2018SupportsCohortUK primary-care database; newly diagnosed depression followed for adverse eventsN=238963 · Final eligible cohort; 87.7% received antidepressants during follow-upAdults aged 20–64 with newly diagnosed depression in UK primary careAdverse outcomes (including falls) by antidepressant class/drug vs non-use

Common misconceptions

  • A hazard ratio is the same thing as a relative risk at the end of follow-up.

    It is a ratio of instantaneous event rates over time, under a proportional-hazards assumption. Absolute gaps can look very different from the HR, as in the cancer-survival paper's 29.6% versus 9.1% net-survival gaps by SES profile despite a prostate EHR of 0.80 for area deprivation.

    1. 1Individual SES and cancer survival
  • If a large trial's HR is about 1, the two treatments have been proven equivalent.

    The propofol versus sevoflurane breast-cancer trial analysed 1,670 patients and still could not prove equivalence for a rare mortality outcome; opioid co-analgesia also differed by arm.

    1. 1Propofol vs sevoflurane and breast cancer survival
  • An HR below 1 from a food or alcohol cohort means that exposure was shown to prevent the disease.

    Those analyses remain observational (olive oil within a trial cohort; alcohol and diet-quality scores in follow-up studies) and the PREDIMED depression analysis was not even the trial's original primary endpoint.

    1. 1Olive oil intake and CVD risk
    2. 2Wine, alcohol, and depression
    3. 3Mediterranean diet and depression risk

Exam-style questions

Short-answer questions that ask you to explain or compare, not recall.

What does a Cox model add that a simple chi-square of 'event yes/no' at the end of a study does not, and why does that matter for the OSA cohort?

It uses time to the event and keeps censored people in the risk set until they leave, so a 68-month median follow-up with 11.5% composite events can still estimate how AHI shifts the hazard rather than collapsing everyone to a yes/no at a common calendar date. People diagnosed later or lost sooner still contribute follow-up time.

Contrast the propofol–sevoflurane HR near 1 with the olive-oil HR of 0.87. Which is a randomised treatment contrast, and what is each number still not allowed to claim?

Anaesthetic assignment was randomised; olive-oil intake was an observational contrast inside PREDIMED. The trial HR near 1 does not prove equivalence for rare deaths or for cancers other than primary breast surgery. The olive-oil HR does not mean olive oil was itself randomised, and residual confounding remains possible.

Why is PREDIMED's merged Mediterranean-diet HR of 0.85 for depression a weak basis for a dietary prescription against depression?

Depression was not the original primary endpoint, the merged-diet HR was not significant, and the clearer nuts-arm signal was a diabetes subgroup among 224 cases. Subgroup HRs after a non-significant main contrast are easy to overfit.

How can a 'protective' alcohol HR of 0.72 and an antidepressant-falls signal both be true in Cox models without telling a student what to drink or which pill to avoid?

Both are observational rate ratios. Moderate alcohol may mark a healthy-user pattern rather than a causal antidepressant; extra falls on named drugs may reflect who is prescribed them (confounding by indication). Cox organises the time-to-event comparison; it does not by itself identify the mechanism or the right clinical action.

The studies

16 studies in this library bear on Cox Proportional Hazards Model, ordered by citations. The first 8 are shown.

  • Night shifts and type 2 diabetes risk

    Longer rotating night-shift work associated with higher type 2 diabetes risk in two large female nurse cohorts.

    PLoS medicine · 2011 · 567 citations

  • OSA severity and cardiovascular risk

    In >10,000 sleep-study patients, higher apnea–hypopnea burden predicted cardiovascular events and death over years of follow-up.

    PLoS medicine · 2014 · 311 citations

  • Healthy lifestyle and multimorbidity risk

    Healthier lifestyle scores tracked lower CVD and diabetes risk and lower multimorbidity transitions.

    BMC medicine · 2020 · 268 citations

  • Mediterranean diet and depression risk

    Overall Mediterranean diet assignment did not significantly reduce depression; a nuts arm suggested benefit in diabetes subgroup analyses.

    BMC medicine · 2013 · 232 citations

  • Olive oil intake and CVD risk

    Higher olive oil intake tracked lower major CVD events and cardiovascular death in PREDIMED participants.

    BMC medicine · 2014 · 228 citations

  • How broadly does smoking raise heart and vessel disease risk?

    In a large Australian cohort, current smoking raised risk across nearly all CVD subtypes—especially peripheral arterial disease—and quitting lowered risk.

    BMC medicine · 2019 · 198 citations

  • Shingles vaccine cut zoster by about half

    In older US adults, herpes zoster vaccination showed ~48% effectiveness against incident zoster and ~59–62% against post-herpetic neuralgia.

    PLoS medicine · 2013 · 160 citations

  • Diet quality and depression risk

    In the SUN cohort, higher adherence to Mediterranean, pro-vegetarian, and AHEI-2010 diet scores tracked with lower risk of incident depression over about 8.5 years.

    BMC medicine · 2015 · 118 citations

Show 8 more studies
  • Wine, alcohol, and depression

    In PREDIMED adults, low-to-moderate alcohol—especially wine—associated with lower incident depression risk versus abstaining, while heavy drinking looked riskier.

    BMC medicine · 2013 · 93 citations

  • Antidepressant adverse events cohort

    In 238,963 adults aged 20–64 with depression, most used antidepressants and several common agents associated with higher fall rates versus non-use.

    BMC medicine · 2018 · 67 citations

  • Widening NSW cancer survival gaps

    Among 651,245 NSW cancer cases, people in more disadvantaged areas had higher cancer death risk, and disparities appeared to widen over time.

    BMC public health · 2017 · 60 citations

  • ANXA1 in breast cancer prognosis

    ANXA1-positive breast tumors clustered with poor-prognosis features (high grade, basal/TN, BRCA1/2) and worse outcomes in some high-risk subgroups such as HER2+.

    BMC medicine · 2015 · 47 citations

  • Does socioeconomic status change cancer survival?

    In Norwegian women, lower education and income predicted worse cancer survival until stage and smoking were accounted for.

    BMC public health · 2009 · 41 citations

  • Individual SES and cancer survival

    UK linked-data analysis shows both area deprivation and individual SES shape cancer survival, with wider absolute gaps in the most deprived areas.

    BMC public health · 2022 · 38 citations

  • Propofol vs sevoflurane and breast cancer survival

    In the CAN randomised trial, five-year overall survival after primary breast cancer surgery was essentially the same with propofol or sevoflurane maintenance anaesthesia.

    EClinicalMedicine · 2023 · 37 citations

  • Screening and breast cancer equity in NZ

    Screen-detected cancers were less common in Māori women, and crude survival was worse, but survival gaps narrowed among screen-detected cases.

    BMC public health · 2015 · 33 citations

Learn alongside