About the Geneva Cancer Registry
The
Geneva Cancer Registry
is a population-based cancer registry that has been systematically recording all
incident cancer cases
among residents of the Canton of Geneva for more than 50 years.
It consolidates information from hospitals, pathology laboratories, death certificates and clinical records to produce high-quality, validated data on cancer incidence, stage at diagnosis, treatment proxies and survival outcomes.
These data feed
regional, national and international
surveillance and research networks, and support public health planning, evaluation of screening programs and monitoring of inequalities in cancer burden.
About this dashboard
Information
This dashboard provides a concise, population-level overview of cancer incidence and mortality in the Canton of Geneva. It is intended for public health monitoring, epidemiological research and general information purposes, and does not replace official registry publications, professional medical advice, or clinical decision-making.
Multiple primary rules
Since April 2026, incidence figures are calculated after applying IARC multiple primary rules: when a patient develops several tumours at the same organ or histological group, only the first is counted. This brings the registry into line with international reporting standards and may result in slightly lower counts compared to previous editions.
References and supporting file:
IARC/IACR 2004 Multiple Primary Rules (PDF)
;
Updated table (ICD-O-3.2 morphology groups for multiple tumours)
.
Impact of multiple primary rules on incidence reporting:
see which tumor types and incidence years are impacted
impact_reporting_sur_incidence.xlsx
.
Identification procedure for multiple primary tumours
The dashboard applies the following sequence, aligned with IARC/IACR 2004 rules and ICD-O updates:
- Assign each tumour to a topography group (ICD-O-3 Table 1). Some sites are grouped (for example paired organs), so left and right are considered one group.
- Assign each tumour to a morphology group (ICD-O-3 Table 2). Groups marked systemic (haematological and Kaposi sarcoma) follow a dedicated counting rule.
- Apply exclusion rules: for solid tumours, keep one tumour per patient x topography group x morphology group (earliest diagnosis kept); for systemic tumours, keep one tumour per patient x morphology group across all sites.
- Apply NOS rule: when a patient has both a specific morphology and a NOS morphology in the same topography group, the NOS tumour is excluded.
- Validate final counts against IARCcrgTools outputs; discordant records are reviewed and reconciled.
This ensures that each independent primary tumour is counted once and only once.
Multiple tumour filter table
:
Link to iarc multiple primary rule view
Quality, Updates & Scope
To avoid potential confusion, only the most recent and most reliable statistics available are displayed.Data, methods, and definitions may change over time (updates, corrections, harmonization) as part of a continuous improvement process.
Data Sources
Data Sources
Incidence - Geneva Cancer Registry: Incident cancer cases are provided by the Geneva Cancer Registry, which records all new malignant tumours diagnosed among residents of the Canton of Geneva. Cases are identified through systematic notification from hospitals, pathology laboratories, death certificates and clinical records.
Mortality - Swiss Federal Statistical Office (FSO): Mortality data are provided by the Swiss Federal Statistical Office (FSO), based on the underlying cause of death as coded on death certificates. These data are independent from registry records and cover all deaths occurring among Geneva residents.
Population Data - Swiss Federal Statistical Office (FSO): Population data are sourced from Stat-Tab, the interactive tables database of the FSO, extracted from the demographic balance table (Bilan demographique selon l'age et le canton). For each year, sex and age group, the population estimate used is the average of the January 1st and December 31st figures, providing a mid-year estimate more representative for rate calculations.
Harmonisation: Incidence and mortality datasets are harmonised to allow consistent comparison by tumour site, sex, age group and calendar year.
Indicators
The following indicators are computed and displayed across the dashboard visualisations.
Incidence
- Number of incident cases: count of new tumours recorded by the Geneva Cancer Registry.
- Crude incidence rate: number of incident cases per 100,000 person-years.
- Age-standardised incidence rate (ASR): incidence rate per 100,000 person-years, directly standardised to the European reference population (ESP 1976), to allow comparison across populations and over time irrespective of age structure.
-
Identified multiple primary cancers were removed from analyses (references:
IARC multiple primary rules, 2004
;
updated ICD-O table
).
Mortality
- Number of deaths: count of cancer deaths with the relevant tumour as underlying cause, provided by FSO.
- Crude mortality rate: number of deaths per 100,000 person-years.
- Age-standardised mortality rate (ASR): mortality rate per 100,000 person-years, directly standardised to the same reference population as for incidence.
Note:
These indicators are displayed only if relevant to the given graph type. For example, all indicators (counts, crude rates, and ASR) are displayed in bar charts, while only counts are displayed in treemaps.
Data visualisation
This section groups all dashboard graphs by analytical family and explains, for each visual, both its public-health interest and how values are computed and displayed. Titles are clickable: from Information you can open each corresponding graph directly, and from each graph you can return here using the Description link.
Most frequent cancers
The Bars chart is designed for ranking and comparison. It shows mean annual counts and the selected rate indicator for each cancer site over the selected time interval, with optional Top-N filtering. This makes it easy to identify high-burden sites while preserving comparability across sexes, periods and indicator types.
Computation follows the dashboard indicator framework (counts, crude rates and ASR when available), using registry/FSO numerators and person-years denominators. The chart is particularly useful for prioritisation because it combines absolute burden and relative risk level in one view.
The Treemap summarizes composition: each tile area is proportional to the selected burden metric for a cancer site, so large tiles correspond to major contributors and small tiles reveal the long tail. It is intended for rapid visual reading of proportional structure rather than precise value comparison.
Because area perception is less precise than aligned bars, this graph complements the Bars tab: treemap for structure at a glance, bars for exact ranking. Filters for period, sex and data type apply consistently.
Privacy rule (Treemap, public mode):
a site is merged into
Others
only when it has fewer than 5 total cases over the selected period for the selected sex.
Annual averages (e.g. cases per year) do not trigger suppression by themselves.
The Relative distribution graph displays how the composition of cancer sites changes over time. For each selected year, the chart shows the percentage contribution of each site to the annual total (incidence or mortality).
Privacy rule (cancer sites):
If a cancer site has between 0 and 4 cases in any year of the selected period for the selected sex, that site is merged into the category
Others
for all years in the period: it is no longer shown as its own stratum, but its cases are included in the
Others
counts and percentage share in the chart and table (stacked area for the chart). Years with zero cases for that site do not trigger this rule.
The Top N sites are ranked by frequency in the latest year available in the selected range. In public mode, these Top N sites remain displayed as separate strata even if they have 1–4 cases in another year of the selected period. All remaining sites are grouped under
Others
together with any sites merged for privacy. Users can also choose to show all remaining sites without a Top-N cap.
In the downloadable table, counts below 5 in any cell are shown as an em dash (—) in public mode, consistent with other dashboard tables.
Age analysis
Age distribution describes how burden is distributed across age groups for one selected cancer site and period. It combines age-specific counts and rates, helping distinguish cancers concentrated at younger ages from those dominated by older ages.
Rates are computed per age class from observed events and corresponding person-years. This graph is especially useful for prevention or screening interpretation, where age concentration is often as informative as total burden.
Privacy rule applied in this graph (public mode): age groups with fewer than 5 observed cases over the full selected period are suppressed in the chart (bar omitted) and shown as an em dash (—) in the table.
Median age (heatmap) compares age-at-diagnosis or age-at-death across cancer sites and 5-year periods in a compact matrix view. Colour intensity maps median age values, making temporal drifts and cross-site contrasts immediately visible.
Only stable site-period combinations are shown (minimum-count rule), which limits noisy medians in sparse strata. This view is therefore suited to identify broad age-pattern shifts over time rather than isolated fluctuations.
The Age-rate profile chart (Age analysis tab) is an interactive scatter plot that positions each cancer site along two dimensions: median age at diagnosis (incidence) or at death (mortality) on the horizontal axis, and a user-selected epidemiological indicator on the vertical axis (annual mean counts, crude rate per 100,000, or age-standardised rate using ESP 1976). Each point is colour-coded by cancer type; optional lines connect a site across consecutive 5-year periods when several periods are selected.
Case definitions and denominators follow the same rules as the median age heatmap (IARC multiple primary rules; person-years for rates). The chart supports incidence or mortality, sex filters, and subsetting of sites and periods.
This view helps identify epidemiological clusters: high-burden cancers affecting younger ages, late-onset but rarer cancers, or sites that stand out from the overall cloud of points, supporting prioritisation and comparative reading alongside absolute trend analyses.
The Age-Period Bubble Profile chart (Age analysis tab) displays the evolution of median age at diagnosis (incidence) or at death (mortality) across successive 5-year periods for each selected cancer site. The horizontal axis represents the 5-year periods, the vertical axis represents median age, bubble colour identifies the cancer site, and bubble size reflects the selected epidemiological indicator (counts, crude rate per 100,000, or ASR per 100,000 using ESP 1976).
Like the Age-rate profile, this view uses the same underlying population-based registry data and indicator definitions, including IARC multiple primary rules and person-years denominators for rate calculations. The chart is filtered by data type (incidence or mortality), sex, 5-year periods and cancer sites, and supports direct cross-site comparison within each period and across time.
For consistency with the median age framework, only site-period combinations meeting the minimum-count criterion are displayed (at least 5 cases/deaths in the corresponding 5-year period for the selected sex, after multiple primary rules). This avoids unstable median-age and rate summaries in very sparse strata and keeps interpretation focused on robust epidemiological signals.
Trends
Evolution displays medium-term trends using non-overlapping multi-year periods to smooth annual noise. It is intended for descriptive trend reading and comparison of direction and pace between sites.
By reducing year-to-year volatility, this chart helps detect sustained increases or decreases that can be masked in annual series, while preserving interpretable period labels.
For methodological details, see
Evolution
in the Statistical analysis methods section.
Trend change uses annual data and segmented regression to identify possible breakpoints in temporal trajectories. It is the inferential complement to Evolution: instead of only showing smoothed curves, it tests whether slope changes are statistically plausible.
This tab supports interpretation of transitions (acceleration, deceleration, reversal) and provides model-based quantities such as segment APCs and goodness-of-fit criteria.
For methodological details, see
Trend Change Regression Analysis (Joinpoint)
,
Summary: Evolution vs. Trend Change (Joinpoint)
and
Annual Percent Change (APC)
in the Statistical analysis methods section.
Annual Percent Change (APC) condenses the difference between two selected periods into an annualised percentage change. It is useful when users need a single synthetic measure to compare recent trend direction and magnitude across sites.
The graph highlights increases, decreases and non-significant changes based on confidence intervals, which supports quick interpretation while keeping statistical uncertainty explicit.
For methodological details, see
Annual Percent Change (APC)
in the Statistical analysis methods section.
Dynamic evolution animates year-by-year changes, allowing users to follow how sites move in relative position over time. It is designed for pattern discovery: rank shifts, convergences/divergences, and abrupt changes become visually salient in motion.
This view complements static trend charts by emphasizing temporal ordering and transitions rather than only start/end differences.
Population
Age pyramid shows the population age structure by sex for the selected year or period context. It is a denominator-focused chart: it helps interpret why crude rates can change when population ageing shifts the at-risk structure.
Reading this graph alongside cancer indicators improves interpretation of demographic effects versus true disease-risk changes.
Global evolution tracks total population over time, optionally by sex. It provides long-run demographic context for incidence/mortality trends, especially when interpreting periods of growth, stagnation or decline in the denominator.
This graph should be read with trend tabs to distinguish epidemiological change from population-volume change.
Evolution by age category decomposes demographic change across age groups, highlighting where growth or contraction occurs in the population pyramid over time.
It is particularly informative for cancer surveillance because many sites are age-sensitive: shifts in older age-group denominators can materially affect crude trends.
Cancer dictionary
List of the tumour categories generated for the dashboard and their corresponding International Classification of Disease (ICD, 10th revision, version 2010) codes.
Statistical analysis methods
Confidence Intervals for Rates
95% confidence intervals (CI) are calculated for both crude rates and age-standardised rates (ASR) to quantify uncertainty around point estimates. These intervals are displayed in downloadable result tables when relevant.
Crude Rate Confidence Intervals
For crude rates, 95% confidence intervals are calculated using the exact Poisson method based on the chi-square distribution:
Formula:
For K = 0 cases:
Lower bound = 0
Upper bound = [χ²(0.975, df=2) / (2×P)] × 100,000
For K > 0 cases:
Lower bound = [χ²(0.025, df=2K) / (2×P)] × 100,000
Upper bound = [χ²(0.975, df=2(K+1)) / (2×P)] × 100,000
where:
- K = total number of cases (or deaths) in the selected period
- P = total person-years at risk in the selected period
- χ²(p, df) = p-th quantile of the chi-square distribution with df degrees of freedom
Age-Standardised Rate (ASR) Confidence Intervals
For age-standardised rates, 95% confidence intervals are calculated using the Gamma distribution method proposed by Fay and Feuer (1997), which is robust for small numbers of cases:
Formula:
For each age group i:
aᵢ = (100,000 × wᵢ) / (Σⱼ wⱼ × Pᵢ)
ASR = Σᵢ (aᵢ × kᵢ)
Var(ASR) = Σᵢ (aᵢ² × kᵢ)
95% CI bounds = Gamma(α/2 or 1-α/2, shape = ASR²/Var, scale = Var/ASR)
where:
- kᵢ = number of cases (or deaths) in age group i
- Pᵢ = person-years at risk in age group i
- wᵢ = weight for age group i from the European Standard Population (ESP 1976)
- α = 0.05 (for 95% confidence intervals)
Reference:
Fay MP, Feuer EJ. Confidence intervals for directly standardized rates: a method based on the gamma distribution. Statistics in Medicine. 1997;16(7):791-801. doi: 10.1002/(SICI)1097-0258(19970415)16:7<791::AID-SIM500>3.0.CO;2-#
Age distribution
The age distribution tab displays age-specific incidence or mortality rates for a given cancer site and time period. For each 5-year age group (0-4, 5-9, ..., 85+), the age-specific rate is calculated as:
Age-specific rate:
Rate(i) = (Cases(i) / Population(i)) × 100,000
where:
- Cases(i) = total number of incident cases (or deaths) in age group i over the selected period
- Population(i) = total person-years in age group i over the selected period (mid-year population summed across years)
Bar charts display counts (bars) and age specific rates (lines).
Data privacy
To protect patient confidentiality, age groups with fewer than 5 observed cases over the entire selected period are handled as follows:
- In charts: bars for suppressed age groups are omitted entirely (no bar is drawn). The age group label remains visible on the x-axis.
- In tables: suppressed counts are replaced by — (em dash), following the convention used in official epidemiological publications (e.g. NICER, Eurostat, WHO).
- Annual averages below 5 (e.g. 1.6 cases/year) are not suppressed: they reflect small but real counts spread over multiple years (e.g. 8 cases over 5 years = 1.6/year) and are therefore displayed as is.
- The same suppression rules apply to all tabs displaying age-stratified data, including the Age distribution table.
Evolution
Time trend with non-overlapping 3-year periods
The Evolution tab displays temporal trends in incidence or mortality rates using non-overlapping 3-year periods. This approach smooths out year-to-year random fluctuations while preserving the overall trend.
Method:
- The selected year range is divided into consecutive non-overlapping 3-year windows: [Y, Y+1, Y+2], [Y+3, Y+4, Y+5], etc.
- For each 3-year window, all individual-level data (cases and population) are pooled across the 3 years.
- Crude rates and age-standardised rates (ASR) are computed on the pooled data using the same formulas as described in the Indicators section.
- Each data point is plotted at the midpoint year of the period (e.g., 1971 represents the 1970–1972 period).
Example:
For the year range 1970–2022, the periods would be: 1970–1972 (plotted at 1971), 1973–1975 (plotted at 1974), 1976–1978 (plotted at 1977), etc.
The user can choose to display the crude rate or the age-standardised rate (ASR). The graph shows one data point per 3-year period, connected by lines, for each selected cancer site.
Trend Change Regression Analysis (Joinpoint)
The Trend Change tab performs a segmented regression analysis to identify statistically significant changes in the trend of cancer rates over time. Unlike the Evolution tab (which uses 3-year periods), the Joinpoint analysis uses annual data to provide maximum temporal resolution for trend detection.
Eligibility rule used in this tab: a site is included when the official confidentiality threshold is mostly respected (yearly counts should be at least 5), with a tolerance of at most 3 years below 5 over the selected period (for the selected sex). Joinpoint estimates remain visible in tables and exports for interpretation.
1. Data preparation
For each calendar year, the dashboard computes:
-
Crude rate:
(Cases / Population) × 100,000, where cases and population are summed across all age groups.
-
Age-standardised rate (ASR):
Directly standardised rate using the European Standard Population (ESP 1976) weights, calculated with the same formula as in the Indicators section.
The user selects which rate measure to model (crude or ASR).
2. Model specification
Two different regression models are used depending on the rate measure selected:
For Crude Rates — Poisson Regression:
log(E[Cases]) = β₀ + β₁ × Year + log(Population)
This is a generalized linear model (GLM) with a Poisson family, log link, and log(Population) as an offset. This directly models the count process and accounts for the population at risk.
For ASR — Log-Linear Regression:
log(ASR) = β₀ + β₁ × Year
This is an ordinary least squares (OLS) linear regression on the log-transformed ASR. The log transformation ensures that predicted rates are always positive and that the model captures multiplicative changes in rates.
3. Segmented regression (joinpoints)
The dashboard fits piecewise linear models with 0, 1, 2, or 3 joinpoints (knots). A joinpoint is a year where the trend changes direction or rate. The
segmented
R package is used to estimate the location of joinpoints and the slopes of each segment.
Segmented model structure:
-
0 change point:
A single linear trend over the entire period (baseline model).
-
1 change point:
Two segments with different slopes, joined at one breakpoint. Example: an increase until 1995, then a decrease.
-
2 change points:
Three segments with different slopes, joined at two breakpoints.
-
3 change points:
Four segments with different slopes, joined at three breakpoints.
The joinpoints (breakpoints) are estimated by the segmented algorithm and are displayed as vertical dashed lines on the plot. Initial knot positions are set at evenly spaced quantiles of the year range.
4. Model selection — Preferred model (BIC)
The preferred model is automatically highlighted in the top-right legend and in the results table using the Bayesian Information Criterion (BIC). BIC balances goodness-of-fit against model complexity: adding knots may improve fit but is penalized for extra parameters. The model with the lowest BIC is highlighted by default, while users may still display models with other numbers of knots.
BIC Formulas:
For ASR (lm):
BIC = log(SSE/n) + 2·(k+2)·log(n)/n
BIC3 = log(SSE/n) + 3·(k+2)·log(n)/n
For Crude rates (Poisson GLM):
BIC = Deviance/n + 2·(k+2)·log(n)/n
BIC3 = Deviance/n + 3·(k+2)·log(n)/n
Weighted BIC (WBIC):
WBIC = (1-w)·BIC + w·BIC3
where:
- SSE = sum of squared residuals (for lm) or deviance (for GLM)
- n = number of data points (years)
- k = number of knots (joinpoints)
- w = weight (0 to 1) based on the magnitude of gradient changes between consecutive segments
★ Preferred model:
In both the table and the plot, the preferred model (lowest BIC) is indicated by a ★ symbol. In the table, the preferred model's rows are highlighted in green. On the plot, the preferred model's trend line is drawn thicker.
5. Annual Percent Change (APC) per segment
For each segment of the joinpoint model, the Annual Percent Change (APC) is calculated using the slope coefficient of that segment (Lambert et al. approach):
APC calculation:
For segmented models, slopes are extracted via segmented::slope(). For the base model (0 knots), the slope is the coefficient β₁ of year.
APC = 100 × [exp(β) - 1]
95% Confidence Interval:
APC_lower = 100 × [exp(β - 1.96 × SE(β)) - 1]
APC_upper = 100 × [exp(β + 1.96 × SE(β)) - 1]
where:
- β = estimated slope (on log scale) for the segment
- SE(β) = standard error of the slope estimate
- For segmented models, slopes and standard errors are obtained from segmented::slope()
6. Statistical significance
Two types of statistical tests are reported:
-
Model significance (F-statistic / p-value):
- For ASR models (lm): an F-test comparing the segmented model to a null model (intercept only). The F-statistic is computed as: F = [(RSS_null - RSS_model) / (df_null - df_model)] / [RSS_model / df_model].
- For crude rate models (Poisson GLM): a chi-square test based on the difference between null deviance and residual deviance.
-
Slope significance (per segment):
A t-test (for lm) or z-test (for GLM) for the slope of each segment. A p-value < 0.05 indicates a statistically significant trend in that segment.
7. Plot interpretation
The Joinpoint plot displays:
-
Black dots:
Observed annual rates (one point per year).
-
Colored lines:
Fitted trend lines from the selected joinpoint models. Different line styles (solid, dashed, dotted, dash-dot) distinguish between models with 0, 1, 2, and 3 knots.
-
Thicker line with ★:
The preferred model (lowest BIC) is drawn with a thicker line and marked with a ★ in the legend.
-
Log scale option:
The y-axis can be switched to logarithmic scale, which is useful for visualizing constant annual percent changes as straight lines.
References:
Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates. Statistics in Medicine. 2000;19(3):335-351. DOI:
10.1002%2F%28SICI%291097-0258%2820000215%2919%3A3%3C335%3A%3AAID-SIM336%3E3.0.CO%3B2-Z
Muggeo VMR. segmented: An R package to fit regression models with break-points / change-points estimation. R package version 2.1-0. 2024. Available at: https://CRAN.R-project.org/package=segmented
Clegg LX, Hankey BF, Tiwari R, Feuer EJ, Edwards BK. Estimating average annual per cent change in trend analysis. Statistics in Medicine. 2009;28(29):3670-82. DOI:
10.1002/sim.3733
National Cancer Institute. Joinpoint Regression Program, Version 4.9.0.0. Statistical Research and Applications Branch. Available at: https://surveillance.cancer.gov/joinpoint/
Limitations:
Joinpoint models assume that the rate changes linearly within each segment; if the underlying trend follows a non-linear trajectory (e.g. exponential growth, logarithmic saturation, or a smooth continuous curve), piecewise linear models may provide a poor approximation regardless of the number of joinpoints. Model selection via BIC identifies the best-fitting model among the linear alternatives considered, but this does not preclude that alternative approaches — such as spline regression or generalised additive models (GAMs) — might fit the data substantially better.
Summary: Evolution vs. Trend Change (Joinpoint)
| Feature |
Evolution Tab |
Trend Change Tab |
| Time resolution |
Non-overlapping 3-year periods |
Annual (1-year) |
| Smoothing |
Yes (3-year pooling) |
No (raw annual rates) |
| Statistical model |
Descriptive (no model fitted) |
Segmented regression (0–3 joinpoints) |
| Trend detection |
Visual only |
Statistical (APC + p-values + BIC) |
| Rate measures |
Crude rate, ASR |
Crude rate, ASR |
| Model selection |
N/A |
BIC (★ preferred model) |
Annual Percent Change (APC)
The Annual Percent Change (APC) tab quantifies the average annual rate of change in cancer incidence or mortality rates between two time periods. It is computed by comparing rates from two 3-year periods and calculating the annualized percentage change using a logarithmic transformation approach.
Formula (logarithmic approach):
Step 1: Calculate the annual slope on the log scale
γ = [log(R₂) - log(R₁)] / n = log(R₂/R₁) / n
Step 2: Transform back to percentage change
APC = 100 × [exp(γ) - 1]
where:
- R₁ = rate in period 1 (first 3-year period)
- R₂ = rate in period 2 (second 3-year period)
- n = number of years between the midpoints of the two periods
- γ = annual slope on the logarithmic scale
95% Confidence Intervals:
Confidence intervals for the APC are calculated using a log transformation approach. The standard error of the log-transformed rates is approximated using the Poisson assumption:
Confidence Interval Calculation:
Step 1: Standard error of log-transformed rates (Poisson approximation)
SE[log(R₁)] ≈ 1/√C₁
SE[log(R₂)] ≈ 1/√C₂
Step 2: Standard error of the log ratio
SE[log(R₂/R₁)] = √[SE²[log(R₁)] + SE²[log(R₂)]]
Step 3: Standard error of the annual slope γ
SE(γ) = SE[log(R₂/R₁)] / n
Step 4: 95% confidence interval on γ, then back-transform to APC
γ_lower = γ - 1.96 × SE(γ)
γ_upper = γ + 1.96 × SE(γ)
APC_lower = 100 × [exp(γ_lower) - 1]
APC_upper = 100 × [exp(γ_upper) - 1]
where C₁ and C₂ are the number of cases in periods 1 and 2, respectively.
A change is considered statistically significant if the 95% confidence interval does not include zero. In the visualisation, significant increases are shown in red, significant decreases in blue, and non-significant changes in white with black borders.
⚠ Important assumption:
The APC calculation assumes that trends are linear between the two periods analyzed. This means that the annual rate of change is assumed to be constant over the entire time interval. If the actual trend is non-linear (e.g., accelerating, decelerating, or with inflection points), the APC may not accurately represent the true pattern of change. Users should interpret APC values with caution, especially when comparing periods that are far apart or when there is evidence of non-linear trends in the data.
Reference:
Clegg LX, Hankey BF, Tiwari R, Feuer EJ, Edwards BK. Estimating average annual per cent change in trend analysis. Statistics in Medicine. 2009;28(29):3670-82. DOI:
10.1002/sim.3733
National Cancer Institute. Joinpoint Regression Program, Version 4.9.0.0 - April 2021. Statistical Research and Applications Branch, National Cancer Institute. Available at: https://surveillance.cancer.gov/joinpoint/
SEER*Stat Software. Surveillance, Epidemiology, and End Results (SEER) Program. National Cancer Institute. Available at: https://seer.cancer.gov/seerstat/
Cancer factsheets - Burden summary tab
The Burden summary page provides a concise overview of cancer burden in Geneva through key incidence and mortality indicators for each selected cancer site, considered independently from the others.
The tab is organised into four sections: disease burden (incidence, mortality and MIR), demographic profile (median age and sex distribution), age-standardised rate landscape, and age distribution based on mean annual counts.
Population data
Population data are sourced from Stat-Tab, the interactive tables database of the Swiss Federal Statistical Office (OFS). The data are extracted from the demographic balance table (
Bilan démographique selon l'âge et le canton
).
To obtain these data, the selection includes: all available years, Canton of Geneva, nationality category 'total', sex (Male, Female), all age categories, and demographic components 'Effectif au 1er janvier' (population on January 1st) and 'Effectif au 31 décembre' (population on December 31st). The population value used in the dashboard is the average of these two figures, providing a mid-year population estimate that is more representative for rate calculations than using either the beginning or end-of-year population alone.
Acknowledgements
The Registry wishes to sincerely thank all cancer patients who agree to make their data available. Thanks to their contribution, it is possible to better understand the burden of the disease and, ultimately, improve cancer prevention and care in the canton.
Partners and Collaborators
The Geneva Cancer Registry warmly thanks all its partners and collaborators for their commitment. This report is the result of their work or collaborations:
- Physicians and other cancer notifiers from hospitals, radiotherapy centers, and private practices
- Public and private pathology laboratories
- The Geneva Cancer Screening Foundation
- Other Swiss cancer registries
- The National Cancer Registration Office (NACR)
- The Wildsorf Foundation
- The Geneva League Against Cancer
- The Swiss League Against Cancer
- The Federal Statistical Office (OFS)
- The Cantonal Health Office (OCS)
- The Cantonal Population Office (OCP)
- The IT services of the University of Geneva, in particular the Information System Service and the IT Service of the Faculty of Medicine
Artificial Intelligence Training
The Geneva Cancer Registry thanks the
Graph Course team
for the high quality of the generative AI courses that contributed to the development of this web application.
Downloadable table variables
Bar chart table description
The downloadable Bar chart table contains the following variables:
-
Data_type:
type of indicator, with choices between incidence or mortality.
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Years_begin:
first year selected in the year range.
-
Years_end:
last year selected in the year range.
-
Site:
cancer site in ICD10.
-
Cases:
total number of cases in the chosen year range.
-
Annual_average_cases:
mean annual number of cases.
-
Crude_rate:
crude incidence or mortality rate per 100,000.
-
Crude_rate_LCL:
lower confidence limit for the crude incidence or mortality rate.
-
Crude_rate_UCL:
upper confidence limit for the crude incidence or mortality rate.
-
ASR:
age-standardised rate (European reference population, 1976).
-
ASR_LCL:
lower confidence limit for the age-standardised rate (European reference population, 1976).
-
ASR_UCL:
upper confidence limit for the age-standardised rate (European reference population, 1976).
Treemap table description
The downloadable Treemap table contains the following variables:
-
Data_type:
type of indicator, with choices between incidence or mortality.
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Years_begin:
first year selected in the year range.
-
Years_end:
last year selected in the year range.
-
Site:
cancer site in ICD10. Sites with fewer than 5 total cases over the selected period are merged into Others.
-
Cases:
total number of cases in the chosen year range.
-
Annual_average_cases:
mean annual number of cases in the chosen year range (informative only; confidentiality threshold uses Cases over the selected period and gender).
-
Pct:
percentage, Cases / Total_tumor_count * 100.
-
Total_tumor_count:
total number of invasive tumours in the selected year range for the selected gender.
-
Annual_average_invasive_cases:
mean annual number of invasive tumours across all sites in the chosen year range (Total_tumor_count divided by period length).
Evolution table description
The downloadable Evolution table contains the following variables:
-
Data_type:
type of indicator, with choices between incidence or mortality.
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Years_begin:
first year selected in the year range.
-
Years_end:
last year selected in the year range.
-
Site:
cancer site in ICD10.
-
Period (e.g. 1970-1972):
3-year period graphed at the midpoint of the period.
-
Cases:
total number of cases in the given 3-year period.
-
Crude_rate:
crude incidence or mortality rate per 100,000.
-
Crude_rate_LCL:
lower confidence limit for the crude incidence or mortality rate.
-
Crude_rate_UCL:
upper confidence limit for the crude incidence or mortality rate.
-
ASR:
age-standardised rate (European reference population, 1976).
-
ASR_LCL:
lower confidence limit for the age-standardised rate (European reference population, 1976).
-
ASR_UCL:
upper confidence limit for the age-standardised rate (European reference population, 1976).
Joinpoint table description
The downloadable Joinpoint table contains model-level and segment-level results with the following variables:
-
Type:
row type (Model Statistics or Segment).
-
Site:
cancer site in ICD10.
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Number of knots / Number_of_knots:
number of joinpoints in the fitted segmented model.
-
Preferred:
indicates the preferred model (lowest BIC).
-
Model status / Model_status:
model convergence status.
-
Segment:
segment index within the model.
-
Start year / Start_year:
starting year of the segment.
-
End year / End_year:
ending year of the segment.
-
Duration:
segment duration in years.
-
APC:
annual percent change for the segment.
-
SE:
standard error of APC slope estimate.
-
APC LCL / APC_LCL:
lower confidence limit for APC.
-
APC UCL / APC_UCL:
upper confidence limit for APC.
-
Slope p-value / Slope_pvalue:
p-value testing whether segment slope differs from 0.
-
BIC, BIC3, WBIC:
information criteria used for model comparison.
-
F-statistic / F_statistic and F p-value / F_pvalue:
overall model significance statistics.
Interpretation note: even when the eligibility tolerance is used (up to 3 years below 5), Joinpoint-derived indicators are shown in the table/export so the statistical trend can be reviewed alongside the curve.
APC table description
The downloadable APC table contains the following variables:
-
Cancer Site:
cancer site in ICD10.
-
Tumor Category:
tumour category code used for grouping.
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Indicator:
selected measure for APC computation (crude rate or ASR).
-
Type:
type of indicator, with choices between incidence or mortality.
-
Period 1 / Period 2:
compared 3-year periods (e.g. 1970-1972 vs 2019-2021).
-
Period 1 Value / Period 2 Value:
rate values used to compute APC for each period.
-
APC (%):
annual percent change between the two periods.
-
CI Lower (%) / CI Upper (%):
confidence interval bounds for APC.
-
Significant:
whether the APC confidence interval excludes zero.
Relative distribution table description
The downloadable Relative distribution table contains the following variables:
-
Data_type:
type of indicator, with choices between incidence or mortality.
-
Gender:
gender at diagnosis or death, with choices between both, male or female.
-
Years_begin:
first year selected in the year range.
-
Years_end:
last year selected in the year range.
-
Year:
calendar year of observation.
-
Site:
cancer site label (Top N or Others; sites with 1–4 cases in any year are merged into Others for the whole period).
-
Counts:
number of incident cases or deaths for the site in that year.
-
Pct:
percentage share of the site in the annual total (Counts / Total_tumor_count * 100).
-
Total_tumor_count:
total number of incident cases or deaths across all sites for that year.
Age distribution table description
The downloadable Age distribution table contains the following variables:
-
Years_begin:
first year selected in the year range.
-
Years_end:
last year selected in the year range.
-
Age group:
5-year age category (e.g., 0-4, 5-9, ..., 85+).
-
Gender:
gender at diagnosis, with choices between both, male or female.
-
Annual avg incident cases:
annual average number of incident cases in the selected period.
-
Annual avg deaths:
annual average number of deaths in the selected period.
-
Population:
annual average population at risk for the selected age group.
-
Incidence rate /100k:
incidence rate per 100,000 person-years.
-
Mortality rate /100k:
mortality rate per 100,000 person-years.
The downloadable Median age (heatmap) table contains the following variables:
-
Data_type:
type of indicator, with choices between incidence or mortality.
-
Gender:
gender at diagnosis or death, with choices between both, male or female.
-
Period:
5-year period displayed in the heatmap.
-
Cancer_site:
cancer site in text.
-
Median_age:
median age at diagnosis (incidence) or at death (mortality).
-
N_cases (optional):
number of cases/deaths used to estimate ages, when available in source data.
-
Annual_average_cases (optional):
mean annual number of cases/deaths over the displayed period, when N_cases is available.
-
Q1_age / Q3_age (optional):
first and third quartiles of age, when available.
-
Mean_age / SD_age (optional):
mean and standard deviation of age, when available.
-
Min_age / Max_age (optional):
minimum and maximum observed ages, when available.
Age-rate profile table description
The downloadable Age-rate profile table contains one row per cancer site and 5-year period shown in the chart, with the following variables:
-
Data_type:
incidence or mortality, according to the chart selection.
-
Gender:
male, female, or both sexes, according to the chart selection.
-
Period:
5-year period label (same windows as the chart).
-
Tumor_category:
tumour category code used for grouping.
-
Cancer_site:
cancer site label.
-
Median_age:
median age at diagnosis (incidence) or at death (mortality), years (one decimal).
-
Crude_rate:
crude rate per 100,000 person-years for that site and period (one decimal).
-
ASR:
age-standardised rate per 100,000 (ESP 1976) for that site and period (one decimal).
-
Selected_indicator:
value of the indicator currently chosen for the Y-axis (counts, crude rate, or ASR), rounded for display (one decimal).
-
Cases:
integer count of cases or deaths after multiple-primary rules, aligned with the heatmap tab.
