On this page:

Cancer in Scotland: Sustaining Change: Cancer Incidence Projections for Scotland (2001-2020) - An aid to planning cancer services

« Previous | Contents | Next »

Listen

Cancer in Scotland: Sustaining Change
Cancer Incidence Projections for Scotland (2001-2020)
An Aid to Planning Cancer Services

APPENDIX A: EMPIRICAL TESTING

To test the validity of the chosen models they were fitted using data from 1961-1990 and the projected numbers of cases were then estimated for 1991-2000 using the same projection method as used for the main analysis applied to the most parsimonious model. The differences between the observed and predicted values ( see Table A1) show that overall the estimate was very close (difference 1,588 cases; 1.2% higher than was actually observed), but for individual cancer sites the accuracy of the predictions did vary considerably.

The prediction underestimated the true number of cases by more than 20% for cancers of rectum in males, melanoma in females, prostate and brain in males. Male rectal cancers and prostate cancers saw a rise in the 1990s that could not have been predicted beforehand. The models performed poorly at predicting the trends for females with melanoma - a very conservative model (predicting 1,533 cases) being chosen over a highly escalating model (2,734 cases). A similar situation was observed for males with brain cancer. Due to this, the younger age groups were modelled differently in the main analyses.

The prediction overestimated the true number of cases by more than 20% for cancers of cervix, melanoma in males, NHL, leukaemia, and other and unspecified tumours in males. The decrease in cervical cancers was a result of service re-organisation that could not have been predicted beforehand. The rate of increase in melanoma in males in the 1980s did not continue in the 1990s and this also could not have been predicted. Trends in the other and unspecified tumours are difficult to comment on because they comprise of a mixed group of tumours. The models performed poorly at predicting the trends for NHL and leukaemia. Again, as a result of this, the younger age groups were modelled differently in the main analyses.

Table A1: Accuracy of the predicted compared to observed estimates for 1996-2000; prediction based on 1961-1990 trends and known population estimates

Cancer group

Sex

Observed Cases

Minimum Difference

Model

N

%

Head and Neck

M

3,441

D2

103

3.0%

F

1,492

D1

-178

11.9%

Oesophagus

M

2,315

D1

155

6.7%

F

1,586

D1

49

3.1%

Stomach

M

2,830

D1

303

10.7%

F

1,974

ND3

-178

9.0%

Colon

M

5,605

D2

-360

6.4%

F

5,728

D1

63

1.1%

Rectum

M

3,478

D1

-717

20.6%

F

2,417

D3

-20

0.8%

Lung

M

13,442

D2

351

2.6%

F

9,878

D2

-118

1.2%

Pancreas

M

1,487

D2

55

3.7%

F

1,666

D3

-220

13.2%

Melanoma of skin

M

1,367

D2

399

29.2%

F

1,912

ND1

-379

19.8%

Breast

F

17,392

D2

310

1.8%

Cervix

F

1,709

D3

831

48.6%

Corpus uteri

F

2,194

ND1

-332

15.2%

Ovary

F

3,045

D1

84

2.8%

Prostate

M

10,061

ND1

-1,890

18.8%

Testis

M

948

ND2

-1

0.1%

Kidney

M

1,603

D1

-175

10.9%

F

1,104

D1

-124

11.3%

Bladder (invasive, insitu and uncertain behaviour)

M

5,193

D1

308

5.9%

F

2,298

D1

108

4.7%

Brain, meninges and CNS

M

878

D1

-282

32.1%

F

729

D1

-82

11.2%

Hodgkin's disease

M

340

D1

15

4.4%

F

288

D1

18

6.3%

NHL

M

1,899

ND1

430

22.6%

F

2,083

D1

428

20.6%

Leukaemia

M

1,534

D1

303

19.8%

F

1,298

D1

308

23.8%

Other and Unspecified

M

7,242

D1

1,625

22.4%

F

7,961

D1

400

5.0%

Total

130,417

1,588

1.2%

Note: Negative values denote underestimates

« Previous | Contents | Next »

Page updated: Thursday, June 9, 2005