Table 1.
Summary of cohort studies included in the review
| Characteristics | Number of studies (N = 82) | (%) |
|---|---|---|
|
Publication year | ||
| 2000 |
2 |
(2) |
| 2001 |
4 |
(5) |
| 2002 |
4 |
(5) |
| 2003 |
5 |
(6) |
| 2004 |
6 |
(7) |
| 2005 |
8 |
(10) |
| 2006 |
11 |
(13) |
| 2007 |
13 |
(16) |
| 2008 |
13 |
(16) |
| 2009 |
16 |
(20) |
|
Number of participants recruited at baseline | ||
| Not stated in paper |
4 |
(5) |
| 1,000–2,000 |
15 |
(18) |
| 2,000–3,000 |
13 |
(16) |
| 3,000–5,000 |
14 |
(17) |
| 5,000–10,000 |
13 |
(16) |
| 10,000–20,000 |
14 |
(17) |
| 20,000 + |
9 |
(11) |
|
Date of baseline recruitment (decade recruitment started) | ||
| Not stated in paper |
5 |
(6) |
| Before 1970 |
9 |
(11) |
| 1970–1979 |
9 |
(11) |
| 1980–1989 |
25 |
(30) |
| 1990–1999 |
30 |
(37) |
| 2000–2009 |
4 |
(5) |
|
Number of follow-up waves used in the analysis | ||
| Number not stated |
12 |
(15) |
| Mean or range given |
7 |
(9) |
| 1 |
10 |
(12) |
| 2 |
9 |
(11) |
| 3 |
18 |
(22) |
| 4 |
12 |
(15) |
| ≥5 |
14 |
(17) |
|
Statistical methods for epidemiological analysis† | ||
| Cox proportional hazards regression^ |
37‡ |
|
| Time-varying covariates |
35 |
|
| Time-invariant covariates |
3 |
|
| Generalised Estimating Equations |
12* |
|
| Linear regression |
3 |
|
| Logistic regression |
10 |
|
| Generalised linear mixed-effects modelling |
16 |
|
| Linear regression |
13 |
|
| Logistic regression |
3 |
|
| Standard linear regression |
3 |
|
| Standard logistic regression |
9 |
|
| Other methods | 6 | |
† The total number of papers is 83 because one paper used two analysis models: a Cox proportional hazards model for a time to event outcome and a linear mixed effects model for a numerical outcome.
^ Note one paper used a parametric survival model.
‡ One paper incorporated their repeated measures of exposures using both a time varying covariate and time-invariant covariates.
*One paper had a numerical and binary outcome.