Hello,
I'm using Stata version 15 to run some random effects meta-analyses, and have identified evidence of small study effects using Egger's test (p=0.001) (using metabias). However when attempting to use the trim and fill method using metatrim, no additional studies have been added, resulting in it having the same WMD and 95% CI as the original estimate. I've pasted the output for metan, metabias, and metatrim below.
I'm not sure if anyone has come across this issue before? I've seen it asked previously in the forum, but those cases didn't seem to have significant Egger's tests, which seemed to explain why the trim and fill method did not change anything. I'm quite new to Stata, so it's possible I could have missed something in the command.
Any advice would be greatly appreciated,
Kind regards,
Elizabeth
Meta-analysis
metan Nuts_n Nuts_mean Nuts_SD Control_n Control_mean Control_SD, nostandard random lcols (Names)
Study | WMD [95% Conf. Interval] % Weight
---------------------+---------------------------------------------------
Study 1 | -5.530 -11.964 0.904 0.01
Study 2 | -3.500 -7.590 0.590 0.01
Study 3 | -2.300 -3.780 -0.820 0.10
Study 4 | -1.290 -2.984 0.404 0.07
Study 5 | -1.000 -2.714 0.714 0.07
Study 6 | -0.750 -1.558 0.058 0.32
Study 7 | -0.687 -1.523 0.149 0.30
Study 8 | -0.600 -2.530 1.330 0.06
Study 9 | -0.500 -1.643 0.643 0.16
Study 10 | -0.400 -1.296 0.496 0.26
Study 11 | -0.342 -3.253 2.569 0.02
Study 12 | -0.305 -0.798 0.188 0.85
Study 13 | -0.300 -1.291 0.691 0.21
Study 14 | -0.200 -0.840 0.440 0.51
Study 15 | -0.180 -0.624 0.264 1.04
Study 16 | -0.150 -0.900 0.600 0.37
Study 17 | -0.104 -0.954 0.747 0.29
Study 18 | -0.100 -1.715 1.515 0.08
Study 19 | -0.100 -1.539 1.339 0.10
Study 20 | -0.100 -0.866 0.666 0.35
Study 21 | 0.000 -0.001 0.001 53.65
Study 22 | 0.010 -0.026 0.046 40.40
Study 23 | 0.050 -1.254 1.354 0.12
Study 24 | 0.100 -0.689 0.889 0.33
Study 25 | 0.500 -0.341 1.341 0.29
Study 26 | 0.600 -2.444 3.644 0.02
---------------------+---------------------------------------------------
D+L pooled WMD | -0.013 -0.058 0.033 100.00
---------------------+---------------------------------------------------
Heterogeneity chi-squared = 31.31 (d.f. = 25) p = 0.179
I-squared (variation in WMD attributable to heterogeneity) = 20.2%
Estimate of between-study variance Tau-squared = 0.0010
Test of WMD=0 : z= 0.54 p = 0.592
Egger's test
metabias _ES _seES, egger
Note: data input format theta se_theta assumed.
Egger's test for small-study effects:
Regress standard normal deviate of intervention
effect estimate against its standard error
Number of studies = 26 Root MSE = .9054
------------------------------------------------------------------------------
Std_Eff | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
slope | .0008036 .0005183 1.55 0.134 -.0002661 .0018733
bias | -.6839088 .1814884 -3.77 0.001 -1.058483 -.3093351
------------------------------------------------------------------------------
Test of H0: no small-study effects P = 0.001
Trim and fill
metatrim _ES _seES, reffect print funnel graph idvar (Names)
Note: default data input format (theta, se_theta) assumed.
Meta-analysis
| Pooled 95% CI Asymptotic No. of
Method | Est Lower Upper z_value p_value studies
-------+----------------------------------------------------
Fixed | 0.000 -0.001 0.001 0.713 0.476 26
Random | -0.013 -0.058 0.033 -0.536 0.592
Test for heterogeneity: Q= 31.313 on 25 degrees of freedom (p= 0.179)
Moment-based estimate of between studies variance = 0.001
Trimming estimator: Linear
Meta-analysis type: Random-effects model
iteration | estimate Tn # to trim diff
----------+--------------------------------------
1 | -0.013 50 0 351
2 | -0.013 50 0 0
Note: no trimming performed; data unchanged
Filled
Meta-analysis
| Pooled 95% CI Asymptotic No. of
Method | Est Lower Upper z_value p_value studies
-------+----------------------------------------------------
Fixed | 0.000 -0.001 0.001 0.713 0.476 26
Random | -0.013 -0.058 0.033 -0.536 0.592
Test for heterogeneity: Q= 31.313 on 25 degrees of freedom (p= 0.179)
Moment-based estimate of between studies variance = 0.001
| Weights Study 95% CI
Study | Fixed Random Est Lower Upper
-----------------------------------------+----------------------------------------
Study 1 | 0.09 0.09 -5.53 -11.96 0.90
Study 2 | 0.23 0.23 -3.50 -7.59 0.59
Study 3 | 1.75 1.75 -2.30 -3.78 -0.82
Study 4 | 1.34 1.34 -1.29 -2.98 0.40
Study 5 | 1.31 1.31 -1.00 -2.71 0.71
Study 6 | 5.88 5.84 -0.75 -1.56 0.06
Study 7 | 5.50 5.47 -0.69 -1.52 0.15
Study 8 | 1.03 1.03 -0.60 -2.53 1.33
Study 9 | 2.94 2.93 -0.50 -1.64 0.64
Study 10 | 4.79 4.76 -0.40 -1.30 0.50
Study 11 | 0.45 0.45 -0.34 -3.25 2.57
Study 12 | 15.80 15.55 -0.31 -0.80 0.19
Study 13 | 3.91 3.89 -0.30 -1.29 0.69
Study 14 | 9.38 9.30 -0.20 -0.84 0.44
Study 15 | 19.53 19.15 -0.18 -0.62 0.26
Study 16 | 6.83 6.78 -0.15 -0.90 0.60
Study 17 | 5.31 5.28 -0.10 -0.95 0.75
Study 18 | 6.55 6.51 -0.10 -0.87 0.67
Study 19 | 1.85 1.85 -0.10 -1.54 1.34
Study 20 | 1.47 1.47 -0.10 -1.71 1.51
Study 21 | 3.2e+06 984.21 0.00 -0.00 0.00
Study 22 | 3000.00 741.25 0.01 -0.03 0.05
Study 23 | 2.26 2.25 0.05 -1.25 1.35
Study 24 | 6.16 6.13 0.10 -0.69 0.89
Study 25 | 5.44 5.41 0.50 -0.34 1.34
Study 27 | 0.41 0.41 0.60 -2.44 3.64
I'm using Stata version 15 to run some random effects meta-analyses, and have identified evidence of small study effects using Egger's test (p=0.001) (using metabias). However when attempting to use the trim and fill method using metatrim, no additional studies have been added, resulting in it having the same WMD and 95% CI as the original estimate. I've pasted the output for metan, metabias, and metatrim below.
I'm not sure if anyone has come across this issue before? I've seen it asked previously in the forum, but those cases didn't seem to have significant Egger's tests, which seemed to explain why the trim and fill method did not change anything. I'm quite new to Stata, so it's possible I could have missed something in the command.
Any advice would be greatly appreciated,
Kind regards,
Elizabeth
Meta-analysis
metan Nuts_n Nuts_mean Nuts_SD Control_n Control_mean Control_SD, nostandard random lcols (Names)
Study | WMD [95% Conf. Interval] % Weight
---------------------+---------------------------------------------------
Study 1 | -5.530 -11.964 0.904 0.01
Study 2 | -3.500 -7.590 0.590 0.01
Study 3 | -2.300 -3.780 -0.820 0.10
Study 4 | -1.290 -2.984 0.404 0.07
Study 5 | -1.000 -2.714 0.714 0.07
Study 6 | -0.750 -1.558 0.058 0.32
Study 7 | -0.687 -1.523 0.149 0.30
Study 8 | -0.600 -2.530 1.330 0.06
Study 9 | -0.500 -1.643 0.643 0.16
Study 10 | -0.400 -1.296 0.496 0.26
Study 11 | -0.342 -3.253 2.569 0.02
Study 12 | -0.305 -0.798 0.188 0.85
Study 13 | -0.300 -1.291 0.691 0.21
Study 14 | -0.200 -0.840 0.440 0.51
Study 15 | -0.180 -0.624 0.264 1.04
Study 16 | -0.150 -0.900 0.600 0.37
Study 17 | -0.104 -0.954 0.747 0.29
Study 18 | -0.100 -1.715 1.515 0.08
Study 19 | -0.100 -1.539 1.339 0.10
Study 20 | -0.100 -0.866 0.666 0.35
Study 21 | 0.000 -0.001 0.001 53.65
Study 22 | 0.010 -0.026 0.046 40.40
Study 23 | 0.050 -1.254 1.354 0.12
Study 24 | 0.100 -0.689 0.889 0.33
Study 25 | 0.500 -0.341 1.341 0.29
Study 26 | 0.600 -2.444 3.644 0.02
---------------------+---------------------------------------------------
D+L pooled WMD | -0.013 -0.058 0.033 100.00
---------------------+---------------------------------------------------
Heterogeneity chi-squared = 31.31 (d.f. = 25) p = 0.179
I-squared (variation in WMD attributable to heterogeneity) = 20.2%
Estimate of between-study variance Tau-squared = 0.0010
Test of WMD=0 : z= 0.54 p = 0.592
Egger's test
metabias _ES _seES, egger
Note: data input format theta se_theta assumed.
Egger's test for small-study effects:
Regress standard normal deviate of intervention
effect estimate against its standard error
Number of studies = 26 Root MSE = .9054
------------------------------------------------------------------------------
Std_Eff | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
slope | .0008036 .0005183 1.55 0.134 -.0002661 .0018733
bias | -.6839088 .1814884 -3.77 0.001 -1.058483 -.3093351
------------------------------------------------------------------------------
Test of H0: no small-study effects P = 0.001
Trim and fill
metatrim _ES _seES, reffect print funnel graph idvar (Names)
Note: default data input format (theta, se_theta) assumed.
Meta-analysis
| Pooled 95% CI Asymptotic No. of
Method | Est Lower Upper z_value p_value studies
-------+----------------------------------------------------
Fixed | 0.000 -0.001 0.001 0.713 0.476 26
Random | -0.013 -0.058 0.033 -0.536 0.592
Test for heterogeneity: Q= 31.313 on 25 degrees of freedom (p= 0.179)
Moment-based estimate of between studies variance = 0.001
Trimming estimator: Linear
Meta-analysis type: Random-effects model
iteration | estimate Tn # to trim diff
----------+--------------------------------------
1 | -0.013 50 0 351
2 | -0.013 50 0 0
Note: no trimming performed; data unchanged
Filled
Meta-analysis
| Pooled 95% CI Asymptotic No. of
Method | Est Lower Upper z_value p_value studies
-------+----------------------------------------------------
Fixed | 0.000 -0.001 0.001 0.713 0.476 26
Random | -0.013 -0.058 0.033 -0.536 0.592
Test for heterogeneity: Q= 31.313 on 25 degrees of freedom (p= 0.179)
Moment-based estimate of between studies variance = 0.001
| Weights Study 95% CI
Study | Fixed Random Est Lower Upper
-----------------------------------------+----------------------------------------
Study 1 | 0.09 0.09 -5.53 -11.96 0.90
Study 2 | 0.23 0.23 -3.50 -7.59 0.59
Study 3 | 1.75 1.75 -2.30 -3.78 -0.82
Study 4 | 1.34 1.34 -1.29 -2.98 0.40
Study 5 | 1.31 1.31 -1.00 -2.71 0.71
Study 6 | 5.88 5.84 -0.75 -1.56 0.06
Study 7 | 5.50 5.47 -0.69 -1.52 0.15
Study 8 | 1.03 1.03 -0.60 -2.53 1.33
Study 9 | 2.94 2.93 -0.50 -1.64 0.64
Study 10 | 4.79 4.76 -0.40 -1.30 0.50
Study 11 | 0.45 0.45 -0.34 -3.25 2.57
Study 12 | 15.80 15.55 -0.31 -0.80 0.19
Study 13 | 3.91 3.89 -0.30 -1.29 0.69
Study 14 | 9.38 9.30 -0.20 -0.84 0.44
Study 15 | 19.53 19.15 -0.18 -0.62 0.26
Study 16 | 6.83 6.78 -0.15 -0.90 0.60
Study 17 | 5.31 5.28 -0.10 -0.95 0.75
Study 18 | 6.55 6.51 -0.10 -0.87 0.67
Study 19 | 1.85 1.85 -0.10 -1.54 1.34
Study 20 | 1.47 1.47 -0.10 -1.71 1.51
Study 21 | 3.2e+06 984.21 0.00 -0.00 0.00
Study 22 | 3000.00 741.25 0.01 -0.03 0.05
Study 23 | 2.26 2.25 0.05 -1.25 1.35
Study 24 | 6.16 6.13 0.10 -0.69 0.89
Study 25 | 5.44 5.41 0.50 -0.34 1.34
Study 27 | 0.41 0.41 0.60 -2.44 3.64
