Determining level of significance between two median and between two catergorial variables, pre and post treatment

[Daniel Kak]

Dear Stata community,

I was wondering if anybody could help me with the following question. I am trying to figure out how to determine two things:

- Level of significance between pre op and post-op CKD (kidney function)
- Level of significance between pre op and post-op for median of eGFR and creatinine (also kidney function),

To help you understand, people either undergo a treatment RN or a treatment PN.

I hope I made myself clear, I am struggling for days now so I was hoping somebody could give me a clue. Thanks in advance!

	RN			PN
	Pre-op	1 year Post-op	p-value	Pre-op	1 year Post-op	p-value
CKD stage n (%)			???			???
1 (<15)	0	0		0	0
2 (15 – 29)	0	1		0	0
3a (30 – 44)	3	20		8	8
3b (45 – 59)	11	30		5	11
4 (60 – 89)	50	27		51	45
5 (>90)	36	3		36	12
eGFR (ml/min/1.73 m2) median [IQR]	83 (69 – 95)	57 (44 – 65)	???	79 (71 – 94)	73 (60 – 87)	???
Creatinine median [IQR]	81 (69 – 92)	107 (92 – 127)	???	84 (72 – 95)	88 (73 – 104)	???
New-onset CKD n (%)	42			11		<0.05
Negative delta-eGFR (%)	33 (23 – 40)			11 (3 – 19)		<0.05
Negative delta-eGFR median [IQR]	28 (18 – 34)			8 ( 2 – 16)		<0.05
Delta serum creatinine (%)	27 (4 – 46)			10 (-34 – 17)		<0.05
Delta serum creatinine median [IQR]	32 (22 – 41)			8 ( 1 – 16)		<0.05

[Bruce Weaver]

Hello Daniel Kak. You might find -qreg2- helpful, as it allows clustered standard errors.

ssc describe qreg2

HTH.

[Daniel Kak]

Dear Bruce Weaver,

Thanks for you quick reply! The qreg2 command is indeed perfect for comparing the two medians!
However, I am still looking for a command to compare the CKD stage pre and post-op. Hopefully you can point me in the right direction with this as well.

[Bruce Weaver]

Your table says: New onset CKD n (%), and shows only one value for each group (i.e., no pre-post). What is CKD? Is it a binary (Yes/No) variable? If not, what is it? Do you have pre-op and post-op values for both groups? Thanks for clarifying.

[Carlo Lazzaro]

Daniel:
you may want to consider interacting a binary predictor for treatment (RN;PN) with another binary predictor (Pre;Post op).

[Daniel Kak]

Bruce Weaver the New onset CKD (Chronic kidney disease) variable is defined as (eGFR pre >60 and eGFR post <60), it is a binary variable. I have pre and post values for both groups. This variable is however not the variable I want to check for, I want to compare distribution in both CKD groups before and after each treatment
Carlo Lazzaro: can you perhaps explain this?

[Carlo Lazzaro]

Daniel:
I meant something along the following lines:

qreg2 eGFR i.group##i.period

where:
1) -i-group- is a two-level categorocal variable (0=RN; 1=PN; or the other way round);
2) -i.time- is a two-level categorical variable (0=Pre_op; 1=Post_op; or the other way round);