Question regarding confidence intervals on a restricted cubic spline model.

[lei]

Hi everyone,

I have a question regarding the restricted cubic spline model shown in the picture here. This is taken from the paper. I am wondering why the confidence interval around 6 looks like that. Can someone explain it in really simple terms? I have no clue how to explain this.

This is what is said under the caption. This diagram is for the paper: http://eurheartj.oxfordjournals.org/content/35/27/1817

Figure 1

Adjusted restricted cubic spline model showing the hazard ratios (95% confidence intervals) for the association of high-sensitivity cardiac troponin T with incident dementia [among the 6321 participants with high-sensitivity cardiac troponin T concentrations within the measuring range (≥3 ng/L)]. Knots at 4, 6, and 14 ng/L (10th, 50th, and 90th percentiles). Histogram shows the distribution of concentrations of high-sensitivity cardiac troponin T within the measuring range. Data truncated at 97.5 percentile of high-sensitivity cardiac troponin T. Model is adjusted for age (continuous), race/field centre (Washington County, Maryland whites; suburbs of Minneapolis, Minnesota whites; Forsyth County, North Carolina whites; Forsyth County, North Carolina blacks; Jackson, Mississippi blacks), sex (male; female), education level (<high school; high school or equivalent; college, graduate or professional school), income (<$35 000/year; ≥$35 000/year; not reported), body mass index (continuous), physical activity (continuous), alcohol consumption (current high; current low; former; never), smoking (current; former; never), total cholesterol (continuous), HDL cholesterol (continuous), diabetes (yes; no), systolic and diastolic blood pressure (continuous), and blood pressure-lowering medication (yes; no).

Thank you.

[Doc_Duke]

The analysis was done in Stata and the graph looks like it may have come directly from the program, so the Stata documentation may give you some insight.

Otherwise, contact the corresponding author and see if you can get further information that way.

[lei]

Hi,

You can also see an example from SAS here http://support.sas.com/resources/papers/proceedings09/252-2009.pdf

therefore it has nothing to do whether it's STATA or SAS.

[JacobSimonsen]

I see no other explanation than the Hazard Ratios is calculated with "6" as a reference value. Therefore the HR at 6 is 1, and the confidence intervals at 6 is just a point [1-1] (or undefined).

[lei]

Thank you. I think that explains it.