Thread by @lung_ca_screen

#NELSON #lungcancerscreening trial overview & comments:

Click here for the full publication: nejm.org/doi/full/10.10…

and editorial:
nejm.org/doi/full/10.10…

Randomized trial 15,792 individuals (13,195 males for 1ry analysis).
Lower risk cohort vs. NLST
-Younger lower age limit 50 vs. 55 in NLST (median 58 years vs. 61.4)
-lower pack-years 15-19 py (38py vs. 48 py in NLST)
But shorter quit interval <10 years (vs. 15 in NLST)

1:1 randomization #LDCT vs. no screening (unlike CXR in #NLST)
4 rounds of LDCT: baseline, yrs 1, 3 & 5.5 (over 2x screening window @ 66 months vs. NLST’s 3 rounds at yr 0, 1, 2 / 24 month period).
Volumetric nodule measurement protocols
10 year f/y (vs. 6.5y in NLST

1ry outcome:
Cumulative rate ratio of 0.76 (95% CI, 0.61 to 0.94; P = 0.01)
Number needed to screen is 125 to avoid 1 lung cancer death.
Point estimates for NLST ineligible participants (0.82) and men 50-54 (0.85) less than 1 nice to see even if CI’s cross unity.

Mortality reduction in women (2ry analysis) shows lower point estimate for death from lung cancer of 0.67 but small sample size / wider confidence intervals does not lead to statistical significance (95% CI, 0.38 to 1.14). Trend consistent across studies including #NLST and #LUSI

48.8% of lung cancers in screening group were stage I/II vs. 23.4% in controls. This is significantly lower than in NLST which found 70.2% stage I/II in LDCT arm. Could this be a result of the longer screening intervals?

Lung cancer is a major cause of mortally in this cohort: 18.4% and 24.4% of all deaths in screening/control arms. All-cause mortality not reduced. Not powered to show this, but a “trend” would have been comforting to see.

Downsides of screening:
9.2% of scans lead to a repeat CT / early recall. This was significantly higher after baseline (19.7%) vs follow-up exams (range 1.9-6.7%). Seems acceptable, but could reduce further with additional refinements to interpretation and management protocols?

Positive scans requiring clinical evaluation in only 2.1%, 43.5% of which led to a lung cancer diagnosis (positive predictive value) suggesting that few individuals need anything more than early follow-up LDCTs for indeterminate lung nodules under the NELSON volumetric protocol.

Overall 1.2% of participants had a false positive test (defined as requiring a clinical evaluation) over the length of the study.

Overdiagnosis: 40 excess cases of lung cancer at 10 years of 18.5% (344 vs. 304), but with ongoing reduction in this excess over time (down to 18 cases / 8.9% by year 11). Unclear how newer management algorithms for managing growing GGOs would affect these rates.

Some criticism exists about overall study management and implementation. (see: ntvg.nl/artikelen/mach…), but not unexpected in trial designed, implemented and analyzed over a 20 yr period. Results not interpreted in isolation, but in addition to other trials published to date.

Study adds to the evidence that #LDCT #lungcancerscreening reduces lung cancer mortality, a fact that should now be considered as confirmed. Meta-analysis of all trials to date to shed more light on all-cause mortality benefit seen in the #NLST &10 year follow-up of #MILD trial.

Longer screening window of 5.5 years may have lead to the greater lung cancer mortality reduction despite a lower risk cohort than NLST, but the relatively low rate of early stage cancers leaves me concerned regarding the longer screen intervals.

#MILD with both longer screening window and annual screening seems to have best results (HR 0.61 at 10 years).

Focus moving forward is to implement #lungcancerscreening internationally and improve screening rates where programs have already been initiated (USA).

Of course many questions remain about optimal subgroups to be screened, intervals, optimal management protocols and cost-effectiveness. But as further knowledge and experience is gained in these areas, outcomes should only improve from what is now a new benchmark.

@threadreaderapp unroll